Merge remote-tracking branch 'origin/main' into update-folly-11.6

Summary:
- Bump version.h from 11.6.0 to 11.7.0
- Add `11.6.fb` to `check_format_compatible.sh`
- Sync HISTORY.md `## 11.6.` section from `11.6.fb`
- Delete 5 consumed `unreleased_history/` note file(s) from main

Part of 11.6 release workflow.

Pull Request resolved: https://github.com/facebook/rocksdb/pull/14913

Reviewed By: anand1976

Differential Revision: D110520723

Pulled By: pdillinger

fbshipit-source-id: 20f2f9b4be03c08d78d6f2c4e6b70812a408d9f0

.github/actions/build-folly/action.yml

@@ -9,7 +9,16 @@ runs:
   steps:
   - name: Build folly and dependencies
     if: ${{ inputs.cache-hit != 'true' }}
-    run: make build_folly
+    run: |
+      clean_path=()
+      IFS=: read -ra path_entries <<< "$PATH"
+      for entry in "${path_entries[@]}"; do
+        if [[ "$entry" != "/usr/lib/ccache" ]]; then
+          clean_path+=("$entry")
+        fi
+      done
+      export PATH="$(IFS=:; echo "${clean_path[*]}")"
+      make build_folly
     shell: bash
   - name: Skip folly build (using cached version)
     if: ${{ inputs.cache-hit == 'true' }}

.github/actions/install-jdk8-on-macos/action.yml

@@ -1,9 +0,0 @@
-name: install-jdk8-on-macos
-runs:
-  using: composite
-  steps:
-  - name: Install JDK 8 on macos
-    run: |-
-      HOMEBREW_NO_AUTO_UPDATE=1 brew tap bell-sw/liberica
-      HOMEBREW_NO_AUTO_UPDATE=1 brew install --cask liberica-jdk8
-    shell: bash

.github/actions/pre-steps/action.yml

@@ -2,6 +2,13 @@ name: pre-steps
 runs:
   using: composite
   steps:
+  - name: Dump ulimits for diagnostics
+    run: |
+      echo "=== ulimits (soft) ==="
+      ulimit -a -S || true
+      echo "=== ulimits (hard) ==="
+      ulimit -a -H || true
+    shell: bash
   - name: Install lld linker for faster builds
     run: apt-get update -y && apt-get install -y lld 2>/dev/null || true
     shell: bash

.github/actions/setup-jdk-on-macos/action.yml

@@ -0,0 +1,17 @@
+name: setup-jdk-on-macos
+runs:
+  using: composite
+  steps:
+  - name: Set up JDK on macos
+    run: |-
+      set -euo pipefail
+
+      JAVA_HOME="$(/usr/libexec/java_home -v 21)"
+
+      echo "JAVA_HOME=${JAVA_HOME}" >> "$GITHUB_ENV"
+      echo "JAVAC_ARGS=--release 8" >> "$GITHUB_ENV"
+      echo "${JAVA_HOME}/bin" >> "$GITHUB_PATH"
+      echo "Using JDK at ${JAVA_HOME}"
+      "${JAVA_HOME}/bin/java" -version
+      "${JAVA_HOME}/bin/javac" -version
+    shell: bash

.github/actions/teardown-ccache/action.yml

@@ -5,7 +5,11 @@ runs:
   steps:
     - name: Trim and print ccache stats
       run: |
+        if [ -z "$CCACHE_DIR" ]; then
+          echo "teardown-ccache: CCACHE_DIR not set, skipping (setup-ccache may not have run)"
+          exit 0
+        fi
         .github/scripts/ccache-trim.sh || true
-        ccache -s
+        ccache -s || echo "teardown-ccache: ccache not found, skipping stats"
       if: always()
       shell: bash

.github/actions/windows-build-steps/action.yml

@@ -3,7 +3,7 @@ inputs:
   suite-run:
     description: Comma-separated list of test suites to run (empty to skip C++ tests)
     required: false
-    default: arena_test,db_basic_test,db_test,db_test2,db_merge_operand_test,bloom_test,c_test,coding_test,crc32c_test,dynamic_bloom_test,env_basic_test,env_test,hash_test,random_test
+    default: arena_test,db_basic_test,db_test,db_etc2_test,db_merge_operand_test,bloom_test,c_test,coding_test,crc32c_test,dynamic_bloom_test,env_basic_test,env_test,hash_test,random_test
   run-java:
     description: Whether to run Java tests
     required: false
@@ -11,32 +11,85 @@ inputs:
 runs:
   using: composite
   steps:
+  - name: Detect Visual Studio generator
+    id: detect_vs
+    shell: pwsh
+    run: |
+      $ErrorActionPreference = "Stop"
+
+      $vswhere = Join-Path ([Environment]::GetFolderPath("ProgramFilesX86")) "Microsoft Visual Studio\Installer\vswhere.exe"
+      if (!(Test-Path $vswhere)) {
+        throw "vswhere.exe was not found at $vswhere"
+      }
+
+      $vsInstallationsJson = & $vswhere -latest -products * -requires Microsoft.VisualStudio.Component.VC.Tools.x86.x64 -format json -utf8
+      $vsInstallations = @($vsInstallationsJson | ConvertFrom-Json)
+      if ($vsInstallations.Count -eq 0) {
+        throw "No Visual Studio instance with C++ tools was found."
+      }
+
+      $vsInstallation = $vsInstallations[0]
+      $vsMajor = [int]$vsInstallation.catalog.productLineVersion
+      if ($vsMajor -eq 0) {
+        $vsMajor = [int]($vsInstallation.installationVersion.Split(".")[0])
+      }
+
+      $generatorLine = cmake --help | Where-Object {
+        $_ -match "^\*?\s*Visual Studio $vsMajor\s+\d{4}\s*="
+      } | Select-Object -First 1
+      if (!$generatorLine) {
+        throw "CMake does not support a Visual Studio generator for VS major version $vsMajor."
+      }
+
+      $cmakeGenerator = ($generatorLine -replace "^\*?\s*", "" -replace "\s*=.*$", "").Trim()
+      $vsVersionRange = "[{0}.0,{1}.0)" -f $vsMajor, ($vsMajor + 1)
+
+      echo "Detected Visual Studio at $($vsInstallation.installationPath)"
+      echo "Using CMake generator: $cmakeGenerator"
+      echo "Using setup-msbuild vs-version: $vsVersionRange"
+
+      "CMAKE_GENERATOR=$cmakeGenerator" >> $env:GITHUB_ENV
+      "vs_version=$vsVersionRange" >> $env:GITHUB_OUTPUT
   - name: Add msbuild to PATH
-    uses: microsoft/setup-msbuild@v1.3.1
+    uses: microsoft/setup-msbuild@v2
+    with:
+      vs-version: ${{ steps.detect_vs.outputs.vs_version }}
   - name: ccache
     uses: hendrikmuhs/ccache-action@v1.2
     with:
       max-size: "10GB"
+      key: ccache-windows-${{ github.workflow }}
+      restore-keys: |
+        ccache-windows-
   - name: Configure ccache
     shell: pwsh
     run: |
       ccache --set-config=base_dir=C:\a\rocksdb\rocksdb
       ccache --set-config=hash_dir=false
+      ccache --set-config=compiler_check=content
   - name: Custom steps
     env:
       THIRDPARTY_HOME: ${{ github.workspace }}/thirdparty
       CMAKE_HOME: C:/Program Files/CMake
       CMAKE_BIN: C:/Program Files/CMake/bin/cmake.exe
       CTEST_BIN: C:/Program Files/CMake/bin/ctest.exe
-      JAVA_HOME: C:/Program Files/BellSoft/LibericaJDK-8
       SNAPPY_HOME: ${{ github.workspace }}/thirdparty/snappy-1.2.2
       SNAPPY_INCLUDE: ${{ github.workspace }}/thirdparty/snappy-1.2.2;${{ github.workspace }}/thirdparty/snappy-1.2.2/build
       SNAPPY_LIB_DEBUG: ${{ github.workspace }}/thirdparty/snappy-1.2.2/build/Debug/snappy.lib
     run: |-
       # NOTE: if ... Exit $LASTEXITCODE lines needed to exit and report failure
       echo ===================== Install Dependencies =====================
-      choco install liberica8jdk -y
-      if(!$?) { Exit $LASTEXITCODE }
+      if (!$env:JAVA_HOME -or !(Test-Path (Join-Path $env:JAVA_HOME "bin\javac.exe"))) {
+        $javac = Get-Command javac -ErrorAction SilentlyContinue
+        if (!$javac) {
+          Write-Error "javac was not found on PATH and JAVA_HOME is not set."
+          Exit 1
+        }
+        $env:JAVA_HOME = Split-Path (Split-Path $javac.Source -Parent) -Parent
+      }
+      echo "JAVA_HOME=$env:JAVA_HOME"
+      & "$env:JAVA_HOME\bin\java.exe" -version
+      & "$env:JAVA_HOME\bin\javac.exe" -version
       mkdir $Env:THIRDPARTY_HOME
       cd $Env:THIRDPARTY_HOME
       echo "Building Snappy dependency..."
@@ -49,11 +102,11 @@ runs:
       cd build
       & cmake -G "$Env:CMAKE_GENERATOR" .. -DSNAPPY_BUILD_TESTS=OFF -DSNAPPY_BUILD_BENCHMARKS=OFF
       if(!$?) { Exit $LASTEXITCODE }
-      msbuild Snappy.sln -maxCpuCount -property:Configuration=Debug -property:Platform=x64
+      cmake --build . --config Debug --parallel 32 -- /p:Platform=x64
       if(!$?) { Exit $LASTEXITCODE }
       echo ======================== Build RocksDB =========================
       cd ${{ github.workspace }}
-      $env:Path = $env:JAVA_HOME + ";" + $env:Path
+      $env:Path = "$env:JAVA_HOME\bin;" + $env:Path
       mkdir build
       cd build
       & cmake -G "$Env:CMAKE_GENERATOR" -DCMAKE_BUILD_TYPE=Debug -DWIN_CI=1 -DPORTABLE="$Env:CMAKE_PORTABLE" -DSNAPPY=1 -DXPRESS=1 -DJNI=1 ..
@@ -61,7 +114,7 @@ runs:
       cd ..
       echo "Building with VS version: $Env:CMAKE_GENERATOR"
       # use more parallel processes than the number of processes available, as most of the compile command would be cache hit
-      msbuild build/rocksdb.sln /m:32 /p:LinkIncremental=false -property:Configuration=Debug -property:Platform=x64
+      cmake --build build --config Debug --parallel 32 -- /p:LinkIncremental=false /p:Platform=x64
       if(!$?) { Exit $LASTEXITCODE }
       echo ========================= Test RocksDB =========================
       $suiteRun = "${{ inputs.suite-run }}"

.github/scripts/build-ai-review-comment.js

@@ -0,0 +1,27 @@
+// Shared markdown builder for AI review comment bodies.
+//
+// Usage:
+//   const build = require('./build-ai-review-comment.js');
+//   return build({ icon, headerTitle, triggerLine, responseBody, footerLines
+//   });
+
+module.exports = function buildAiReviewComment(
+    {icon, headerTitle, triggerLine, responseBody, footerLines}) {
+  return [
+    `## ${icon} ${headerTitle}`,
+    '',
+    triggerLine,
+    '',
+    '---',
+    '',
+    responseBody,
+    '',
+    '---',
+    '',
+    '<details>',
+    '<summary>ℹ️ About this response</summary>',
+    '',
+    ...footerLines,
+    '</details>',
+  ].join('\n');
+};

.github/scripts/parse-claude-review.js

@@ -7,11 +7,25 @@
 // Parameters:
 //   executionFile - path to the JSON execution log from claude-code-base-action
 //   conclusion    - 'success' or 'failure' from the action output
-//   meta          - { trigger, headSha, reviewer, isQuery }
+//   meta          - { trigger, autoMode, headSha, reviewer, isQuery, isPartial
+//   }
 
 const fs = require('fs');
+const buildComment = require('./build-ai-review-comment.js');
 
 module.exports = function parseClaude({executionFile, conclusion, meta}) {
+  function getTriggerLine() {
+    if (meta.trigger !== 'auto') {
+      return `*Requested by @${meta.reviewer}*`;
+    }
+    const shortSha = meta.headSha ? meta.headSha.substring(0, 7) : 'unknown';
+    if (meta.autoMode === 'early') {
+      return `*Auto-triggered after CI reached the early-review threshold — reviewing commit ${
+          shortSha}*`;
+    }
+    return `*Auto-triggered after CI passed — reviewing commit ${shortSha}*`;
+  }
+
   let responseBody = '';
 
   try {
@@ -23,6 +37,29 @@ module.exports = function parseClaude({executionFile, conclusion, meta}) {
 
     const resultMessage = executionLog.find(m => m.type === 'result');
 
+    // Helper: extract the last substantial assistant text from the log.
+    // Used as a fallback when Claude ran out of turns and the recovery
+    // session also failed to produce a result.
+    function getLastAssistantText(log, minLength = 200) {
+      for (let i = log.length - 1; i >= 0; i--) {
+        const m = log[i];
+        if (m.type !== 'assistant') continue;
+        const content = m.message && m.message.content;
+        if (!Array.isArray(content)) continue;
+        for (let j = content.length - 1; j >= 0; j--) {
+          if (content[j].type === 'text' && content[j].text &&
+              content[j].text.trim().length >= minLength) {
+            const text = content[j].text.trim();
+            // Truncate to avoid enormous PR comments
+            return text.length > 50000 ? text.substring(0, 50000) +
+                    '\n\n*[Truncated — full output in execution log artifact]*' :
+                                         text;
+          }
+        }
+      }
+      return null;
+    }
+
     if (!resultMessage) {
       responseBody = '⚠️ No result message found in execution log.';
     } else if (resultMessage.subtype === 'success' && resultMessage.result) {
@@ -31,6 +68,19 @@ module.exports = function parseClaude({executionFile, conclusion, meta}) {
       const errorInfo =
           resultMessage.result || resultMessage.error || 'Unknown error';
       responseBody = `❌ **Claude encountered an error:**\n\n${errorInfo}`;
+    } else if (resultMessage.subtype === 'error_max_turns') {
+      // The workflow runs a recovery session when this happens, so this
+      // branch is typically only hit if recovery wasn't attempted (e.g.,
+      // no findings file was written). Extract what we can.
+      const partial = getLastAssistantText(executionLog);
+      if (partial) {
+        responseBody =
+            `⚠️ **Review incomplete — Claude hit the turn limit.**\n\nBelow is the last partial output. You can request a fresh review with \`/claude-review\`.\n\n---\n\n${
+                partial}`;
+      } else {
+        responseBody =
+            '⚠️ **Review incomplete — Claude hit the turn limit before producing output.** You can request a fresh review with `/claude-review`.';
+      }
     } else if (resultMessage.result) {
       responseBody = `⚠️ **Completed with status: ${
           resultMessage.subtype}**\n\n${resultMessage.result}`;
@@ -41,38 +91,28 @@ module.exports = function parseClaude({executionFile, conclusion, meta}) {
     responseBody = `❌ Error parsing Claude response: ${error.message}`;
   }
 
-  const icon = conclusion === 'success' ? '✅' : '⚠️';
+  const isPartial = !!meta.isPartial;
+  const icon = isPartial ? '🟡' : (conclusion === 'success' ? '✅' : '⚠️');
   const headerTitle = meta.isQuery ? 'Claude Response' : 'Claude Code Review';
-  const triggerLine = meta.trigger === 'auto' ?
-      `*Auto-triggered after CI passed — reviewing commit ${
-          meta.headSha.substring(0, 7)}*` :
-      `*Requested by @${meta.reviewer}*`;
+  const triggerLine = getTriggerLine();
 
-  return [
-    `## ${icon} ${headerTitle}`,
-    '',
+  return buildComment({
+    icon,
+    headerTitle,
     triggerLine,
-    '',
-    '---',
-    '',
     responseBody,
-    '',
-    '---',
-    '',
-    '<details>',
-    '<summary>ℹ️ About this response</summary>',
-    '',
-    'Generated by [Claude Code](https://github.com/anthropics/claude-code-action).',
-    'Review methodology: `claude_md/code_review.md`',
-    '',
-    '**Limitations:**',
-    '- Claude may miss context from files not in the diff',
-    '- Large PRs may be truncated',
-    '- Always apply human judgment to AI suggestions',
-    '',
-    '**Commands:**',
-    '- `/claude-review [context]` — Request a code review',
-    '- `/claude-query <question>` — Ask about the PR or codebase',
-    '</details>',
-  ].join('\n');
+    footerLines: [
+      'Generated by [Claude Code](https://github.com/anthropics/claude-code-action).',
+      'Review methodology: `claude_md/code_review.md`',
+      '',
+      '**Limitations:**',
+      '- Claude may miss context from files not in the diff',
+      '- Large PRs may be truncated',
+      '- Always apply human judgment to AI suggestions',
+      '',
+      '**Commands:**',
+      '- `/claude-review [context]` — Request a code review',
+      '- `/claude-query <question>` — Ask about the PR or codebase',
+    ],
+  });
 };

.github/scripts/parse-codex-review.js

@@ -0,0 +1,98 @@
+// Parse Codex review artifacts and produce a markdown review comment.
+//
+// Usage from actions/github-script:
+//   const parse = require('./.github/scripts/parse-codex-review.js');
+//   const markdown = parse({ responseFile, recoveryFile, findingsFile, logFile,
+//   exitCode, meta });
+//
+// Parameters:
+//   responseFile - path to Codex final response output
+//   recoveryFile - path to recovery output formatted from review-findings.md
+//   findingsFile - path to incremental findings file written during review
+//   logFile      - path to Codex stdout/stderr log
+//   exitCode     - Codex process exit code
+//   meta         - { trigger, autoMode, headSha, reviewer, isQuery, isPartial }
+
+const fs = require('fs');
+const buildComment = require('./build-ai-review-comment.js');
+
+module.exports = function parseCodex(
+    {responseFile, recoveryFile, findingsFile, logFile, exitCode, meta}) {
+  function getTriggerLine() {
+    if (meta.trigger !== 'auto') {
+      return `*Requested by @${meta.reviewer}*`;
+    }
+    const shortSha = meta.headSha ? meta.headSha.substring(0, 7) : 'unknown';
+    if (meta.autoMode === 'early') {
+      return `*Auto-triggered after CI reached the early-review threshold — reviewing commit ${
+          shortSha}*`;
+    }
+    return `*Auto-triggered after CI passed — reviewing commit ${shortSha}*`;
+  }
+
+  function readIfPresent(path) {
+    if (!path || !fs.existsSync(path)) {
+      return '';
+    }
+    const text = fs.readFileSync(path, 'utf8').trim();
+    return text;
+  }
+
+  function tailFile(path, maxChars = 12000) {
+    const text = readIfPresent(path);
+    if (!text) {
+      return '';
+    }
+    if (text.length <= maxChars) {
+      return text;
+    }
+    return text.slice(text.length - maxChars);
+  }
+
+  let responseBody = '';
+  const recovered = readIfPresent(recoveryFile);
+  const direct = readIfPresent(responseFile);
+  const findings = readIfPresent(findingsFile);
+
+  if (recovered) {
+    responseBody = recovered;
+  } else if (direct) {
+    responseBody = direct;
+  } else if (findings) {
+    responseBody =
+        '⚠️ **Review incomplete — Codex did not produce a final response.**\n\n' +
+        'Below are the incremental findings recovered from `review-findings.md`.\n\n---\n\n' +
+        findings;
+  } else {
+    const logTail = tailFile(logFile);
+    responseBody = logTail ?
+        `❌ **Codex review failed before producing findings.**\n\n\`\`\`\n${
+            logTail}\n\`\`\`` :
+        '❌ Codex review failed before producing any output.';
+  }
+
+  const isPartial = !!meta.isPartial;
+  const code = Number.parseInt(exitCode || '1', 10);
+  const icon = isPartial ? '🟡' : (code === 0 ? '✅' : '⚠️');
+  const triggerLine = getTriggerLine();
+
+  return buildComment({
+    icon,
+    headerTitle: meta.isQuery ? 'Codex Response' : 'Codex Code Review',
+    triggerLine,
+    responseBody,
+    footerLines: [
+      'Generated by Codex CLI.',
+      'Review methodology: `claude_md/code_review.md`',
+      '',
+      '**Limitations:**',
+      '- Codex may miss context from files not in the diff',
+      '- Large PRs may be truncated',
+      '- Always apply human judgment to AI suggestions',
+      '',
+      '**Commands:**',
+      '- `/codex-review [context]` — Request a code review',
+      '- `/codex-query <question>` — Ask about the PR or codebase',
+    ],
+  });
+};

.github/scripts/post-pr-comment.js

@@ -14,9 +14,31 @@
 //   marker   - HTML comment marker for dedup (e.g. '<!-- claude-review -->')
 //              If an existing comment with this marker is found, it is updated.
 //              If not found, a new comment is created.
+//   legacyMarkers - optional list of legacy markers that should be migrated by
+//              being considered part of the same comment family.
+//   prunePrefix - optional marker prefix whose older comments should be
+//                 superseded. If obsoleteTitle is set, older comments are
+//                 collapsed instead of deleted.
+//   preserveLatest - optional count of active comments to keep when
+//                    prunePrefix is set.
+//   obsoleteMarker - optional HTML marker used to detect already-obsolete
+//                    comments.
+//   obsoleteTitle - optional heading to use when collapsing superseded
+//                   comments into a details block.
 
-module.exports = async function postPrComment(
-    {github, context, core, prNumber, body, marker}) {
+module.exports = async function postPrComment({
+  github,
+  context,
+  core,
+  prNumber,
+  body,
+  marker,
+  legacyMarkers = [],
+  prunePrefix = '',
+  preserveLatest = 0,
+  obsoleteMarker = '',
+  obsoleteTitle = '',
+}) {
   if (!prNumber || !body) {
     core.warning('Missing prNumber or body; skipping comment.');
     return;
@@ -28,30 +50,155 @@ module.exports = async function postPrComment(
   // Ensure marker is embedded in the body
   const markedBody = body.includes(marker) ? body : `${marker}\n${body}`;
 
-  // Search for existing comment with this marker
-  const {data: comments} = await github.rest.issues.listComments({
-    owner,
-    repo,
-    issue_number: prNumber,
-    per_page: 100,
-  });
-  const existing = comments.find(c => c.body.includes(marker));
-
-  if (existing) {
-    await github.rest.issues.updateComment({
+  async function listComments() {
+    return await github.paginate(github.rest.issues.listComments, {
       owner,
       repo,
-      comment_id: existing.id,
-      body: markedBody,
+      issue_number: prNumber,
+      per_page: 100,
     });
-    core.info(`Updated existing comment ${existing.id}`);
-  } else {
-    await github.rest.issues.createComment({
+  }
+
+  function commentActivityTime(comment) {
+    const timestamp =
+        Date.parse(comment.updated_at || comment.created_at || '');
+    return Number.isNaN(timestamp) ? 0 : timestamp;
+  }
+
+  function commentSortDescending(left, right) {
+    const timeDelta = commentActivityTime(right) - commentActivityTime(left);
+    if (timeDelta !== 0) {
+      return timeDelta;
+    }
+    return Number(right.id || 0) - Number(left.id || 0);
+  }
+
+  function isObsoleteComment(comment) {
+    return !!obsoleteMarker && typeof comment.body === 'string' &&
+        comment.body.includes(obsoleteMarker);
+  }
+
+  function buildObsoleteBody(comment) {
+    const originalBody =
+        typeof comment.body === 'string' && comment.body.trim() ?
+        comment.body :
+        '*No original review body preserved.*';
+    const title = obsoleteTitle || 'AI Review - OBSOLETE';
+    return `${obsoleteMarker}\n## ${
+        title}\n\n<details>\n<summary>Superseded by a newer AI review. Expand to see the original review.</summary>\n\n${
+        originalBody}\n\n</details>`;
+  }
+
+  async function deleteCommentIfPresent(comment, reason) {
+    try {
+      await github.rest.issues.deleteComment({
+        owner,
+        repo,
+        comment_id: comment.id,
+      });
+      core.info(`${reason} ${comment.id}`);
+    } catch (error) {
+      if (error.status === 404) {
+        core.info(`Comment ${comment.id} was already deleted by another run.`);
+        return;
+      }
+      throw error;
+    }
+  }
+
+  async function supersedeCommentIfPresent(comment, reason) {
+    if (!obsoleteTitle) {
+      await deleteCommentIfPresent(comment, reason);
+      return;
+    }
+    if (isObsoleteComment(comment)) {
+      return;
+    }
+    try {
+      await github.rest.issues.updateComment({
+        owner,
+        repo,
+        comment_id: comment.id,
+        body: buildObsoleteBody(comment),
+      });
+      core.info(`${reason} ${comment.id}`);
+    } catch (error) {
+      if (error.status === 404) {
+        core.info(`Comment ${comment.id} was already deleted by another run.`);
+        return;
+      }
+      throw error;
+    }
+  }
+
+  let comments = await listComments();
+  const existing = comments.find(
+      comment =>
+          typeof comment.body === 'string' && comment.body.includes(marker));
+  let currentCommentId = null;
+
+  if (existing) {
+    try {
+      const response = await github.rest.issues.updateComment({
+        owner,
+        repo,
+        comment_id: existing.id,
+        body: markedBody,
+      });
+      currentCommentId = existing.id;
+      core.info(`Updated existing comment ${existing.id}`);
+    } catch (error) {
+      if (error.status !== 404) {
+        throw error;
+      }
+      core.info(
+          `Comment ${existing.id} disappeared before update; ` +
+          'creating a fresh comment instead.');
+    }
+  }
+
+  if (!currentCommentId) {
+    const response = await github.rest.issues.createComment({
       owner,
       repo,
       issue_number: prNumber,
       body: markedBody,
     });
+    currentCommentId = response.data.id;
     core.info('Created new PR comment');
   }
+
+  if (prunePrefix || legacyMarkers.length > 0) {
+    comments = await listComments();
+    const relatedComments = comments.filter(
+        comment => comment.id !== currentCommentId &&
+            typeof comment.body === 'string' &&
+            ((prunePrefix && comment.body.includes(prunePrefix)) ||
+             legacyMarkers.some(
+                 legacyMarker => comment.body.includes(legacyMarker))));
+    const activeRelatedComments =
+        comments
+            .filter(
+                comment => typeof comment.body === 'string' &&
+                    !isObsoleteComment(comment) &&
+                    (comment.id === currentCommentId ||
+                     relatedComments.some(
+                         relatedComment => relatedComment.id === comment.id)))
+            .sort(commentSortDescending);
+
+    const keep =
+        new Set(activeRelatedComments.slice(0, Math.max(preserveLatest, 1))
+                    .map(comment => comment.id));
+
+    const supersedeCandidates = obsoleteTitle ?
+        activeRelatedComments.filter(comment => !keep.has(comment.id)) :
+        relatedComments.filter(comment => !keep.has(comment.id));
+
+    for (const comment of supersedeCandidates) {
+      if (keep.has(comment.id)) {
+        continue;
+      }
+      await supersedeCommentIfPresent(comment, 'Superseded old AI review');
+    }
+  }
 };

.github/scripts/post-pr-comment.test.js

@@ -0,0 +1,347 @@
+const test = require('node:test');
+const assert = require('node:assert/strict');
+
+const postPrComment = require('./post-pr-comment.js');
+
+const OBSOLETE_MARKER = '<!-- claude-review-obsolete -->';
+const OBSOLETE_TITLE = 'Claude Code Review - OBSOLETE';
+
+function makeComment(id, body, createdAt, updatedAt) {
+  return {
+    id,
+    body,
+    created_at: createdAt,
+    updated_at: updatedAt || createdAt,
+  };
+}
+
+function createHarness(initialComments, options = {}) {
+  let comments = initialComments.map(comment => ({...comment}));
+  let nextCommentId = options.nextCommentId || 1000;
+  let paginateCount = 0;
+
+  const calls = {
+    update: [],
+    create: [],
+    delete: [],
+  };
+
+  const github = {
+    paginate: async () => {
+      paginateCount++;
+      if (options.onPaginate) {
+        const updated = options.onPaginate({
+          paginateCount,
+          comments: comments.map(comment => ({...comment})),
+        });
+        if (updated) {
+          comments = updated.map(comment => ({...comment}));
+        }
+      }
+      return comments.map(comment => ({...comment}));
+    },
+    rest: {
+      issues: {
+        listComments: () => {
+          throw new Error('listComments should only be used through paginate');
+        },
+        updateComment: async ({comment_id, body}) => {
+          calls.update.push({comment_id, body});
+          const error =
+              options.updateErrors && options.updateErrors[comment_id];
+          if (error) {
+            throw error;
+          }
+          const index =
+              comments.findIndex(comment => comment.id === comment_id);
+          if (index === -1) {
+            const notFound = new Error(`Comment ${comment_id} not found`);
+            notFound.status = 404;
+            throw notFound;
+          }
+          const updated = {
+            ...comments[index],
+            body,
+            updated_at: options.updateTimestamp || '2026-04-24T00:00:00Z',
+          };
+          comments[index] = updated;
+          return {data: {...updated}};
+        },
+        createComment: async ({issue_number, body}) => {
+          calls.create.push({issue_number, body});
+          const created = makeComment(
+              nextCommentId++, body,
+              options.createTimestamp || '2026-04-24T00:00:00Z');
+          comments.push(created);
+          return {data: {id: created.id}};
+        },
+        deleteComment: async ({comment_id}) => {
+          calls.delete.push({comment_id});
+          const error =
+              options.deleteErrors && options.deleteErrors[comment_id];
+          if (error) {
+            throw error;
+          }
+          comments = comments.filter(comment => comment.id !== comment_id);
+        },
+      },
+    },
+  };
+
+  return {
+    github,
+    calls,
+    getComments: () => comments.map(comment => ({...comment})),
+  };
+}
+
+function createCore() {
+  return {
+    info: () => {},
+    warning: () => {},
+  };
+}
+
+function escapeRegExp(value) {
+  return value.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
+}
+
+function assertObsoleteComment(comment, originalBody) {
+  assert.ok(comment);
+  assert.match(comment.body, new RegExp(escapeRegExp(OBSOLETE_MARKER)));
+  assert.match(comment.body, new RegExp(`## ${escapeRegExp(OBSOLETE_TITLE)}`));
+  assert.match(comment.body, /<details>/);
+  assert.match(comment.body, /Superseded by a newer AI review/);
+  assert.match(comment.body, new RegExp(escapeRegExp(originalBody)));
+}
+
+const context = {
+  repo: {
+    owner: 'facebook',
+    repo: 'rocksdb',
+  },
+};
+
+test(
+    'creates a fresh review comment and supersedes legacy comments',
+    async () => {
+      const harness = createHarness(
+          [
+            makeComment(
+                1, '<!-- claude-review-auto -->\nold legacy', 'not-a-date'),
+            makeComment(
+                2, '<!-- claude-review-auto -->\nnew legacy',
+                '2026-04-21T00:00:00Z'),
+          ],
+          {
+            updateTimestamp: '2026-04-24T00:00:00Z',
+          });
+
+      await postPrComment({
+        github: harness.github,
+        context,
+        core: createCore(),
+        prNumber: 14659,
+        body: 'review body',
+        marker: '<!-- claude-review-auto-run-500 -->',
+        legacyMarkers: ['<!-- claude-review-auto -->'],
+        prunePrefix: '<!-- claude-review-auto-',
+        preserveLatest: 1,
+        obsoleteMarker: OBSOLETE_MARKER,
+        obsoleteTitle: OBSOLETE_TITLE,
+      });
+
+      assert.equal(harness.calls.create.length, 1);
+      assert.deepEqual(
+          harness.calls.update.map(call => call.comment_id), [2, 1]);
+      assert.equal(harness.calls.delete.length, 0);
+
+      const comments = harness.getComments();
+      assert.equal(comments.length, 3);
+      assert.match(
+          comments.find(comment => comment.id === 1000).body,
+          /<!-- claude-review-auto-run-500 -->/);
+      assertObsoleteComment(
+          comments.find(comment => comment.id === 2),
+          '<!-- claude-review-auto -->\nnew legacy');
+      assertObsoleteComment(
+          comments.find(comment => comment.id === 1),
+          '<!-- claude-review-auto -->\nold legacy');
+    });
+
+test(
+    'updates an exact-match comment and supersedes leftover legacy comments',
+    async () => {
+      const harness = createHarness(
+          [
+            makeComment(
+                3, '<!-- claude-review-auto-abcdef0 -->\ncurrent body',
+                '2026-04-24T00:00:00Z'),
+            makeComment(
+                4, '<!-- claude-review-auto -->\nlegacy body',
+                '2026-04-20T00:00:00Z'),
+          ],
+          {
+            updateTimestamp: '2026-04-24T01:00:00Z',
+          });
+
+      await postPrComment({
+        github: harness.github,
+        context,
+        core: createCore(),
+        prNumber: 14659,
+        body: 'refreshed body',
+        marker: '<!-- claude-review-auto-abcdef0 -->',
+        legacyMarkers: ['<!-- claude-review-auto -->'],
+        prunePrefix: '<!-- claude-review-auto-',
+        preserveLatest: 1,
+        obsoleteMarker: OBSOLETE_MARKER,
+        obsoleteTitle: OBSOLETE_TITLE,
+      });
+
+      assert.deepEqual(
+          harness.calls.update.map(call => call.comment_id), [3, 4]);
+      assert.equal(harness.calls.create.length, 0);
+      assert.equal(harness.calls.delete.length, 0);
+
+      const comments = harness.getComments();
+      assert.match(
+          comments.find(comment => comment.id === 3).body,
+          /<!-- claude-review-auto-abcdef0 -->\nrefreshed body/);
+      assertObsoleteComment(
+          comments.find(comment => comment.id === 4),
+          '<!-- claude-review-auto -->\nlegacy body');
+    });
+
+test(
+    'creates a new comment if the target disappears before update',
+    async () => {
+      const missing = new Error('gone');
+      missing.status = 404;
+      const harness = createHarness(
+          [
+            makeComment(
+                10, '<!-- claude-review-auto-abcdef0 -->\nold body',
+                '2026-04-20T00:00:00Z'),
+          ],
+          {
+            updateErrors: {
+              10: missing,
+            },
+          });
+
+      await postPrComment({
+        github: harness.github,
+        context,
+        core: createCore(),
+        prNumber: 14659,
+        body: 'replacement body',
+        marker: '<!-- claude-review-auto-abcdef0 -->',
+      });
+
+      assert.equal(harness.calls.update.length, 1);
+      assert.equal(harness.calls.create.length, 1);
+      assert.match(
+          harness.calls.create[0].body, /<!-- claude-review-auto-abcdef0 -->/);
+    });
+
+test(
+    'ignores 404 when superseding a comment already deleted by another run',
+    async () => {
+      const missing = new Error('gone');
+      missing.status = 404;
+      const harness = createHarness(
+          [
+            makeComment(
+                20, '<!-- claude-review-auto-oldest -->\noldest',
+                '2026-04-20T00:00:00Z'),
+            makeComment(
+                21, '<!-- claude-review-auto-newer -->\nnewer',
+                '2026-04-21T00:00:00Z'),
+          ],
+          {
+            nextCommentId: 30,
+            createTimestamp: '2026-04-24T00:00:00Z',
+            deleteErrors: {
+              20: missing,
+            },
+          });
+
+      await postPrComment({
+        github: harness.github,
+        context,
+        core: createCore(),
+        prNumber: 14659,
+        body: 'fresh body',
+        marker: '<!-- claude-review-auto-latest -->',
+        prunePrefix: '<!-- claude-review-auto-',
+        preserveLatest: 1,
+        obsoleteMarker: OBSOLETE_MARKER,
+        obsoleteTitle: OBSOLETE_TITLE,
+      });
+
+      assert.equal(harness.calls.create.length, 1);
+      assert.equal(harness.calls.delete.length, 0);
+      assert.deepEqual(
+          harness.calls.update.map(call => call.comment_id), [21, 20]);
+
+      const comments = harness.getComments();
+      assertObsoleteComment(
+          comments.find(comment => comment.id === 21),
+          '<!-- claude-review-auto-newer -->\nnewer');
+    });
+
+test(
+    'supersedes the current review when a newer concurrent one appears',
+    async () => {
+      const harness = createHarness(
+          [
+            makeComment(
+                40, '<!-- claude-review-auto-current -->\ncurrent',
+                '2026-04-20T00:00:00Z'),
+            makeComment(
+                41, '<!-- claude-review-auto-older -->\nolder',
+                '2026-04-19T00:00:00Z'),
+          ],
+          {
+            updateTimestamp: '2026-04-24T00:00:00Z',
+            onPaginate: ({paginateCount, comments}) => {
+              if (paginateCount !== 2) {
+                return comments;
+              }
+              return comments.concat([
+                makeComment(
+                    42, '<!-- claude-review-auto-newer -->\nnewer',
+                    '2026-04-25T00:00:00Z'),
+              ]);
+            },
+          });
+
+      await postPrComment({
+        github: harness.github,
+        context,
+        core: createCore(),
+        prNumber: 14659,
+        body: 'updated current body',
+        marker: '<!-- claude-review-auto-current -->',
+        prunePrefix: '<!-- claude-review-auto-',
+        preserveLatest: 1,
+        obsoleteMarker: OBSOLETE_MARKER,
+        obsoleteTitle: OBSOLETE_TITLE,
+      });
+
+      assert.deepEqual(
+          harness.calls.update.map(call => call.comment_id), [40, 40, 41]);
+      assert.equal(harness.calls.delete.length, 0);
+
+      const comments = harness.getComments();
+      assert.match(
+          comments.find(comment => comment.id === 42).body,
+          /<!-- claude-review-auto-newer -->\nnewer/);
+      assertObsoleteComment(
+          comments.find(comment => comment.id === 40),
+          '<!-- claude-review-auto-current -->\nupdated current body');
+      assertObsoleteComment(
+          comments.find(comment => comment.id === 41),
+          '<!-- claude-review-auto-older -->\nolder');
+    });

.github/workflows/ai-review-comment.yml

@@ -0,0 +1,185 @@
+# Shared comment-posting workflow for AI reviews.
+
+name: AI Review Comment
+
+on:
+  workflow_call:
+    inputs:
+      provider:
+        description: Provider key, for example "claude" or "codex"
+        required: true
+        type: string
+      result_artifact_name:
+        description: Artifact name produced by the analysis workflow
+        required: true
+        type: string
+      comment_file:
+        description: Markdown comment file produced by the analysis workflow
+        required: true
+        type: string
+
+permissions:
+  pull-requests: write
+  issues: write
+
+jobs:
+  comment:
+    if: github.event.workflow_run.conclusion == 'success'
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout scripts
+        uses: actions/checkout@v4
+        with:
+          sparse-checkout: .github/scripts
+          sparse-checkout-cone-mode: true
+
+      - name: Download review artifact
+        id: download
+        uses: actions/download-artifact@v4
+        with:
+          name: ${{ inputs.result_artifact_name }}
+          run-id: ${{ github.event.workflow_run.id }}
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+        continue-on-error: true
+
+      - name: Post or update PR comment
+        if: steps.download.outcome == 'success'
+        env:
+          PROVIDER: ${{ inputs.provider }}
+          COMMENT_FILE: ${{ inputs.comment_file }}
+          RUN_ID: ${{ github.event.workflow_run.id }}
+        uses: actions/github-script@v7
+        with:
+          script: |
+            const fs = require('fs');
+            if (!fs.existsSync(process.env.COMMENT_FILE) ||
+                !fs.existsSync('pr_number.txt')) {
+              core.info('No review results found; skipping.');
+              return;
+            }
+
+            const post = require('./.github/scripts/post-pr-comment.js');
+            const provider = process.env.PROVIDER;
+            const providerTitle = provider === 'codex' ? 'Codex' : 'Claude';
+            const body = fs.readFileSync(process.env.COMMENT_FILE, 'utf8');
+            const prNumber = parseInt(
+              fs.readFileSync('pr_number.txt', 'utf8').trim(),
+              10
+            );
+            const trigger = fs.existsSync('trigger_type.txt') ?
+              fs.readFileSync('trigger_type.txt', 'utf8').trim() :
+              'auto';
+            const headSha = fs.existsSync('head_sha.txt') ?
+              fs.readFileSync('head_sha.txt', 'utf8').trim() :
+              '';
+            const shortSha = headSha ? headSha.substring(0, 7) : '';
+            const runId = process.env.RUN_ID;
+
+            let marker = '';
+            let legacyMarkers = [];
+            let prunePrefix = '';
+            let preserveLatest = 0;
+            let obsoleteMarker = '';
+            let obsoleteTitle = '';
+
+            if (trigger === 'auto') {
+              if (!shortSha) {
+                core.warning(
+                  'Missing head_sha.txt for auto review; skipping comment.'
+                );
+                return;
+              }
+              marker = `<!-- ${provider}-review-auto-run-${runId} -->`;
+              legacyMarkers = [`<!-- ${provider}-review-auto -->`];
+              prunePrefix = `<!-- ${provider}-review-auto-`;
+              preserveLatest = 1;
+              obsoleteMarker = `<!-- ${provider}-review-obsolete -->`;
+              obsoleteTitle = `${providerTitle} Code Review - OBSOLETE`;
+            } else {
+              marker = `<!-- ${provider}-review-manual-run-${runId} -->`;
+            }
+
+            await post({
+              github,
+              context,
+              core,
+              prNumber,
+              body,
+              marker,
+              legacyMarkers,
+              prunePrefix,
+              preserveLatest,
+              obsoleteMarker,
+              obsoleteTitle,
+            });
+
+      - name: Add reaction to trigger comment
+        if: steps.download.outcome == 'success'
+        uses: actions/github-script@v7
+        with:
+          script: |
+            const fs = require('fs');
+            if (!fs.existsSync('trigger_type.txt')) return;
+            const trigger = fs.readFileSync('trigger_type.txt', 'utf8').trim();
+            if (trigger !== 'manual') return;
+            if (!fs.existsSync('comment_id.txt')) return;
+            const commentId = parseInt(
+              fs.readFileSync('comment_id.txt', 'utf8').trim(),
+              10
+            );
+            if (!commentId) return;
+
+            await github.rest.reactions.createForIssueComment({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              comment_id: commentId,
+              content: 'rocket'
+            });
+
+  failure-notice:
+    if: github.event.workflow_run.conclusion == 'failure'
+    runs-on: ubuntu-latest
+    steps:
+      - name: Download artifact for PR number
+        id: download
+        uses: actions/download-artifact@v4
+        with:
+          name: ${{ inputs.result_artifact_name }}
+          run-id: ${{ github.event.workflow_run.id }}
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+        continue-on-error: true
+
+      - name: React with failure
+        if: steps.download.outcome == 'success'
+        uses: actions/github-script@v7
+        with:
+          script: |
+            const fs = require('fs');
+            if (!fs.existsSync('comment_id.txt')) return;
+            const commentId = parseInt(
+              fs.readFileSync('comment_id.txt', 'utf8').trim(),
+              10
+            );
+            if (!commentId) return;
+
+            await github.rest.reactions.createForIssueComment({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              comment_id: commentId,
+              content: 'confused'
+            });
+
+  unauthorized-notice:
+    runs-on: ubuntu-latest
+    if: >-
+      github.event.workflow_run.event == 'issue_comment' &&
+      github.event.workflow_run.conclusion == 'skipped'
+    steps:
+      - name: Log skipped unauthorized request
+        uses: actions/github-script@v7
+        with:
+          script: |
+            core.info(
+              'Analysis workflow was skipped, which usually means the ' +
+              'issue_comment requester was not in the authorized list.'
+            );

.github/workflows/clang-tidy-comment.yml

@@ -20,7 +20,7 @@ jobs:
 
     - name: Download clang-tidy results
       id: download
-      uses: actions/download-artifact@v4.0.0
+      uses: actions/download-artifact@v4.1.3
       with:
         name: clang-tidy-result
         run-id: ${{ github.event.workflow_run.id }}

.github/workflows/clang-tidy.yml

@@ -11,7 +11,7 @@ jobs:
     runs-on:
       labels: 4-core-ubuntu
     container:
-      image: ghcr.io/facebook/rocksdb_ubuntu:24.0
+      image: ghcr.io/facebook/rocksdb_ubuntu:24.1
     steps:
     - uses: actions/checkout@v4.1.0
       with:
@@ -36,14 +36,14 @@ jobs:
         echo "skip=false" >> "$GITHUB_OUTPUT"
     - name: Install clang-tidy
       if: steps.diff-base.outputs.skip != 'true'
-      run: apt-get update && apt-get install -y clang-tidy
+      run: apt-get update && apt-get install -y clang-tidy-21 && ln -sf /usr/bin/clang-tidy-21 /usr/local/bin/clang-tidy
     - name: Generate compile_commands.json
       if: steps.diff-base.outputs.skip != 'true'
       run: |
         mkdir build && cd build
         cmake -DCMAKE_EXPORT_COMPILE_COMMANDS=ON \
-              -DCMAKE_C_COMPILER=clang-18 \
-              -DCMAKE_CXX_COMPILER=clang++-18 ..
+              -DCMAKE_C_COMPILER=clang-21 \
+              -DCMAKE_CXX_COMPILER=clang++-21 ..
         cd ..
         ln -sf build/compile_commands.json compile_commands.json
     - name: Run clang-tidy on changed files

.github/workflows/claude-review-comment.yml

@@ -1,12 +1,8 @@
 # Claude Code Review — Comment Posting Workflow
 #
-# Companion to claude-review.yml. Downloads the review artifact and posts
-# it as a PR comment. This workflow has write permissions but never runs
-# AI or processes untrusted code.
-#
-# Security: Separating analysis (has ANTHROPIC_API_KEY, no write perms)
-# from posting (has GITHUB_TOKEN write perms, no AI) prevents a crafted
-# PR from tricking Claude into exfiltrating tokens.
+# Thin wrapper around .github/workflows/ai-review-comment.yml so the provider
+# specific workflow name stays stable while the shared implementation lives in
+# one reusable workflow.
 
 name: Post Claude Review Comment
 
@@ -20,127 +16,10 @@ permissions:
   issues: write
 
 jobs:
-  comment:
-    if: github.event.workflow_run.conclusion == 'success'
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout scripts
-        uses: actions/checkout@v4
-        with:
-          sparse-checkout: .github/scripts
-          sparse-checkout-cone-mode: true
-
-      - name: Download review artifact
-        id: download
-        uses: actions/download-artifact@v4
-        with:
-          name: claude-review-result
-          run-id: ${{ github.event.workflow_run.id }}
-          github-token: ${{ secrets.GITHUB_TOKEN }}
-        continue-on-error: true
-
-      - name: Post or update PR comment
-        if: steps.download.outcome == 'success'
-        uses: actions/github-script@v7
-        with:
-          script: |
-            const fs = require('fs');
-            if (!fs.existsSync('claude-review-comment.md') || !fs.existsSync('pr_number.txt')) {
-              core.info('No review results found; skipping.');
-              return;
-            }
-
-            const body = fs.readFileSync('claude-review-comment.md', 'utf8');
-            const prNumber = parseInt(fs.readFileSync('pr_number.txt', 'utf8').trim());
-            const trigger = fs.existsSync('trigger_type.txt')
-              ? fs.readFileSync('trigger_type.txt', 'utf8').trim()
-              : 'auto';
-
-            // Use different markers for auto vs manual so they don't
-            // overwrite each other. Auto-reviews update in place (one
-            // per PR). Manual reviews always create new comments.
-            const marker = trigger === 'auto'
-              ? '<!-- claude-review-auto -->'
-              : `<!-- claude-review-manual-${Date.now()} -->`;
-
-            const post = require('./.github/scripts/post-pr-comment.js');
-
-            await post({
-              github,
-              context,
-              core,
-              prNumber,
-              body,
-              marker,
-            });
-
-      - name: Add reaction to trigger comment
-        if: steps.download.outcome == 'success'
-        uses: actions/github-script@v7
-        with:
-          script: |
-            const fs = require('fs');
-            if (!fs.existsSync('trigger_type.txt')) return;
-            const trigger = fs.readFileSync('trigger_type.txt', 'utf8').trim();
-            if (trigger !== 'manual') return;
-            if (!fs.existsSync('comment_id.txt')) return;
-            const commentId = parseInt(fs.readFileSync('comment_id.txt', 'utf8').trim());
-            if (!commentId) return;
-
-            await github.rest.reactions.createForIssueComment({
-              owner: context.repo.owner,
-              repo: context.repo.repo,
-              comment_id: commentId,
-              content: 'rocket'
-            });
-
-  # Notify on failure (for manual triggers)
-  failure-notice:
-    if: github.event.workflow_run.conclusion == 'failure'
-    runs-on: ubuntu-latest
-    steps:
-      - name: Download artifact for PR number
-        id: download
-        uses: actions/download-artifact@v4
-        with:
-          name: claude-review-result
-          run-id: ${{ github.event.workflow_run.id }}
-          github-token: ${{ secrets.GITHUB_TOKEN }}
-        continue-on-error: true
-
-      - name: React with failure
-        if: steps.download.outcome == 'success'
-        uses: actions/github-script@v7
-        with:
-          script: |
-            const fs = require('fs');
-            if (!fs.existsSync('comment_id.txt')) return;
-            const commentId = parseInt(fs.readFileSync('comment_id.txt', 'utf8').trim());
-            if (!commentId) return;
-
-            await github.rest.reactions.createForIssueComment({
-              owner: context.repo.owner,
-              repo: context.repo.repo,
-              comment_id: commentId,
-              content: 'confused'
-            });
-
-  # Unauthorized user notice
-  unauthorized-notice:
-    runs-on: ubuntu-latest
-
-    if: >-
-      github.event.workflow_run.event == 'issue_comment' &&
-      github.event.workflow_run.conclusion == 'skipped'
-
-    steps:
-      - name: Check if unauthorized
-        id: check
-        uses: actions/github-script@v7
-        with:
-          script: |
-            // This job only runs when the analysis workflow was skipped,
-            // which happens when the user is not in the authorized list.
-            // We cannot easily get the original comment from workflow_run,
-            // so this is a best-effort notice via workflow logs.
-            core.info('Analysis workflow was skipped — likely unauthorized user.');
+  review-comment:
+    uses: ./.github/workflows/ai-review-comment.yml
+    with:
+      provider: claude
+      result_artifact_name: claude-review-result
+      comment_file: claude-review-comment.md
+    secrets: inherit

.github/workflows/claude-review.yml

@@ -1,19 +1,8 @@
 # Claude Code Review — Analysis Workflow
 #
-# This workflow runs Claude to review PRs. It produces a markdown review
-# as an artifact but does NOT post comments (no write permissions).
-# The companion workflow (claude-review-comment.yml) handles posting.
-#
-# Triggers:
-#   1. workflow_run: Auto-review after pr-jobs passes
-#   2. issue_comment: Manual /claude-review or /claude-query by maintainers
-#   3. workflow_dispatch: Manual testing
-#
-# Security:
-#   - This job has contents:read ONLY (no PR write, no issue write)
-#   - ANTHROPIC_API_KEY is used here but never coexists with write tokens
-#   - Claude uses read-only tools (View, GlobTool, GrepTool)
-#   - Review methodology: claude_md/code_review.md
+# Thin wrapper around .github/workflows/ai-review-analysis.yml so provider
+# specific triggers and workflow_dispatch inputs stay stable while the shared
+# implementation lives in one reusable workflow.
 
 name: Claude Code Review
 
@@ -22,6 +11,11 @@ on:
     workflows: ["facebook/rocksdb/pr-jobs"]
     types: [completed]
 
+  # The early pull_request_target path is limited to same-repo PRs by the job
+  # condition below. Fork PRs skip this path and rely on workflow_run instead.
+  pull_request_target:
+    types: [opened, reopened, synchronize]
+
   issue_comment:
     types: [created]
 
@@ -32,445 +26,44 @@ on:
         required: true
         type: number
       model:
-        description: Claude model to use
+        description: Claude model to use (defaults to latest Opus)
         required: false
         type: choice
         options:
           - claude-opus-4-6
-          - claude-sonnet-4-20250514
+          - claude-sonnet-4-6
         default: claude-opus-4-6
+      thinking_budget:
+        description: Override MAX_THINKING_TOKENS (blank = auto-classify, capped at 24000)
+        required: false
+        type: string
+        default: ""
 
 permissions:
   contents: read
+  # The shared analysis workflow polls check state, inspects prior runs, and
+  # reads PR metadata. Keep those permissions read-only in the caller.
+  actions: read
+  checks: read
+  pull-requests: read
 
 jobs:
-  # ======================== Auto Review ======================== #
-  auto-review:
-    name: Auto Claude Review
-    runs-on: ubuntu-latest
-    permissions:
-      contents: read
-      actions: read
-
+  review:
     if: >-
-      github.event_name == 'workflow_run' &&
-      github.event.workflow_run.conclusion == 'success' &&
-      github.event.workflow_run.event == 'pull_request'
-
-    steps:
-      - name: Get PR info
-        id: pr_info
-        uses: actions/github-script@v7
-        with:
-          script: |
-            const headSha = context.payload.workflow_run.head_sha;
-            let prNumber = null;
-
-            const prs = context.payload.workflow_run.pull_requests;
-            if (prs && prs.length > 0) {
-              prNumber = prs[0].number;
-            } else {
-              const { data: pulls } = await github.rest.pulls.list({
-                owner: context.repo.owner,
-                repo: context.repo.repo,
-                state: 'open',
-                sort: 'updated',
-                direction: 'desc',
-                per_page: 30
-              });
-              const match = pulls.find(p => p.head.sha === headSha);
-              if (match) prNumber = match.number;
-            }
-
-            if (!prNumber) {
-              core.setFailed(`Could not find PR for SHA ${headSha}`);
-              return;
-            }
-
-            const pr = await github.rest.pulls.get({
-              owner: context.repo.owner,
-              repo: context.repo.repo,
-              pull_number: prNumber
-            });
-
-            core.setOutput('pr_number', prNumber.toString());
-            core.setOutput('head_repo', pr.data.head.repo.clone_url);
-            core.setOutput('head_ref', pr.data.head.ref);
-            core.setOutput('head_sha', pr.data.head.sha);
-            core.setOutput('base_sha', pr.data.base.sha);
-            core.setOutput('base_ref', pr.data.base.ref);
-            core.setOutput('pr_title', pr.data.title);
-            core.setOutput('pr_body', pr.data.body || '');
-            core.setOutput('pr_author', pr.data.user.login);
-
-      - name: Check for existing review
-        id: check_existing
-        uses: actions/github-script@v7
-        env:
-          PR_NUMBER: ${{ steps.pr_info.outputs.pr_number }}
-          HEAD_SHA: ${{ steps.pr_info.outputs.head_sha }}
-        with:
-          script: |
-            const comments = await github.rest.issues.listComments({
-              owner: context.repo.owner,
-              repo: context.repo.repo,
-              issue_number: parseInt(process.env.PR_NUMBER),
-              per_page: 100
-            });
-            const shortSha = process.env.HEAD_SHA.substring(0, 7);
-            const existing = comments.data.find(c =>
-              c.body.includes('<!-- claude-review -->') &&
-              c.body.includes(shortSha)
-            );
-            core.setOutput('skip', existing ? 'true' : 'false');
-
-      - name: Checkout base repository
-        if: steps.check_existing.outputs.skip != 'true'
-        uses: actions/checkout@v4
-        with:
-          ref: ${{ steps.pr_info.outputs.base_ref }}
-          fetch-depth: 0
-          persist-credentials: false
-
-      - name: Generate diff and prompt
-        if: steps.check_existing.outputs.skip != 'true'
-        env:
-          HEAD_REPO: ${{ steps.pr_info.outputs.head_repo }}
-          HEAD_REF: ${{ steps.pr_info.outputs.head_ref }}
-          HEAD_SHA: ${{ steps.pr_info.outputs.head_sha }}
-          BASE_SHA: ${{ steps.pr_info.outputs.base_sha }}
-          PR_TITLE: ${{ steps.pr_info.outputs.pr_title }}
-          PR_BODY: ${{ steps.pr_info.outputs.pr_body }}
-          PR_AUTHOR: ${{ steps.pr_info.outputs.pr_author }}
-        run: |
-          git remote add fork "${HEAD_REPO}" || true
-          git fetch fork "${HEAD_REF}" --depth=100
-
-          git diff "${BASE_SHA}...${HEAD_SHA}" > /tmp/pr.diff
-          DIFF_STATS=$(git diff --stat "${BASE_SHA}...${HEAD_SHA}" | tail -1)
-
-          DIFF_SIZE=$(wc -c < /tmp/pr.diff)
-          IS_TRUNCATED=false
-          if [ "$DIFF_SIZE" -gt 100000 ]; then
-            head -c 100000 /tmp/pr.diff > /tmp/pr_truncated.diff
-            mv /tmp/pr_truncated.diff /tmp/pr.diff
-            IS_TRUNCATED=true
-          fi
-
-          cat > /tmp/prompt.txt << 'PROMPT_EOF'
-          Read claude_md/code_review.md for the full review methodology, then
-          review the pull request described below. Follow all perspectives and
-          the output format defined in that document.
-          PROMPT_EOF
-
-          cat >> /tmp/prompt.txt << PROMPT_EOF
-
-          ## Pull Request Information
-          - **Title:** ${PR_TITLE}
-          - **Author:** ${PR_AUTHOR}
-          - **Changes:** ${DIFF_STATS}
-          PROMPT_EOF
-
-          if [ "${IS_TRUNCATED}" = "true" ]; then
-            echo "- **Note:** Diff truncated due to size." >> /tmp/prompt.txt
-          fi
-
-          printf '\n## PR Description\n' >> /tmp/prompt.txt
-          printf '%s\n' "${PR_BODY}" >> /tmp/prompt.txt
-          printf '\n---\n\n## Diff to Review\n\n' >> /tmp/prompt.txt
-          cat /tmp/pr.diff >> /tmp/prompt.txt
-
-      - name: Run Claude
-        if: steps.check_existing.outputs.skip != 'true'
-        id: claude
-        uses: anthropics/claude-code-base-action@beta
-        with:
-          prompt_file: /tmp/prompt.txt
-          model: claude-opus-4-6
-          max_turns: "30"
-          allowed_tools: "View,GlobTool,GrepTool"
-          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
-
-      - name: Build review comment
-        if: steps.check_existing.outputs.skip != 'true'
-        uses: actions/github-script@v7
-        env:
-          EXECUTION_FILE: ${{ steps.claude.outputs.execution_file }}
-          CONCLUSION: ${{ steps.claude.outputs.conclusion }}
-          HEAD_SHA: ${{ steps.pr_info.outputs.head_sha }}
-        with:
-          script: |
-            const parse = require('./.github/scripts/parse-claude-review.js');
-            const body = parse({
-              executionFile: process.env.EXECUTION_FILE,
-              conclusion: process.env.CONCLUSION,
-              meta: {
-                trigger: 'auto',
-                headSha: process.env.HEAD_SHA,
-                reviewer: '',
-              }
-            });
-            const fs = require('fs');
-            fs.writeFileSync('claude-review-comment.md', body);
-
-      - name: Save PR number
-        if: steps.check_existing.outputs.skip != 'true'
-        run: echo "${{ steps.pr_info.outputs.pr_number }}" > pr_number.txt
-
-      - name: Save trigger metadata
-        if: steps.check_existing.outputs.skip != 'true'
-        run: echo "auto" > trigger_type.txt
-
-      - name: Upload review artifact
-        if: always() && steps.check_existing.outputs.skip != 'true'
-        uses: actions/upload-artifact@v4
-        with:
-          name: claude-review-result
-          path: |
-            claude-review-comment.md
-            pr_number.txt
-            trigger_type.txt
-          if-no-files-found: ignore
-          retention-days: 1
-
-      - name: Upload execution log
-        if: always() && steps.check_existing.outputs.skip != 'true'
-        uses: actions/upload-artifact@v4
-        with:
-          name: claude-review-execution-log
-          path: ${{ steps.claude.outputs.execution_file }}
-          retention-days: 7
-          if-no-files-found: warn
-
-  # ======================== Manual Review ======================== #
-  manual-review:
-    name: Manual Claude Review
-    runs-on: ubuntu-latest
-    permissions:
-      contents: read
-
-    if: >-
-      github.event_name == 'workflow_dispatch' ||
-      (
-        github.event_name == 'issue_comment' &&
-        github.event.issue.pull_request &&
-        (contains(github.event.comment.body, '/claude-review') || contains(github.event.comment.body, '/claude-query')) &&
-        contains(fromJSON('["pdillinger", "jaykorean", "hx235", "dannyhchen", "joshkang97", "zaidoon1", "omkarhgawde", "xingbowang", "anand1976", "virajthakur", "nmk70", "archang19", "mszeszko-meta", "yoshinorim", "cbi42"]'), github.event.comment.user.login)
-      )
-
-    steps:
-      - name: Detect command type
-        id: command
-        uses: actions/github-script@v7
-        with:
-          script: |
-            if (context.eventName === 'workflow_dispatch') {
-              core.setOutput('type', 'review');
-              core.setOutput('query', '');
-              return;
-            }
-            const body = context.payload.comment.body;
-            if (body.includes('/claude-review')) {
-              core.setOutput('type', 'review');
-              const match = body.match(/\/claude-review\s+([\s\S]*)/);
-              core.setOutput('query', match ? match[1].trim() : '');
-            } else if (body.includes('/claude-query')) {
-              const match = body.match(/\/claude-query\s+([\s\S]*)/);
-              const query = match ? match[1].trim() : '';
-              if (!query) {
-                core.setFailed('No query provided. Usage: /claude-query <your question>');
-                return;
-              }
-              core.setOutput('type', 'query');
-              core.setOutput('query', query);
-            } else {
-              core.setFailed('Unknown command');
-            }
-
-      - name: Get PR details
-        id: pr_info
-        uses: actions/github-script@v7
-        env:
-          INPUT_PR_NUMBER: ${{ inputs.pr_number }}
-        with:
-          script: |
-            const prNumber = context.eventName === 'workflow_dispatch'
-              ? parseInt(process.env.INPUT_PR_NUMBER, 10)
-              : context.issue.number;
-
-            const pr = await github.rest.pulls.get({
-              owner: context.repo.owner,
-              repo: context.repo.repo,
-              pull_number: prNumber
-            });
-
-            core.setOutput('pr_number', prNumber.toString());
-            core.setOutput('head_repo', pr.data.head.repo.clone_url);
-            core.setOutput('head_ref', pr.data.head.ref);
-            core.setOutput('head_sha', pr.data.head.sha);
-            core.setOutput('base_sha', pr.data.base.sha);
-            core.setOutput('base_ref', pr.data.base.ref);
-            core.setOutput('pr_title', pr.data.title);
-            core.setOutput('pr_body', pr.data.body || '');
-            core.setOutput('pr_author', pr.data.user.login);
-
-      - name: Checkout base repository
-        uses: actions/checkout@v4
-        with:
-          ref: ${{ steps.pr_info.outputs.base_ref }}
-          fetch-depth: 0
-          persist-credentials: false
-
-      - name: Generate diff and prompt
-        env:
-          HEAD_REPO: ${{ steps.pr_info.outputs.head_repo }}
-          HEAD_REF: ${{ steps.pr_info.outputs.head_ref }}
-          HEAD_SHA: ${{ steps.pr_info.outputs.head_sha }}
-          BASE_SHA: ${{ steps.pr_info.outputs.base_sha }}
-          PR_TITLE: ${{ steps.pr_info.outputs.pr_title }}
-          PR_BODY: ${{ steps.pr_info.outputs.pr_body }}
-          PR_AUTHOR: ${{ steps.pr_info.outputs.pr_author }}
-          DIFF_STATS: ""
-          COMMAND_TYPE: ${{ steps.command.outputs.type }}
-          ADDITIONAL_CONTEXT: ${{ steps.command.outputs.query }}
-          USER_QUERY: ${{ steps.command.outputs.query }}
-        run: |
-          git remote add fork "${HEAD_REPO}" || true
-          git fetch fork "${HEAD_REF}" --depth=100
-
-          git diff "${BASE_SHA}...${HEAD_SHA}" > /tmp/pr.diff
-          DIFF_STATS=$(git diff --stat "${BASE_SHA}...${HEAD_SHA}" | tail -1)
-
-          DIFF_SIZE=$(wc -c < /tmp/pr.diff)
-          IS_TRUNCATED=false
-          if [ "$DIFF_SIZE" -gt 100000 ]; then
-            head -c 100000 /tmp/pr.diff > /tmp/pr_truncated.diff
-            mv /tmp/pr_truncated.diff /tmp/pr.diff
-            IS_TRUNCATED=true
-          fi
-
-          if [ "${COMMAND_TYPE}" = "query" ]; then
-            # Query prompt
-            cat > /tmp/prompt.txt << 'PROMPT_EOF'
-          You are an expert C++ engineer for RocksDB.
-          Read CLAUDE.md in the repo root and claude_md/ files for project context.
-          Answer thoroughly with exact file paths and line numbers.
-
-          ---
-
-          PROMPT_EOF
-
-            cat >> /tmp/prompt.txt << PROMPT_EOF
-          ## Pull Request Context
-          - **Title:** ${PR_TITLE}
-          - **Author:** ${PR_AUTHOR}
-          - **Changes:** ${DIFF_STATS}
-
-          ## PR Description
-          PROMPT_EOF
-            printf '%s\n' "${PR_BODY}" >> /tmp/prompt.txt
-            printf '\n## Question\n' >> /tmp/prompt.txt
-            printf '%s\n' "${USER_QUERY}" >> /tmp/prompt.txt
-            printf '\n---\n## PR Diff (for reference)\n' >> /tmp/prompt.txt
-            cat /tmp/pr.diff >> /tmp/prompt.txt
-          else
-            # Review prompt
-            cat > /tmp/prompt.txt << 'PROMPT_EOF'
-          Read claude_md/code_review.md for the full review methodology, then
-          review the pull request described below. Follow all perspectives and
-          the output format defined in that document.
-          PROMPT_EOF
-
-            cat >> /tmp/prompt.txt << PROMPT_EOF
-
-          ## Pull Request Information
-          - **Title:** ${PR_TITLE}
-          - **Author:** ${PR_AUTHOR}
-          - **Changes:** ${DIFF_STATS}
-          PROMPT_EOF
-
-            if [ "${IS_TRUNCATED}" = "true" ]; then
-              echo "- **Note:** Diff truncated due to size." >> /tmp/prompt.txt
-            fi
-
-            printf '\n## PR Description\n' >> /tmp/prompt.txt
-            printf '%s\n' "${PR_BODY}" >> /tmp/prompt.txt
-
-            if [ -n "${ADDITIONAL_CONTEXT}" ]; then
-              printf '\n## Additional Instructions from Reviewer\n' >> /tmp/prompt.txt
-              printf '%s\n' "${ADDITIONAL_CONTEXT}" >> /tmp/prompt.txt
-            fi
-
-            printf '\n---\n\n## Diff to Review\n\n' >> /tmp/prompt.txt
-            cat /tmp/pr.diff >> /tmp/prompt.txt
-          fi
-
-      - name: Run Claude
-        id: claude
-        uses: anthropics/claude-code-base-action@beta
-        with:
-          prompt_file: /tmp/prompt.txt
-          model: ${{ github.event_name == 'workflow_dispatch' && inputs.model || 'claude-opus-4-6' }}
-          max_turns: "30"
-          allowed_tools: "View,GlobTool,GrepTool"
-          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
-
-      - name: Build review comment
-        uses: actions/github-script@v7
-        env:
-          EXECUTION_FILE: ${{ steps.claude.outputs.execution_file }}
-          CONCLUSION: ${{ steps.claude.outputs.conclusion }}
-          HEAD_SHA: ${{ steps.pr_info.outputs.head_sha }}
-          REVIEWER: ${{ github.event_name == 'workflow_dispatch' && github.actor || github.event.comment.user.login }}
-          COMMAND_TYPE: ${{ steps.command.outputs.type }}
-        with:
-          script: |
-            const parse = require('./.github/scripts/parse-claude-review.js');
-            const isReview = process.env.COMMAND_TYPE === 'review';
-            const body = parse({
-              executionFile: process.env.EXECUTION_FILE,
-              conclusion: process.env.CONCLUSION,
-              meta: {
-                trigger: 'manual',
-                headSha: process.env.HEAD_SHA,
-                reviewer: process.env.REVIEWER,
-                isQuery: !isReview,
-              }
-            });
-            const fs = require('fs');
-            fs.writeFileSync('claude-review-comment.md', body);
-
-      - name: Save PR number
-        run: echo "${{ steps.pr_info.outputs.pr_number }}" > pr_number.txt
-
-      - name: Save trigger metadata
-        env:
-          REVIEWER: ${{ github.event_name == 'workflow_dispatch' && github.actor || github.event.comment.user.login }}
-          COMMENT_ID: ${{ github.event.comment.id }}
-        run: |
-          echo "manual" > trigger_type.txt
-          echo "${REVIEWER}" > reviewer.txt
-          echo "${COMMENT_ID}" > comment_id.txt
-
-      - name: Upload review artifact
-        if: always()
-        uses: actions/upload-artifact@v4
-        with:
-          name: claude-review-result
-          path: |
-            claude-review-comment.md
-            pr_number.txt
-            trigger_type.txt
-            reviewer.txt
-            comment_id.txt
-          if-no-files-found: ignore
-          retention-days: 1
-
-      - name: Upload execution log
-        if: always()
-        uses: actions/upload-artifact@v4
-        with:
-          name: claude-review-execution-log
-          path: ${{ steps.claude.outputs.execution_file }}
-          retention-days: 7
-          if-no-files-found: warn
+      github.event_name != 'pull_request_target' ||
+      github.event.pull_request.head.repo.full_name == github.repository
+    uses: ./.github/workflows/ai-review-analysis.yml
+    with:
+      provider: claude
+      display_name: Claude
+      review_command: /claude-review
+      query_command: /claude-query
+      result_artifact_name: claude-review-result
+      comment_file: claude-review-comment.md
+      default_model: claude-opus-4-6
+      selected_model: ${{ github.event_name == 'workflow_dispatch' && inputs.model || '' }}
+      classifier_model: claude-sonnet-4-6
+      recovery_model: claude-sonnet-4-6
+      thinking_budget: ${{ github.event_name == 'workflow_dispatch' && inputs.thinking_budget || '' }}
+      dispatch_pr_number: ${{ github.event_name == 'workflow_dispatch' && inputs.pr_number || '' }}
+    secrets: inherit

.github/workflows/codex-review-comment.yml

@@ -0,0 +1,25 @@
+# Codex Code Review — Comment Posting Workflow
+#
+# Thin wrapper around .github/workflows/ai-review-comment.yml so the provider
+# specific workflow name stays stable while the shared implementation lives in
+# one reusable workflow.
+
+name: Post Codex Review Comment
+
+on:
+  workflow_run:
+    workflows: ["Codex Code Review"]
+    types: [completed]
+
+permissions:
+  pull-requests: write
+  issues: write
+
+jobs:
+  review-comment:
+    uses: ./.github/workflows/ai-review-comment.yml
+    with:
+      provider: codex
+      result_artifact_name: codex-review-result
+      comment_file: codex-review-comment.md
+    secrets: inherit

.github/workflows/codex-review.yml

@@ -0,0 +1,68 @@
+# Codex Code Review — Analysis Workflow
+#
+# Thin wrapper around .github/workflows/ai-review-analysis.yml so provider
+# specific triggers and workflow_dispatch inputs stay stable while the shared
+# implementation lives in one reusable workflow.
+
+name: Codex Code Review
+
+on:
+  workflow_run:
+    workflows: ["facebook/rocksdb/pr-jobs"]
+    types: [completed]
+
+  # The early pull_request_target path is limited to same-repo PRs by the job
+  # condition below. Fork PRs skip this path and rely on workflow_run instead.
+  pull_request_target:
+    types: [opened, reopened, synchronize]
+
+  issue_comment:
+    types: [created]
+
+  workflow_dispatch:
+    inputs:
+      pr_number:
+        description: PR number to review
+        required: true
+        type: number
+      model:
+        description: Codex model to use
+        required: false
+        type: choice
+        options:
+          - gpt-5.5
+          - gpt-5.3-codex
+          - gpt-5.2-codex
+        default: gpt-5.5
+      thinking_budget:
+        description: Override MAX_THINKING_TOKENS (blank = auto-classify)
+        required: false
+        type: string
+        default: ""
+
+permissions:
+  contents: read
+  # The shared analysis workflow polls check state, inspects prior runs, and
+  # reads PR metadata. Keep those permissions read-only in the caller.
+  actions: read
+  checks: read
+  pull-requests: read
+
+jobs:
+  review:
+    if: >-
+      github.event_name != 'pull_request_target' ||
+      github.event.pull_request.head.repo.full_name == github.repository
+    uses: ./.github/workflows/ai-review-analysis.yml
+    with:
+      provider: codex
+      display_name: Codex
+      review_command: /codex-review
+      query_command: /codex-query
+      result_artifact_name: codex-review-result
+      comment_file: codex-review-comment.md
+      default_model: gpt-5.5
+      selected_model: ${{ github.event_name == 'workflow_dispatch' && inputs.model || '' }}
+      thinking_budget: ${{ github.event_name == 'workflow_dispatch' && inputs.thinking_budget || '' }}
+      dispatch_pr_number: ${{ github.event_name == 'workflow_dispatch' && inputs.pr_number || '' }}
+    secrets: inherit

.github/workflows/nightly.yml

@@ -41,22 +41,41 @@ jobs:
     - uses: "./.github/actions/pre-steps"
     - run: make V=1 -j32 check
     - uses: "./.github/actions/post-steps"
-  build-linux-clang-18-asan-ubsan-with-folly:
+  build-linux-clang-21-asan-ubsan-with-folly:
     if: ${{ github.repository_owner == 'facebook' }}
     runs-on:
       labels: 16-core-ubuntu
     container:
-      image: ghcr.io/facebook/rocksdb_ubuntu:24.0
+      image: ghcr.io/facebook/rocksdb_ubuntu:24.1
       options: --shm-size=16gb
     env:
-      CC: clang-18
-      CXX: clang++-18
+      CC: clang-21
+      CXX: clang++-21
     steps:
     - uses: actions/checkout@v4.1.0
     - uses: "./.github/actions/pre-steps"
     - uses: "./.github/actions/cache-getdeps-downloads"
     - uses: "./.github/actions/setup-folly"
-    - uses: "./.github/actions/build-folly"
+    - name: Build folly and dependencies
+      run: |
+        clean_path=()
+        IFS=: read -ra path_entries <<< "$PATH"
+        for entry in "${path_entries[@]}"; do
+          if [[ "$entry" != "/usr/lib/ccache" ]]; then
+            clean_path+=("$entry")
+          fi
+        done
+        export PATH="$(IFS=:; echo "${clean_path[*]}")"
+        ccache_bin="$(command -v ccache || true)"
+        if [[ -n "$ccache_bin" && -x "$ccache_bin" ]]; then
+          mv "$ccache_bin" "${ccache_bin}.disabled"
+        fi
+        export CC=gcc
+        export CXX=g++
+        export USE_CCACHE=0
+        export CCACHE_DISABLE=1
+        make build_folly
+      shell: bash
     - run: LIB_MODE=static USE_CLANG=1 USE_FOLLY=1 COMPILE_WITH_UBSAN=1 COMPILE_WITH_ASAN=1 make -j32 check
     - uses: "./.github/actions/post-steps"
   build-linux-cmake-with-folly:
@@ -94,11 +113,10 @@ jobs:
     - run: "USE_FOLLY=1 LIB_MODE=static DEBUG_LEVEL=0 V=1 make -j20 release"
     - run: "(mkdir build && cd build && cmake -DUSE_FOLLY=1 -DWITH_GFLAGS=1 -DROCKSDB_BUILD_SHARED=0 -DCMAKE_BUILD_TYPE=Release .. && make VERBOSE=1 -j20 && ctest -j20)"
     - uses: "./.github/actions/post-steps"
-  build-windows-vs2022-avx2:
+  build-windows-msvc-avx2:
     if: ${{ github.repository_owner == 'facebook' }}
-    runs-on: windows-2022
+    runs-on: windows-8-core
     env:
-      CMAKE_GENERATOR: Visual Studio 17 2022
       CMAKE_PORTABLE: AVX2
     steps:
     - uses: actions/checkout@v4.1.0
@@ -145,13 +163,13 @@ jobs:
     runs-on:
       labels: 4-core-ubuntu
     container:
-      image: ghcr.io/facebook/rocksdb_ubuntu:24.0
+      image: ghcr.io/facebook/rocksdb_ubuntu:24.1
       options: --shm-size=16gb
     steps:
     - uses: actions/checkout@v4.1.0
     - uses: "./.github/actions/pre-steps"
     - name: Build rocksdb lib
-      run: CC=clang-18 CXX=clang++-18 USE_CLANG=1 make -j4 static_lib
+      run: CC=clang-21 CXX=clang++-21 USE_CLANG=1 make -j4 static_lib
     - name: Build fuzzers
       run: cd fuzz && make sst_file_writer_fuzzer db_fuzzer db_map_fuzzer
     - uses: "./.github/actions/post-steps"

.github/workflows/pr-jobs.yml

@@ -72,6 +72,8 @@ jobs:
       run: make check-buck-targets
     - name: Simple source code checks
       run: make check-sources
+    - name: Validate GitHub Actions YAML
+      run: make check-workflow-yaml
     - name: Sanity check check_format_compatible.sh
       run: |-
         export TEST_TMPDIR=/dev/shm/rocksdb
@@ -105,7 +107,7 @@ jobs:
     runs-on:
       labels: 4-core-ubuntu
     container:
-      image: ghcr.io/facebook/rocksdb_ubuntu:24.0
+      image: ghcr.io/facebook/rocksdb_ubuntu:24.1
       options: --shm-size=16gb
     steps:
     - uses: actions/checkout@v4.1.0
@@ -295,23 +297,40 @@ jobs:
     - uses: "./.github/actions/teardown-ccache"
       if: always()
     - uses: "./.github/actions/post-steps"
-  build-linux-clang-18-no_test_run:
-    if: needs.config.outputs.only_job == 'build-linux-clang-18-no_test_run' || (needs.config.outputs.only_job == '' && github.repository_owner == 'facebook')
+  build-linux-clang-21-no_test_run:
+    if: needs.config.outputs.only_job == 'build-linux-clang-21-no_test_run' || (needs.config.outputs.only_job == '' && github.repository_owner == 'facebook')
     needs: config
     runs-on:
       labels: 16-core-ubuntu
     container:
-      image: ghcr.io/facebook/rocksdb_ubuntu:24.0
+      image: ghcr.io/facebook/rocksdb_ubuntu:24.1
       options: --shm-size=16gb
     steps:
     - uses: actions/checkout@v4.1.0
     - uses: "./.github/actions/pre-steps"
     - uses: "./.github/actions/setup-ccache"
       with:
-        cache-key-prefix: clang-18
-    - run: CC=clang-18 CXX=clang++-18 USE_CLANG=1 make -j32 all microbench
+        cache-key-prefix: clang-21
+    - run: CC=clang-21 CXX=clang++-21 USE_CLANG=1 make -j32 all microbench
     - run: make clean
-    - run: CC=clang-18 CXX=clang++-18 USE_CLANG=1 DEBUG_LEVEL=0 make -j32 release
+    - run: CC=clang-21 CXX=clang++-21 USE_CLANG=1 DEBUG_LEVEL=0 make -j32 release
+    - name: Install clang-format-21 for C API gen checks
+      run: apt-get update && apt-get install -y clang-format-21
+    - name: Check C API codegen (link-complete, backward-compatible, up to date)
+      # Run the make target in STRICT mode so this core CI job HARD-FAILS rather
+      # than silently skipping if a prerequisite (clang++, clang-format, or the
+      # compat baseline ref) is ever missing. See `make check-c-api-gen`.
+      run: |
+        # The container checkout is owned by a different user, so git refuses to
+        # operate on it ("dubious ownership") -- mark it safe for the `git fetch`
+        # and the `git show` the compat checker runs.
+        git config --global --add safe.directory "$GITHUB_WORKSPACE"
+        git config --global --add safe.directory '*'
+        git fetch --no-tags --depth=1 origin main
+        CXX=clang++-21 make check-c-api-gen \
+          CHECK_C_API_GEN_STRICT=1 \
+          API_COMPAT_REF=FETCH_HEAD \
+          CLANG_FORMAT_BINARY=clang-format-21
     - uses: "./.github/actions/teardown-ccache"
       if: always()
     - uses: "./.github/actions/post-steps"
@@ -321,7 +340,7 @@ jobs:
     runs-on:
       labels: 16-core-ubuntu
     container:
-      image: ghcr.io/facebook/rocksdb_ubuntu:24.0
+      image: ghcr.io/facebook/rocksdb_ubuntu:24.1
       options: --shm-size=16gb
     steps:
     - uses: actions/checkout@v4.1.0
@@ -371,22 +390,25 @@ jobs:
           crash_duration: 240
     container:
       image: ghcr.io/facebook/rocksdb_ubuntu:22.1
-      options: --shm-size=16gb
+      options: --shm-size=16gb --ulimit nofile=1048576:1048576
     steps:
     - uses: actions/checkout@v4.1.0
     - uses: "./.github/actions/pre-steps"
     - uses: "./.github/actions/setup-ccache"
       with:
         cache-key-prefix: mini-crashtest
-    - run: ulimit -S -n `ulimit -H -n` && make V=1 -j8 CRASH_TEST_EXT_ARGS='--duration=${{ matrix.crash_duration }} --max_key=2500000' ${{ matrix.crash_test_target }}
+    - run: |
+        ulimit -S -n $(ulimit -Hn) # Raise open files soft limit to hard limit
+        echo "Updated ulimit -Hn: $(ulimit -Hn) -Sn: $(ulimit -Sn)"
+        make V=1 -j8 CRASH_TEST_EXT_ARGS='--duration=${{ matrix.crash_duration }} --max_key=2500000' ${{ matrix.crash_test_target }}
     - uses: "./.github/actions/teardown-ccache"
       if: always()
     - uses: "./.github/actions/post-steps"
       with:
         artifact-prefix: "${{ github.job }}-${{ matrix.crash_test_target }}"
   # ======================= Linux with Sanitizers ===================== #
-  build-linux-clang18-asan-ubsan:
-    if: needs.config.outputs.only_job == 'build-linux-clang18-asan-ubsan' || (needs.config.outputs.only_job == '' && github.repository_owner == 'facebook')
+  build-linux-clang21-asan-ubsan:
+    if: needs.config.outputs.only_job == 'build-linux-clang21-asan-ubsan' || (needs.config.outputs.only_job == '' && github.repository_owner == 'facebook')
     needs: config
     runs-on:
       labels: 32-core-ubuntu
@@ -395,15 +417,15 @@ jobs:
       matrix:
         shard: [0, 1, 2]
     container:
-      image: ghcr.io/facebook/rocksdb_ubuntu:24.0
+      image: ghcr.io/facebook/rocksdb_ubuntu:24.1
       options: --shm-size=16gb
     steps:
     - uses: actions/checkout@v4.1.0
     - uses: "./.github/actions/pre-steps"
     - uses: "./.github/actions/setup-ccache"
       with:
-        cache-key-prefix: clang18-asan-ubsan
-    - run: COMPILE_WITH_ASAN=1 COMPILE_WITH_UBSAN=1 CC=clang-18 CXX=clang++-18 ROCKSDB_DISABLE_ALIGNED_NEW=1 USE_CLANG=1 make V=1 -j40 check
+        cache-key-prefix: clang21-asan-ubsan
+    - run: COMPILE_WITH_ASAN=1 COMPILE_WITH_UBSAN=1 CC=clang-21 CXX=clang++-21 ROCKSDB_DISABLE_ALIGNED_NEW=1 USE_CLANG=1 make V=1 -j40 check
       env:
         CI_SHARD_INDEX: ${{ matrix.shard }}
         CI_TOTAL_SHARDS: 3
@@ -412,8 +434,8 @@ jobs:
     - uses: "./.github/actions/post-steps"
       with:
         artifact-prefix: "${{ github.job }}-shard${{ matrix.shard }}"
-  build-linux-clang18-mini-tsan:
-    if: needs.config.outputs.only_job == 'build-linux-clang18-mini-tsan' || (needs.config.outputs.only_job == '' && github.repository_owner == 'facebook')
+  build-linux-clang21-mini-tsan:
+    if: needs.config.outputs.only_job == 'build-linux-clang21-mini-tsan' || (needs.config.outputs.only_job == '' && github.repository_owner == 'facebook')
     needs: config
     runs-on:
       labels: 32-core-ubuntu
@@ -422,15 +444,15 @@ jobs:
       matrix:
         shard: [0, 1, 2]
     container:
-      image: ghcr.io/facebook/rocksdb_ubuntu:24.0
+      image: ghcr.io/facebook/rocksdb_ubuntu:24.1
       options: --shm-size=16gb
     steps:
     - uses: actions/checkout@v4.1.0
     - uses: "./.github/actions/pre-steps"
     - uses: "./.github/actions/setup-ccache"
       with:
-        cache-key-prefix: clang18-tsan
-    - run: COMPILE_WITH_TSAN=1 CC=clang-18 CXX=clang++-18 ROCKSDB_DISABLE_ALIGNED_NEW=1 USE_CLANG=1 make V=1 -j32 check
+        cache-key-prefix: clang21-tsan
+    - run: COMPILE_WITH_TSAN=1 CC=clang-21 CXX=clang++-21 ROCKSDB_DISABLE_ALIGNED_NEW=1 USE_CLANG=1 make V=1 -j32 check
       env:
         CI_SHARD_INDEX: ${{ matrix.shard }}
         CI_TOTAL_SHARDS: 3
@@ -520,8 +542,8 @@ jobs:
     - uses: "./.github/actions/post-steps"
   # ======================== Windows with Tests ======================= #
   # NOTE: some windows jobs are in "nightly" to save resources
-  build-windows-vs2022:
-    if: needs.config.outputs.only_job == 'build-windows-vs2022' || (needs.config.outputs.only_job == '' && github.repository_owner == 'facebook')
+  build-windows-msvc:
+    if: needs.config.outputs.only_job == 'build-windows-msvc' || (needs.config.outputs.only_job == '' && github.repository_owner == 'facebook')
     needs: config
     runs-on: windows-8-core
     strategy:
@@ -532,13 +554,12 @@ jobs:
           suite_run: db_test
           run_java: "false"
         - test_shard: other
-          suite_run: arena_test,db_basic_test,db_test2,db_merge_operand_test,bloom_test,c_test,coding_test,crc32c_test,dynamic_bloom_test,env_basic_test,env_test,hash_test,random_test
+          suite_run: arena_test,db_basic_test,db_etc2_test,db_merge_operand_test,bloom_test,c_test,coding_test,crc32c_test,dynamic_bloom_test,env_basic_test,env_test,hash_test,random_test
           run_java: "false"
         - test_shard: java
           suite_run: ""
           run_java: "true"
     env:
-      CMAKE_GENERATOR: Visual Studio 17 2022
       CMAKE_PORTABLE: 1
     steps:
     - uses: actions/checkout@v4.1.0
@@ -598,7 +619,6 @@ jobs:
     needs: config
     runs-on: macos-15-xlarge
     env:
-      JAVA_HOME: "/Library/Java/JavaVirtualMachines/liberica-jdk-8.jdk/Contents/Home"
       ROCKSDB_DISABLE_JEMALLOC: 1
     steps:
     - uses: actions/checkout@v4.1.0
@@ -610,7 +630,7 @@ jobs:
         cache-key-prefix: macos-java
     - uses: "./.github/actions/increase-max-open-files-on-macos"
     - uses: "./.github/actions/install-gflags-on-macos"
-    - uses: "./.github/actions/install-jdk8-on-macos"
+    - uses: "./.github/actions/setup-jdk-on-macos"
     - uses: "./.github/actions/pre-steps-macos"
     - name: Set Java Environment
       run: |-
@@ -626,8 +646,6 @@ jobs:
     if: needs.config.outputs.only_job == 'build-macos-java-static' || (needs.config.outputs.only_job == '' && github.repository_owner == 'facebook')
     needs: config
     runs-on: macos-15-xlarge
-    env:
-      JAVA_HOME: "/Library/Java/JavaVirtualMachines/liberica-jdk-8.jdk/Contents/Home"
     steps:
     - uses: actions/checkout@v4.1.0
     - uses: maxim-lobanov/setup-xcode@v1.6.0
@@ -638,7 +656,7 @@ jobs:
         cache-key-prefix: macos-java-static
     - uses: "./.github/actions/increase-max-open-files-on-macos"
     - uses: "./.github/actions/install-gflags-on-macos"
-    - uses: "./.github/actions/install-jdk8-on-macos"
+    - uses: "./.github/actions/setup-jdk-on-macos"
     - uses: "./.github/actions/pre-steps-macos"
     - name: Set Java Environment
       run: |-
@@ -654,8 +672,6 @@ jobs:
     if: needs.config.outputs.only_job == 'build-macos-java-static-universal' || (needs.config.outputs.only_job == '' && github.repository_owner == 'facebook')
     needs: config
     runs-on: macos-15-xlarge
-    env:
-      JAVA_HOME: "/Library/Java/JavaVirtualMachines/liberica-jdk-8.jdk/Contents/Home"
     steps:
     - uses: actions/checkout@v4.1.0
     - uses: maxim-lobanov/setup-xcode@v1.6.0
@@ -666,7 +682,7 @@ jobs:
         cache-key-prefix: macos-java-static-universal
     - uses: "./.github/actions/increase-max-open-files-on-macos"
     - uses: "./.github/actions/install-gflags-on-macos"
-    - uses: "./.github/actions/install-jdk8-on-macos"
+    - uses: "./.github/actions/setup-jdk-on-macos"
     - uses: "./.github/actions/pre-steps-macos"
     - name: Set Java Environment
       run: |-

.gitignore

@@ -1,5 +1,7 @@
 make_config.mk
 rocksdb.pc
+.build_signature
+build_tools/__pycache__/
 
 *.a
 *.arc
@@ -52,7 +54,7 @@ GRTAGS
 GTAGS
 rocksdb_dump
 rocksdb_undump
-db_test2
+db_etc2_test
 trace_analyzer
 block_cache_trace_analyzer
 io_tracer_parser
@@ -88,6 +90,8 @@ fbcode/
 fbcode
 buckifier/*.pyc
 buckifier/__pycache__
+tools/__pycache__/
+tools/c_api_gen/__pycache__/
 .arcconfig
 
 compile_commands.json
@@ -102,3 +106,11 @@ cmake-build-*
 third-party/folly/
 .cache
 *.sublime-*
+
+# Claude Code local settings
+.claude/settings.local.json
+
+tools/__pycache__/
+# Keep documentation trackable even if broader ignores match names like "*_test".
+!docs/
+!docs/**

AGENTS.md

@@ -0,0 +1,9 @@
+# Agent Instructions
+
+This repository's authoritative agent instructions live in `CLAUDE.md`.
+
+Read and follow [`CLAUDE.md`](./CLAUDE.md) in full before making changes or
+reviewing code in this checkout.
+
+If there is any ambiguity between this file and `CLAUDE.md`, `CLAUDE.md` takes
+precedence.

BUCK

@@ -32,12 +32,15 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "db/blob/blob_file_cache.cc",
         "db/blob/blob_file_garbage.cc",
         "db/blob/blob_file_meta.cc",
+        "db/blob/blob_file_partition_manager.cc",
         "db/blob/blob_file_reader.cc",
         "db/blob/blob_garbage_meter.cc",
+        "db/blob/blob_gen2_format.cc",
         "db/blob/blob_log_format.cc",
         "db/blob/blob_log_sequential_reader.cc",
         "db/blob/blob_log_writer.cc",
         "db/blob/blob_source.cc",
+        "db/blob/blob_write_batch_transformer.cc",
         "db/blob/prefetch_buffer_collection.cc",
         "db/builder.cc",
         "db/c.cc",
@@ -109,6 +112,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "db/version_util.cc",
         "db/wal_edit.cc",
         "db/wal_manager.cc",
+        "db/wide/read_path_blob_resolver.cc",
         "db/wide/wide_column_serialization.cc",
         "db/wide/wide_columns.cc",
         "db/wide/wide_columns_helper.cc",
@@ -331,6 +335,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "utilities/secondary_index/simple_secondary_index.cc",
         "utilities/simulator_cache/cache_simulator.cc",
         "utilities/simulator_cache/sim_cache.cc",
+        "utilities/sorted_run_builder/sorted_run_builder.cc",
         "utilities/table_properties_collectors/compact_for_tiering_collector.cc",
         "utilities/table_properties_collectors/compact_on_deletion_collector.cc",
         "utilities/trace/file_trace_reader_writer.cc",
@@ -374,10 +379,10 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "utilities/write_batch_with_index/write_batch_with_index_internal.cc",
     ], deps=[
         "//folly/container:f14_hash",
-        "//folly/experimental/coro:blocking_wait",
-        "//folly/experimental/coro:collect",
-        "//folly/experimental/coro:coroutine",
-        "//folly/experimental/coro:task",
+        "//folly/coro:blocking_wait",
+        "//folly/coro:collect",
+        "//folly/coro:coroutine",
+        "//folly/coro:task",
         "//folly/synchronization:distributed_mutex",
     ], headers=glob(["**/*.h"]), link_whole=False, extra_test_libs=False)
 
@@ -4804,6 +4809,12 @@ cpp_unittest_wrapper(name="db_blob_corruption_test",
             extra_compiler_flags=[])
 
 
+cpp_unittest_wrapper(name="db_blob_direct_write_test",
+            srcs=["db/blob/db_blob_direct_write_test.cc"],
+            deps=[":rocksdb_test_lib"],
+            extra_compiler_flags=[])
+
+
 cpp_unittest_wrapper(name="db_blob_index_test",
             srcs=["db/blob/db_blob_index_test.cc"],
             deps=[":rocksdb_test_lib"],
@@ -4858,6 +4869,12 @@ cpp_unittest_wrapper(name="db_encryption_test",
             extra_compiler_flags=[])
 
 
+cpp_unittest_wrapper(name="db_etc2_test",
+            srcs=["db/db_etc2_test.cc"],
+            deps=[":rocksdb_test_lib"],
+            extra_compiler_flags=[])
+
+
 cpp_unittest_wrapper(name="db_etc3_test",
             srcs=["db/db_etc3_test.cc"],
             deps=[":rocksdb_test_lib"],
@@ -4942,6 +4959,12 @@ cpp_unittest_wrapper(name="db_merge_operator_test",
             extra_compiler_flags=[])
 
 
+cpp_unittest_wrapper(name="db_open_with_config_test",
+            srcs=["db/db_open_with_config_test.cc"],
+            deps=[":rocksdb_test_lib"],
+            extra_compiler_flags=[])
+
+
 cpp_unittest_wrapper(name="db_options_test",
             srcs=["db/db_options_test.cc"],
             deps=[":rocksdb_test_lib"],
@@ -5008,12 +5031,6 @@ cpp_unittest_wrapper(name="db_test",
             extra_compiler_flags=[])
 
 
-cpp_unittest_wrapper(name="db_test2",
-            srcs=["db/db_test2.cc"],
-            deps=[":rocksdb_test_lib"],
-            extra_compiler_flags=[])
-
-
 cpp_unittest_wrapper(name="db_universal_compaction_test",
             srcs=["db/db_universal_compaction_test.cc"],
             deps=[":rocksdb_test_lib"],
@@ -5032,6 +5049,12 @@ cpp_unittest_wrapper(name="db_wide_basic_test",
             extra_compiler_flags=[])
 
 
+cpp_unittest_wrapper(name="db_wide_blob_direct_write_test",
+            srcs=["db/wide/db_wide_blob_direct_write_test.cc"],
+            deps=[":rocksdb_test_lib"],
+            extra_compiler_flags=[])
+
+
 cpp_unittest_wrapper(name="db_with_timestamp_basic_test",
             srcs=["db/db_with_timestamp_basic_test.cc"],
             deps=[":rocksdb_test_lib"],
@@ -5532,6 +5555,12 @@ cpp_unittest_wrapper(name="slice_transform_test",
             extra_compiler_flags=[])
 
 
+cpp_unittest_wrapper(name="sorted_run_builder_test",
+            srcs=["utilities/sorted_run_builder/sorted_run_builder_test.cc"],
+            deps=[":rocksdb_test_lib"],
+            extra_compiler_flags=[])
+
+
 cpp_unittest_wrapper(name="sst_dump_test",
             srcs=["tools/sst_dump_test.cc"],
             deps=[":rocksdb_test_lib"],
@@ -5767,3 +5796,5 @@ cpp_unittest_wrapper(name="write_unprepared_transaction_test",
 
 
 export_file(name = "tools/db_crashtest.py")
+
+export_file(name = "tools/fault_injection_log_parser.py")

CLAUDE.md

@@ -38,7 +38,7 @@ This document provides guidance for generating and reviewing code in the RocksDB
 
 **SIMD and Vectorization:** Leverage SIMD instructions (SSE, AVX) for data-parallel operations when appropriate. Structure data to enable auto-vectorization by the compiler. Consider explicit SIMD intrinsics for critical hot paths like checksum computation, encoding/decoding, and bulk data processing.
 
-**Branch Prediction:** Minimize unpredictable branches in hot paths. Use `LIKELY`/`UNLIKELY` macros to hint branch prediction. Consider branchless alternatives for simple conditionals. Order switch cases and if-else chains by frequency.
+**Branch Prediction:** Minimize unpredictable branches in hot paths. Use `LIKELY`/`UNLIKELY` macros to hint branch prediction, e.g. for error cases and other rare or otherwise costly cases, but NOT for predicting popular configurations. Consider branchless alternatives for simple conditionals. Order switch cases and if-else chains by frequency.
 
 **Memory and Resource Management:** Be mindful of memory allocations, especially in hot paths. Use RAII patterns, smart pointers, and RocksDB's memory management utilities appropriately.
 
@@ -149,6 +149,26 @@ Cache management is critical for RocksDB's performance.
 
 When reviewing RocksDB code (or preparing code for review), use this checklist:
 
+### Contract Boundaries
+- [ ] Is each behavior owned by the right layer? High-level policy (for example,
+  "compaction wants this I/O mode") should live at the caller/policy layer, while
+  lower layers should expose generic mechanisms (for example, "open a fresh
+  reader", "skip shared cache insertion", or "use these FileOptions").
+- [ ] Do comments and names describe local contracts rather than leaking a
+  specific caller's rationale into reusable APIs? Generic code should not need to
+  know about one current use case unless the API itself is intentionally
+  use-case-specific.
+- [ ] Does each flag or parameter control one coherent behavior? If one boolean
+  starts implying ownership, cache policy, I/O mode, prefetching, and caller
+  identity, split it into explicit flags or an options struct.
+- [ ] Could a future caller use this lower-level API without accidentally
+  inheriting assumptions from compaction, backup, user reads, or a particular
+  table format? If not, tighten the contract with assertions, clearer names, or
+  a narrower API.
+- [ ] Are implementation details not being used as policy signals? Prefer an
+  explicit contract over inferring behavior from incidental fields such as file
+  options, cache handles, or current table-reader state.
+
 ### Correctness
 - [ ] Does the change preserve database semantics (e.g., snapshot isolation, key ordering)?
 - [ ] Are all error cases handled appropriately?
@@ -204,18 +224,93 @@ The following patterns emerged as frequent sources of review feedback:
 
 10. **Platform Compatibility:** Ensure changes work correctly on all supported platforms (Linux, Windows, macOS) and with all supported compilers (GCC, Clang, MSVC).
 
+11. **Contract Boundary Leaks:** When a change plumbs a new option or use-case
+specific behavior through multiple subsystems, review the call chain for
+contract leaks. Caller-specific rationale belongs at the call site or public API
+documentation; reusable layers should expose precise, layer-local capabilities.
+Watch especially for comments mentioning one caller in generic code, booleans
+that silently bundle several behaviors, and downstream code inferring policy
+from an implementation detail instead of an explicit option.
+
 ---
 
 ## Important tips
 
 ### Build system
-* There are 3 build system. Make, CMake, BUCK(meta internal).
+* There are 3 build system. Make for git clones, BUCK (meta internal) for hg
+  clones, and CMake for some special cases.
 * When a new .cc file is added, update Makefile, CMakeLists.txt, src.mk, BUCK.
 * Don't manually edit BUCK file, after updating src.mk, run
     /usr/local/bin/python3 buckifier/buckify_rocksdb.py to update it
-* Use make to build and run the test. CMake and BUCK are not used locally.
-* Use `make dbg` command to build all of the unit test in debug mode.
 * For -j in make command, use the number of CPU cores to decide it.
+* When searching for references to something (a symbol, library, etc.), do not
+  restrict or truncate your search based on presumed relevance or scope. It is
+  important and time-saving to keep the repo reasonably consistent across
+  different build systems, programming languages, and even between
+  documentation and implementation.
+
+### Avoiding mixed build modes with Make (use `AUTO_CLEAN=1`)
+
+Object files are written to the same paths regardless of build flags, so
+reusing objects from a prior build with different flags causes confusing
+linker errors, etc. This problem is essentially avoidable by ALWAYS using
+`AUTO_CLEAN=1 make -j<n> <something>` for manual make invocations. This
+will automatically clean object files if the build parameters/flavor have
+changed. The `build_tools/rockstest.sh` / `rocksptest.sh` helpers described
+below set `AUTO_CLEAN=1` for you.
+
+### Source checks
+* Run `make check-sources` before committing. This catches non-ASCII
+    characters in source files and other source-level issues that CI will
+    reject. In particular, **do not use Unicode characters** (em dashes,
+    smart quotes, etc.) in comments or strings -- use ASCII equivalents
+    (`--` instead of em dash, `'` instead of smart quote, etc.).
+
+### License headers
+* Every new source file needs a license header. For a file that does **not**
+    carry an outside/third-party copyright, use the standard Meta dual-licensed
+    header (the dual-license designation is required -- a bare
+    "All Rights Reserved" copyright is not an acceptable open-source header):
+    ```
+    //  Copyright (c) Meta Platforms, Inc. and affiliates.
+    //  This source code is licensed under both the GPLv2 (found in the
+    //  COPYING file in the root directory) and Apache 2.0 License
+    //  (found in the LICENSE.Apache file in the root directory).
+    ```
+    Use a `#` comment prefix instead of `//` for shell, Python, and Makefile
+    fragments.
+* Files derived from an external source (e.g. LevelDB) keep their original
+    upstream copyright line in addition to the header above.
+
+### RTTI and dynamic_cast
+* Production code and `db_stress` must build in **release mode
+    (`-fno-rtti`)**. Do not use `dynamic_cast` anywhere except unit tests.
+    Use `static_cast_with_check` from `util/cast_util.h` (validates with
+    `dynamic_cast` in debug builds, plain `static_cast` in release).
+* Unit tests (`*_test.cc`) are built in debug mode with RTTI enabled.
+
+### Cross-platform / portability
+Local `make` only exercises Linux with GCC/Clang, but CI
+(`.github/workflows/pr-jobs.yml` and `nightly.yml`) gates on a much wider
+matrix, so portability breaks are invisible locally until CI fails. Code must
+build (and where noted, run tests) across:
+
+| Axis | Must support |
+|------|--------------|
+| OS | Linux (x86_64 + ARM), macOS, Windows |
+| Compiler | GCC, Clang (libstdc++ **and** libc++), AppleClang, **MSVC (VS2022)**, MinGW (Linux cross-compile, build-only, no gflags) |
+| Build system | Make, CMake, and BUCK (internal) -- keep all in sync (see "Build system" above) |
+| Config | release (`-fno-rtti`), `ASSERT_STATUS_CHECKED`, ASAN/UBSAN/TSAN, folly, unity build, JNI/Java |
+
+Treat these as constraints to satisfy and infer the specifics from them before
+adding any system header, libc call, or compiler-specific construct. The most
+common trap: anything that compiles under GCC/Clang on Linux but not under
+**MSVC/MinGW** -- e.g. unguarded POSIX-only headers/functions (`<unistd.h>`,
+`<sys/*.h>`, `getpid`, `_exit`, ...) or GCC/Clang extensions
+(`__attribute__`, `__builtin_*`, VLAs, `alloca`). Prefer the `port::`/`Env`
+abstractions; otherwise guard with `#ifdef OS_WIN` (POSIX `<unistd.h>` ->
+Windows `<process.h>`). Because libc++ is also tested, include what you use
+rather than relying on libstdc++ transitive includes.
 
 ### Unit Test
 * After all of the unit tests are added, review them and try to extract common
@@ -224,9 +319,26 @@ The following patterns emerged as frequent sources of review feedback:
 * Don't use sleep to wait for certain events to happen. This will cause test to
     be flaky. Instead, use sync point to synchronize thread progress.
 * Cap unit test execution with 60 seconds timeout.
-* When there are multiple unit tests need to be executed, try to use
-    gtest_parallel.py if available. E.g.
-    python3 ${GTEST_PARALLEL}/gtest_parallel.py ./table_test
+* To build and run unit tests locally, prefer these helper scripts:
+    * `build_tools/rocksptest.sh <test_binary> [more_binaries...] [args...]`
+        builds the binary(ies) with parallel make and `AUTO_CLEAN=1` and runs
+        them under gtest-parallel, sharding the test cases across CPUs. Prefer
+        this whenever running more than a couple of test cases, e.g.
+        `build_tools/rocksptest.sh table_test` or
+        `build_tools/rocksptest.sh db_test env_test --gtest_filter=*Foo*`.
+    * `build_tools/rockstest.sh <test_binary> [args...]` builds with parallel
+        make and `AUTO_CLEAN=1` and runs the binary directly (serially).
+        Use it only for a very small number of test cases, e.g.
+        `build_tools/rockstest.sh db_test --gtest_filter=*MixedSlowdown*`.
+* After writing a test, stress-test for flakiness (AUTO_CLEAN handles the
+    rebuild needed by the `COERCE_CONTEXT_SWITCH=1` flag change):
+    ```bash
+    COERCE_CONTEXT_SWITCH=1 build_tools/rockstest.sh {test_binary} -r100 \
+        --gtest_filter="*YourTestName*"
+    ```
+* For CI-style flaky tests that do not reproduce with `gtest_parallel.py`,
+    `--gtest_repeat`, or normal coerce-mode runs, inspect
+    `tools/gtest_parallel_repro.py --help`.
 
 ### Unit test dedup guidelines
 * Extract helper functions for repeated patterns such as object
@@ -261,6 +373,14 @@ The following patterns emerged as frequent sources of review feedback:
 ### Stress test
 * When adding a new feature, make sure stress test covers the new option.
 
+### Component docs
+* For component-level design notes and implementation walkthroughs, start with
+    `docs/components/index.md`.
+* Documentation under `docs/components/` is organized by subsystem in
+    `docs/components/<area>/`.
+* Each subsystem directory should have an `index.md` entry point plus focused
+    chapter files for deeper topics.
+
 ### DB bench update
 * When adding a performance related feature, support it in db_bench
 
@@ -271,9 +391,13 @@ The following patterns emerged as frequent sources of review feedback:
 * Blog post authors must be defined in `docs/_data/authors.yml` to be displayed
 
 ### Final verification of the change
-* Execute make clean to clean all of the changes.
-* Execute make check to build all of the changes and execute all of the tests.
-    Note that executing all of the tests could take multiple minutes.
+* Execute `AUTO_CLEAN=1 make check` to build all of the changes and execute all
+    of the tests. `AUTO_CLEAN=1` ensures a clean rebuild if your previous build
+    used different parameters. Note that executing all of the tests could take
+    multiple minutes.
+* Run `AUTO_CLEAN=1 ASSERT_STATUS_CHECKED=1 make check` to verify all Status
+    objects are properly checked. This catches missing error handling that can
+    lead to silent data corruption.
 
 ### Monitoring make check progress
 * Use `make check-progress` to get machine-parseable JSON progress while
@@ -281,7 +405,7 @@ The following patterns emerged as frequent sources of review feedback:
     builds without timeout issues.
 * Run `make check` in background, then poll progress:
     ```bash
-    make check &
+    AUTO_CLEAN=1 make check &
     # Poll periodically:
     make check-progress
     ```
@@ -306,9 +430,10 @@ The following patterns emerged as frequent sources of review feedback:
 
 ### Executing benchmark using db_bench
 * Since the goal is to measure performance, we need to build a release binary
-    using `make clean && DEBUG_LEVEL=0 make db_bench`. If there is an engine
-    crash due to bug, we need to switch back to debug build. Make sure to run
-    `make clean` before running `make dbg`.
+    using `AUTO_CLEAN=1 DEBUG_LEVEL=0 make db_bench`. If there is an engine
+    crash due to a bug, switch back to a debug build with
+    `AUTO_CLEAN=1 make dbg`; `AUTO_CLEAN=1` handles the release<->debug rebuild
+    automatically.
 
 ### Formatting code
 * After making change, use `make format-auto` to auto-apply formatting without

CMakeLists.txt

@@ -420,13 +420,6 @@ if(WITH_NUMA)
   list(APPEND THIRDPARTY_LIBS NUMA::NUMA)
 endif()
 
-option(WITH_TBB "build with Threading Building Blocks (TBB)" OFF)
-if(WITH_TBB)
-  find_package(TBB REQUIRED)
-  add_definitions(-DTBB)
-  list(APPEND THIRDPARTY_LIBS TBB::TBB)
-endif()
-
 # Stall notifications eat some performance from inserts
 option(DISABLE_STALL_NOTIF "Build with stall notifications" OFF)
 if(DISABLE_STALL_NOTIF)
@@ -654,6 +647,8 @@ if(USE_FOLLY)
     FMT_INST_PATH)
     exec_program(ls ARGS -d ${FOLLY_INST_PATH}/../gflags* OUTPUT_VARIABLE
     GFLAGS_INST_PATH)
+    exec_program(ls ARGS -d ${FOLLY_INST_PATH}/../glog* OUTPUT_VARIABLE
+    GLOG_INST_PATH)
     set(Boost_DIR ${BOOST_INST_PATH}/lib/cmake/Boost-1.83.0)
     if(EXISTS ${FMT_INST_PATH}/lib64)
       set(fmt_DIR ${FMT_INST_PATH}/lib64/cmake/fmt)
@@ -661,19 +656,89 @@ if(USE_FOLLY)
       set(fmt_DIR ${FMT_INST_PATH}/lib/cmake/fmt)
     endif()
     set(gflags_DIR ${GFLAGS_INST_PATH}/lib/cmake/gflags)
+    if(EXISTS ${GLOG_INST_PATH}/lib64/cmake/glog)
+      set(GLOG_CMAKE_DIR ${GLOG_INST_PATH}/lib64/cmake/glog)
+    elseif(EXISTS ${GLOG_INST_PATH}/lib/cmake/glog)
+      set(GLOG_CMAKE_DIR ${GLOG_INST_PATH}/lib/cmake/glog)
+    endif()
+    if(NOT GLOG_CMAKE_DIR)
+      find_library(GETDEPS_GLOG_LIBRARY NAMES glogd glog
+        PATHS ${GLOG_INST_PATH}/lib64 ${GLOG_INST_PATH}/lib
+        NO_DEFAULT_PATH)
+      if(NOT GETDEPS_GLOG_LIBRARY OR
+         NOT EXISTS ${GLOG_INST_PATH}/include/glog/logging.h)
+        message(FATAL_ERROR "Could not find getdeps glog under "
+          "${GLOG_INST_PATH}")
+      endif()
+      set(GLOG_CMAKE_DIR ${CMAKE_CURRENT_BINARY_DIR}/getdeps-glog-cmake)
+      file(MAKE_DIRECTORY ${GLOG_CMAKE_DIR})
+      file(WRITE ${GLOG_CMAKE_DIR}/glog-config.cmake
+        "if(NOT TARGET glog::glog)\n"
+        "  add_library(glog::glog UNKNOWN IMPORTED)\n"
+        "  set_target_properties(glog::glog PROPERTIES\n"
+        "    IMPORTED_LOCATION \"${GETDEPS_GLOG_LIBRARY}\"\n"
+        "    INTERFACE_INCLUDE_DIRECTORIES \"${GLOG_INST_PATH}/include\")\n"
+        "endif()\n"
+        "set(glog_FOUND TRUE)\n"
+        "set(Glog_FOUND TRUE)\n"
+        "set(GLOG_FOUND TRUE)\n"
+        "set(GLOG_LIBRARY \"${GETDEPS_GLOG_LIBRARY}\")\n"
+        "set(GLOG_LIBRARIES \"${GETDEPS_GLOG_LIBRARY}\")\n"
+        "set(GLOG_INCLUDE_DIR \"${GLOG_INST_PATH}/include\")\n")
+    endif()
+    set(glog_DIR ${GLOG_CMAKE_DIR})
+    set(Glog_DIR ${GLOG_CMAKE_DIR})
+    # Unlike gflags, glog may also be installed system-wide. Pre-resolve the
+    # getdeps copy so folly-config.cmake cannot fall back to another version.
+    find_package(glog CONFIG REQUIRED PATHS ${GLOG_CMAKE_DIR}
+                 NO_DEFAULT_PATH)
 
     exec_program(sed ARGS -i 's/gflags_shared//g'
     ${FOLLY_INST_PATH}/lib/cmake/folly/folly-targets.cmake)
 
     include(${FOLLY_INST_PATH}/lib/cmake/folly/folly-config.cmake)
 
-    # Fix gflags library name for debug builds
+    # Fix gflags library name for debug builds. The getdeps gflags shared
+    # library is versioned (e.g. libgflags_debug.so.2.3), so resolve the
+    # actual file by glob rather than hard-coding a version.
     if(CMAKE_BUILD_TYPE STREQUAL "Debug")
+      file(GLOB GFLAGS_DEBUG_SHARED_LIBS
+        "${GFLAGS_INST_PATH}/lib/libgflags_debug.so.*")
+      if(NOT GFLAGS_DEBUG_SHARED_LIBS)
+        message(FATAL_ERROR
+          "Could not find getdeps libgflags_debug.so under ${GFLAGS_INST_PATH}/lib")
+      endif()
+      list(GET GFLAGS_DEBUG_SHARED_LIBS 0 GFLAGS_DEBUG_SHARED_LIB)
       set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -Wl,-rpath=${GFLAGS_INST_PATH}/lib")
-      set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} ${GFLAGS_INST_PATH}/lib/libgflags_debug.so.2.2")
+      set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} ${GFLAGS_DEBUG_SHARED_LIB}")
     endif()
   endif()
 
+  # Folly itself uses gflags transitively, but RocksDB tools and benchmarks
+  # also call gflags APIs directly, so they need an explicit link dependency.
+  if(WITH_GFLAGS)
+    if(NOT TARGET gflags::gflags AND NOT TARGET gflags_shared AND
+       NOT gflags_LIBRARIES)
+      find_package(gflags CONFIG QUIET)
+      if(NOT gflags_FOUND)
+        find_package(gflags REQUIRED)
+      endif()
+    endif()
+    if(TARGET gflags::gflags)
+      set(GFLAGS_LIB gflags::gflags)
+    elseif(TARGET gflags_shared)
+      set(GFLAGS_LIB gflags_shared)
+    elseif(DEFINED GFLAGS_TARGET AND TARGET ${GFLAGS_TARGET})
+      set(GFLAGS_LIB ${GFLAGS_TARGET})
+    elseif(gflags_LIBRARIES)
+      set(GFLAGS_LIB ${gflags_LIBRARIES})
+    else()
+      message(FATAL_ERROR
+        "WITH_GFLAGS is enabled, but no gflags library could be resolved")
+    endif()
+    list(APPEND THIRDPARTY_LIBS ${GFLAGS_LIB})
+  endif()
+
   add_compile_definitions(USE_FOLLY FOLLY_NO_CONFIG HAVE_CXX11_ATOMIC)
   list(APPEND THIRDPARTY_LIBS Folly::folly)
   set(FOLLY_LIBS Folly::folly)
@@ -707,14 +772,17 @@ set(SOURCES
         db/blob/blob_file_addition.cc
         db/blob/blob_file_builder.cc
         db/blob/blob_file_cache.cc
+        db/blob/blob_file_partition_manager.cc
         db/blob/blob_file_garbage.cc
         db/blob/blob_file_meta.cc
         db/blob/blob_file_reader.cc
         db/blob/blob_garbage_meter.cc
+        db/blob/blob_gen2_format.cc
         db/blob/blob_log_format.cc
         db/blob/blob_log_sequential_reader.cc
         db/blob/blob_log_writer.cc
         db/blob/blob_source.cc
+        db/blob/blob_write_batch_transformer.cc
         db/blob/prefetch_buffer_collection.cc
         db/builder.cc
         db/c.cc
@@ -786,6 +854,7 @@ set(SOURCES
         db/version_util.cc
         db/wal_edit.cc
         db/wal_manager.cc
+        db/wide/read_path_blob_resolver.cc
         db/wide/wide_column_serialization.cc
         db/wide/wide_columns.cc
         db/wide/wide_columns_helper.cc
@@ -967,6 +1036,7 @@ set(SOURCES
         utilities/cassandra/merge_operator.cc
         utilities/checkpoint/checkpoint_impl.cc
         utilities/compaction_filters.cc
+        utilities/sorted_run_builder/sorted_run_builder.cc
         utilities/compaction_filters/remove_emptyvalue_compactionfilter.cc
         utilities/counted_fs.cc
         utilities/debug.cc
@@ -1136,8 +1206,23 @@ if(USE_FOLLY_LITE)
   FMT_INCLUDE_DIR)
   include_directories(${FMT_INCLUDE_DIR})
 
+  exec_program(python3 ${PROJECT_SOURCE_DIR}/third-party/folly ARGS
+  build/fbcode_builder/getdeps.py show-inst-dir glog OUTPUT_VARIABLE
+  GLOG_INST_PATH)
+  if(EXISTS ${GLOG_INST_PATH}/lib64)
+    set(GLOG_LIB_DIR ${GLOG_INST_PATH}/lib64)
+  else()
+    set(GLOG_LIB_DIR ${GLOG_INST_PATH}/lib)
+  endif()
+
   add_definitions(-DUSE_FOLLY -DFOLLY_NO_CONFIG)
-  list(APPEND THIRDPARTY_LIBS glog)
+  find_library(GLOG_LIBRARY NAMES glog PATHS ${GLOG_LIB_DIR} NO_DEFAULT_PATH)
+  if(NOT GLOG_LIBRARY)
+    find_library(GLOG_LIBRARY NAMES glog)
+  endif()
+  if(GLOG_LIBRARY)
+    list(APPEND THIRDPARTY_LIBS ${GLOG_LIBRARY})
+  endif()
 endif()
 
 set(ROCKSDB_STATIC_LIB rocksdb${ARTIFACT_SUFFIX})
@@ -1383,6 +1468,7 @@ if(WITH_TESTS)
         db/blob/blob_garbage_meter_test.cc
         db/blob/blob_source_test.cc
         db/blob/db_blob_basic_test.cc
+        db/blob/db_blob_direct_write_test.cc
         db/blob/db_blob_compaction_test.cc
         db/blob/db_blob_corruption_test.cc
         db/blob/db_blob_index_test.cc
@@ -1408,6 +1494,7 @@ if(WITH_TESTS)
         db/db_clip_test.cc
         db/db_dynamic_level_test.cc
         db/db_encryption_test.cc
+        db/db_etc2_test.cc
         db/db_etc3_test.cc
         db/db_flush_test.cc
         db/db_inplace_update_test.cc
@@ -1420,6 +1507,7 @@ if(WITH_TESTS)
         db/db_memtable_test.cc
         db/db_merge_operator_test.cc
         db/db_merge_operand_test.cc
+        db/db_open_with_config_test.cc
         db/db_options_test.cc
         db/db_properties_test.cc
         db/db_range_del_test.cc
@@ -1430,7 +1518,6 @@ if(WITH_TESTS)
         db/db_table_properties_test.cc
         db/db_tailing_iter_test.cc
         db/db_test.cc
-        db/db_test2.cc
         db/db_logical_block_size_cache_test.cc
         db/db_universal_compaction_test.cc
         db/db_wal_test.cc
@@ -1472,6 +1559,7 @@ if(WITH_TESTS)
         db/wal_manager_test.cc
         db/wal_edit_test.cc
         db/wide/db_wide_basic_test.cc
+        db/wide/db_wide_blob_direct_write_test.cc
         db/wide/wide_column_serialization_test.cc
         db/wide/wide_columns_helper_test.cc
         db/write_batch_test.cc
@@ -1552,6 +1640,7 @@ if(WITH_TESTS)
         utilities/cassandra/cassandra_row_merge_test.cc
         utilities/cassandra/cassandra_serialize_test.cc
         utilities/checkpoint/checkpoint_test.cc
+        utilities/sorted_run_builder/sorted_run_builder_test.cc
         utilities/env_timed_test.cc
         utilities/fault_injection_fs_test.cc
         utilities/memory/memory_test.cc

HISTORY.md

@@ -1,6 +1,127 @@
 # Rocksdb Change Log
 > NOTE: Entries for next release do not go here. Follow instructions in `unreleased_history/README.txt`
 
+## 11.6.0 (07/02/2026)
+### New Features
+* Added EXPERIMENTAL embedded blob SST support through `SstFileWriter::OpenWithEmbeddedBlobs()`, storing eligible large values as same-file blob records in block-based SST files and resolving them transparently for reads. This niche feature currently supports uncompressed embedded blobs only; compression options are placeholders and compression support is deferred to follow-up work.
+
+### Public API Changes
+* Expanded the C API (`include/rocksdb/c.h`) with a large set of new `rocksdb_*` functions, mostly option getters/setters plus table-properties, job/event-listener, and metadata accessors, a WAL filter, a ReadOptions table filter, and a backup exclude-files callback. Many are now produced by a new semi-automated generator (`tools/c_api_gen/`) from the C++ headers; `include/rocksdb/c.h` remains a single self-contained header and the signatures of pre-existing functions are unchanged.
+
+### Bug Fixes
+* Reverted PR14831 that made range_lock_manager aware of reverse-order CF
+* Fixed a bug in `RandomAccessFileReader::ReadAsync` where an already-aligned direct-IO read request with a null `scratch` and a caller-provided `aligned_buf` would take the "already aligned" fast path and submit the null buffer to the underlying async read (e.g. a null iovec base to io_uring, failing with EFAULT). This could surface as spurious iterator failures during async prefetch (MultiScan with async IO) on direct-IO databases. The async path now allocates a backing buffer in this case, matching the synchronous `Read` path.
+* Fixed a bug where closing a read-only DB instance could delete live SST files created by a concurrent read-write DB sharing the same directory.
+
+## 11.5.0 (06/16/2026)
+### New Features
+* External table readers that open files through `ExternalTableOptions::fs` now update RocksDB SST/file-read statistics and file IO listener callbacks, making external file IO activity visible in existing read metrics.
+* Added `DB::PrepareFileIngestion()`, a two-phase form of `IngestExternalFile`/`IngestExternalFiles`. `PrepareFileIngestion()` performs all of the work that does not require the DB mutex (validating the arguments, reading each external file's metadata, and linking/copying the files into the DB) and returns an opaque `FileIngestionHandle`. `DB::CommitFileIngestionHandle()` (or `DB::CommitFileIngestionHandles()` for several handles at once) then makes the prepared files visible; multiple handles are committed atomically in a single MANIFEST write, or none are. `FileIngestionHandle::Abort()` (or simply destroying the handle) cancels a prepared ingestion and rolls it back. This gives flexibility to the application to Prepare a file separately from when its committed, and may shorten an applications critical path on `IngestExternalFile`.
+* Added two stats histograms reporting the per-call latency (in microseconds) of each successful `IngestExternalFile`/`IngestExternalFiles` call, split by phase: `rocksdb.ingest.external.file.prepare.micros` (argument validation and reading/validating/linking the external files, which does not block live writes) and `rocksdb.ingest.external.file.run.micros` (the ingestion performed under the DB mutex while live writes are blocked). Failed ingestions are not recorded.
+* Added `DBOptions::use_direct_io_for_compaction_reads` (default false). When enabled, compaction-input SST reads use `O_DIRECT` while user reads remain buffered, avoiding page-cache eviction of hot user-read data by sequential compaction scans. Pair with `use_direct_io_for_flush_and_compaction = true` on write-heavy workloads for direct I/O on both compaction inputs and outputs. Rejected at Open when combined with `allow_mmap_reads = true`. No-op when `use_direct_reads = true` is already set (since all reads are already direct).
+
+### Public API Changes
+* `DB::GetPreparedFileInfoForExternalSstIngestion()` can now prepare metadata for a live DB-generated SST file so it can be passed to `IngestExternalFileArg::file_infos`. The input path must exactly match a live table file owned by the source DB, and the returned metadata handle must outlive the ingestion.
+* `ExternalSstFileInfo` gained a `prepared_file_info` member produced by `SstFileWriter::Finish`, and `IngestExternalFileArg` gained a `file_infos` field: a vector of borrowed `const PreparedFileInfo*` pointers, parallel to `external_files`. The owning `ExternalSstFileInfo::prepared_file_info` handle must outlive the ingestion. When set, `IngestExternalFiles()` reuses the supplied per-file metadata instead of re-opening and scanning each file to recompute it, avoiding that extra I/O.
+* Corrected the public C++ option spelling from `memtable_veirfy_per_key_checksum_on_seek` to `memtable_verify_per_key_checksum_on_seek`. RocksDB continues to accept the old misspelled OPTIONS-file key for compatibility, while newly serialized OPTIONS files use the corrected key.
+* Added experimental read-scoped block buffer provider API, configured through `ReadOptions::read_scoped_block_buffer_provider`, for supported block-based table iterator scans and MultiScan reads to use caller-provided read-scoped storage for final data-block contents. When configured, supported provider-backed scan data-block reads bypass the data-block cache. The provider is ignored when mmap reads are enabled. RocksDB may still use ordinary temporary scratch for serialized block bytes, such as when a block may be compressed. Get and MultiGet do not currently provide API guarantees for this provider.
+* Added `FileSystem::SyncFile()` and `Env::SyncFile()` public APIs for syncing or fsyncing a file by name without requiring callers to reopen it as writable.
+
+### Behavior Changes
+* The default `compression` for column families is now `kLZ4Compression` rather than `kSnappyCompression`. This affects only column families that do not explicitly set `compression`, and only newly-written SST files. The change is fully compatible: existing data remains readable with no migration needed, as RocksDB selects the decompressor per block. LZ4 offers slightly better compression ratios and decompression CPU efficiency vs. Snappy and similar compression CPU, tested across various server CPUs. When support is not compiled in, the fallback path for default compression is LZ4 -> Snappy -> NoCompression.
+* Disable parallel compression (ignore `CompressionOptions::parallel_threads`) for fast built-in compressors: Snappy, LZ4 (accelerated, not LZ4HC), and accelerated levels of ZSTD (level < 0). These do not generally benefit from parallel compression. This behavior can be overridden with a custom Compressor from a custom CompressionManager.
+* `PosixFileSystem::OptimizeForCompactionTableRead` now delegates to the base `FileSystem::OptimizeForCompactionTableRead` implementation before applying its Linux-only compaction-readahead clamp (added for #12038). The returned `FileOptions::use_direct_reads` therefore reflects `DBOptions::use_direct_reads`, consistent with the base class and with other `FileSystem` implementations. Previously the override ignored the base implementation and keyed both the returned options and the Linux readahead clamp off the incoming `FileOptions` rather than `DBOptions`. At the in-tree call site the incoming `FileOptions::use_direct_reads` already matches `DBOptions::use_direct_reads`, so this is effectively a no-op for existing configurations; it removes a latent inconsistency where a caller passing `FileOptions` whose `use_direct_reads` disagreed with `DBOptions` would have had the global flag silently ignored for compaction-input reads on Linux. (Note: the `#12038` readahead clamp still keys off the global `use_direct_reads`; it is not skipped for the compaction-only `use_direct_io_for_compaction_reads` path, since that flag enables O_DIRECT after this hook runs.)
+* Brought more unity to `kLZ4Compression` and `kLZ4HCCompression` by giving each access to the other's compression levels. Negative (fast) values previously only available to LZ4 and positive (slow) values previously only available to LZ4HC are now available to both, so configuring a non-default `compression_opts.level` now selects the LZ4 compressor variant. This is a behavior change for previously tolerated but dubious configurations such as positive compression level with `kLZ4Compression`. `kLZ4Compression` and `kLZ4HCCompression` keep their effective default levels, equivalent to -1 and 9 respectively.
+* Removed a discontinuity in `kZSTD` compression levels at `compression_opts.level == 0` by mapping that setting to `level = -1` rather than to `level = 3` as the ZSTD library does internally. This improves the compression auto-tuning landscape.
+
+### Bug Fixes
+* Fixed a bug where the range lock manager would crash with an assertion failure when using a reverse comparator column family.
+
+### Performance Improvements
+* Reduced commit latency when ingesting many external files by allowing `IngestExternalFileOptions::file_opening_threads` to open table readers for committed ingested files using multiple threads.
+* Reduced commit latency for large external file ingestions into the last level by adding `IngestExternalFileOptions::prefetch_lmax_index_and_filter_blocks`, which can skip commit-time index and filter block prefetching for cache-backed table metadata.
+
+## 11.4.0 (06/02/2026)
+### Public API Changes
+* Added `rocksdb_options_set_memtable_batch_lookup_optimization()` and `rocksdb_options_get_memtable_batch_lookup_optimization()` to the C API, exposing the existing `AdvancedColumnFamilyOptions::memtable_batch_lookup_optimization` field. This allows C API users (and downstream language bindings) to enable the skip-list memtable's batch-lookup optimization for `MultiGet`, which caches the search path between consecutive keys and reduces per-key cost from O(log N) to O(log d) where d is the distance between consecutive keys.
+* Added `rocksdb_readoptions_set_optimize_multiget_for_io()` and `rocksdb_readoptions_get_optimize_multiget_for_io()` to the C API, exposing the existing `ReadOptions::optimize_multiget_for_io` field. This allows C API users (and downstream language bindings) to opt out of the multi-level parallel MultiGet path, which only takes effect when the library is built with `USE_COROUTINES`.
+* Added `rocksdb_block_based_table_index_block_search_type_auto` enum constant and the `rocksdb_block_based_options_set_uniform_cv_threshold()` setter to the C API. The constant exposes `BlockSearchType::kAuto`, and the setter exposes `BlockBasedTableOptions::uniform_cv_threshold`. Both are required for `kAuto` index-block search to take effect from the C API: without setting `uniform_cv_threshold >= 0`, the per-block `is_uniform` footer bit is never written, and `kAuto` falls back to binary search at read time.
+* Added `EventListener::OnDBShutdownBegin`, a callback that fires once when a DB begins shutdown, including when RocksDB cleans up a failed `DB::Open()` attempt.
+* Added a new PerfContext counter `blob_cache_read_byte` for blob cache bytes read.
+
+### Behavior Changes
+* Remote compaction now falls back to local compaction when the primary cannot deserialize a successful `CompactionServiceResult` before installing any remote output files.
+* StringToMap (rocksdb/convenience.h) now preserves the outer braces of nested values so each map entry is in self-contained form and can be embedded directly into another `key=value;` string (e.g. SetOptions) without losing inner ';' delimiters. A new symmetric MapToString utility is provided.
+
+### Bug Fixes
+* Fixed a bug where `AbortAllCompactions()` could leave automatic compaction work unscheduled when aborting before the background worker picked it, causing compaction and write stalls until DB restart.
+* Fix a bug where db had a false positive compaction corruption error due to remote compaction service result with `has_accurate_num_input_records=false` not being serialized.
+* Fixed WritePrepared TransactionDB cleanup after retryable commit or rollback write failures so prepared transactions remain rollbackable instead of leaving unresolved prepared state.
+* Fix a rare corruption bug for tiered compaction that incorrectly moved last level range tombstones into proximal level. This corruption error is surfaced when force_consistency_checks is enabled.
+
+### Performance Improvements
+* Reduce the likelihood of user-facing metadata operations blocking on MANIFEST rotation when file creation is slow, such as on remote storage. MANIFEST write batches containing foreground edits from `CreateColumnFamily()`, `DropColumnFamily()`, `CreateColumnFamilyWithImport()`, `IngestExternalFile()`, and `DeleteFilesInRanges()` now get a relaxed file size threshold for triggering MANIFEST rotation; background-only MANIFEST write batches, such as compaction and flush, continue to use the normal threshold. This change should make auto-tuning manifest file size more attractive (see `max_manifest_file_size` and `max_manifest_space_amp_pct` options).
+
+
+## 11.3.0 (05/15/2026)
+### New Features
+* Add experimental DB option `async_wal_precreate` to precreate the next WAL file in a background thread and reduce foreground WAL rotation latency. The option is sanitized to false when WAL recycling is enabled.
+* Added a new `EventListener::OnCompactionPreCommit` callback that fires after a compaction job finishes but before its input files are released (i.e. while `FileMetaData::being_compacted` is still true). Listeners that maintain bookkeeping of which files are currently being compacted can clean up such state in this new callback to avoid races with concurrent compaction picking, where another thread might pick up the same files for a new compaction immediately after `being_compacted` is flipped back to false but before `OnCompactionCompleted` fires. The default implementation is a no-op so this is not a breaking change.
+* Add mutable DBOption `optimize_manifest_for_recovery` (default false). When enabled, RocksDB can reduce recovery work after a clean shutdown, which may lower DB::Open latency on warm reopens.
+* Added public utility APIs `ParseCompressionNameForDisplay()` to convert `TableProperties::compression_name` into a human-readable compression name for both legacy and format_version 7+ SST metadata, including custom `CompressionManager`-provided display names for custom compression types.
+* Add `reuse_manifest_on_open` DBOption (default false). When enabled, DB::Open reuses the existing MANIFEST file for append instead of creating a fresh one, avoiding the cost of serializing the entire database state into a new MANIFEST on the first post-open write. To prevent this feature from interfering with manifest file size auto-tuning, an extra forward-compatible field is now always added to the MANIFEST (to track the last "compacted" size).
+
+### Behavior Changes
+* Read-only open with `error_if_wal_file_exists=true` now tolerates empty WAL files so empty precreated WALs do not prevent inspection.
+* WriteCommitted TransactionDB now matches WritePrepared and WriteUnprepared compaction filtering in both single- and two-write-queue modes: a compaction filter's FilterMergeOperand will not be invoked on merge operands at or below the latest published sequence number.
+
+### Bug Fixes
+* Fixed blob-backed wide-column merge reads to preserve correct status
+propagation and resolution across memtable, read-only, and secondary DB
+paths.
+* Fixed a bug where `DB::GetCreationTimeOfOldestFile()` could return inaccurate results instead of the real creation time when called shortly after opening a legacy DB (one whose manifest lacks `file_creation_time`) with `open_files_async = true`. The API now waits for background SST file loading to complete only when needed; modern DBs are unaffected.
+* Fixed merge reads against wide-column/blob-backed base values to preserve precise failure statuses, including `GetMergeOperands()` and direct-write memtable reads.
+* Fix bug in range tombstone synthesis that covers live keys added during an IngestExternalFile
+* Reject the empty string as a column family name in `DB::CreateColumnFamily` / `DB::CreateColumnFamilies`. Previously such calls returned OK and a usable handle, but the column family was not persisted in the manifest, so any data written to it was silently lost on DB reopen.
+* Fixed a bug where a WriteCommitted TransactionDB using commit-bypass WBWI ingestion could drop an entry that is still visible at the published sequence boundary.
+
+## 11.2.0 (04/18/2026)
+### New Features
+* Added experimental `DBOptions::fast_sst_open` option. When enabled, RocksDB retrieves opaque file system metadata for SST files after flush, compaction, and external file ingestion, persists it in the MANIFEST, and passes it back to the file system on subsequent file opens to accelerate DB open time.
+* Added new option `min_tombstones_for_range_conversion` in `AdvancedColumnFamilyOptions`. When set to a non-zero value N, forward or reverse iteration will convert N or more contiguous point tombstones into a range tombstone in the mutable memtable. Future read operations will then be able to benefit from range tombstone optimizations. There are some limitations when it comes to table_filters, prefix_filters, and UDTs. See header comments for more details.
+* Added read-triggered compaction: a new column family option `read_triggered_compaction_threshold` (default 0, disabled) that marks SST files for compaction when their read frequency (`num_collapsible_entry_reads_sampled / file_size`) exceeds the threshold. This helps reduce read amplification for frequently-read ("hot") keys. A new DB option `max_compaction_trigger_wakeup_seconds` (default 43200s / 12 hours) controls the maximum interval for periodic compaction score re-evaluation, which is necessary for this feature to work on quiet (no-write) databases.
+
+### Public API Changes
+* Added new `WideColumnBlobResolver` interface and `CompactionFilter::FilterV4()` method, including `ResolveColumn()` / `ResolveColumns()` helpers for lazy loading blob column values in wide-column entities during compaction. This allows compaction filters to resolve blob values on-demand, avoiding unnecessary I/O for blob columns they don't need to access.
+* Changed experimental feature `ExternalTableReader::Get` and `ExternalTableReader::MultiGet` to use `PinnableSlice` instead of `std::string` for output values, enabling zero-copy pinning. This will break existing implementations.
+* Added `SstFileReader::Get` and `SstFileReader::MultiGet` overloads that accept `PinnableSlice`/`std::vector<PinnableSlice>*`, enabling zero-copy reads when the underlying `TableReader` supports pinning.
+
+### Behavior Changes
+* Prefix filter changes - when seeking to a key that is out of domain, and total_order_seek is false, total_order_seek is treated as if it were true. When prefix_same_as_start = true, now iterating past a key that is out of domain invalidates the iterator. The existing behavior does not check for InDomain in the DBIter, so Transform() can produce undefined behavior (e.g. key of size 3 on FixedPrefixTransform(4)).
+* Wide-column entities with blob-backed columns now use a new V2 on-disk encoding; older RocksDB versions that do not support wide-column blob separation will reject DBs or SSTs containing those entities.
+
+### Bug Fixes
+* Fix blob garbage accounting for blob direct-write flushes so flush-time filtering and overwrite elision correctly register obsolete blob bytes in blob metadata.
+* Fix a memory accounting leak in IODispatcher where ReadIndex() moved block values out of ReadSet without releasing the associated prefetch memory, causing subsequent prefetches to be blocked when max_prefetch_memory_bytes was set.
+* Fix MultiScan to fall back to synchronous coalesced reads when async I/O is unsupported at runtime.
+
+
+## 11.1.0 (03/23/2026)
+### New Features
+* Add a new option `open_files_async`. The existing behavior is on DB open, we open all sst files and do basic validations. For very large DBs on remote filesystems with many ssts, this may take very long. This option performs these validations instead in the background. Open errors found by this async background task are surfaced as a new background error kAsyncFileOpen.
+* Added `BlockBasedTableOptions::kAuto` index block search type that automatically selects between binary and interpolation search on a per-index-block basis. During SST construction, each index block's key distribution uniformity is analyzed using the coefficient of variation of key gaps, and index blocks with uniform keys use interpolation search while others fall back to binary search. The uniformity threshold is configurable via `BlockBasedTableOptions::uniform_cv_threshold` (default: 0.2).
+* Introduced enforce_write_buffer_manager_during_recovery option to allow WriteBufferManager to be enforced during WAL recovery. (Default: true)
+* Add `memtable_batch_lookup_optimization` option to use batch lookup optimization for memtable MultiGet. For skip list memtables, after each key lookup, the search path is cached and reused for the next key, reducing per-key cost from O(log N) to O(log d) where d is the distance between consecutive keys. Benchmarks show ~7% improvement in memtable-resident MultiGet throughput.
+* Added `BlockBasedTableOptions::PrepopulateBlockCache::kFlushAndCompaction` to prepopulate the block cache during both flush and compaction. Compaction-warmed blocks are inserted at `BOTTOM` priority (vs `LOW` for flush) so they are evicted first under cache pressure. Recommended only for use cases where most or all of the database is expected to reside in cache (e.g., tiered or remote storage where the working set fits in cache).
+* Added new mutable DB option `verify_manifest_content_on_close` (default: false). When enabled, on DB close the MANIFEST file is read back and all records are validated (CRC checksums and logical content). If corruption is detected, a fresh MANIFEST is written from in-memory state.
+
+### Behavior Changes
+* num_reads_sampled now factors in re-seeks and next/prev() on file iterators for files in L1+. next/prev() is discounted by 64x compared to seek due to being a much cheaper call.
+* Remote compaction workers now skip WAL recovery when opening the secondary DB instance, since only the LSM state from MANIFEST is needed for compaction. This reduces I/O and speeds up remote compaction startup.
+
+### Bug Fixes
+* Fix a bug in round-robin compaction that missed selecting input files that are needed to guarantee data correctness and cause crashing in debug builds or silent data corruption in release builds for Get().
+
 ## 11.0.0 (02/23/2026)
 ### New Features
 * Added support for storing wide-column entity column values in blob files. When `min_blob_size` is configured, large column values in wide-column entities will be stored in blob files, reducing SST file size and improving read performance.

Makefile

@@ -288,7 +288,7 @@ endif
 endif
 
 export JAVAC_ARGS
-CLEAN_FILES += make_config.mk rocksdb.pc
+CLEAN_FILES += rocksdb.pc
 
 ifeq ($(V), 1)
 $(info $(shell uname -a))
@@ -327,6 +327,9 @@ missing_make_config_paths := $(shell				\
 $(foreach path, $(missing_make_config_paths), \
 	$(warning Warning: $(path) does not exist))
 
+# This (the first rule) must depend on "all".
+default: all
+
 ifeq ($(PLATFORM), OS_AIX)
 # no debug info
 else ifneq ($(PLATFORM), IOS)
@@ -384,6 +387,10 @@ endif
 # TSAN doesn't work well with jemalloc. If we're compiling with TSAN, we should use regular malloc.
 ifdef COMPILE_WITH_TSAN
 	DISABLE_JEMALLOC=1
+	# Use a suppressions file instead of the process-wide TSAN default
+	# suppressions hook, which belongs to the final application.
+	TSAN_OPTIONS?=suppressions=$(CURDIR)/tools/tsan_suppressions.txt
+	export TSAN_OPTIONS
 	EXEC_LDFLAGS += -fsanitize=thread
 	PLATFORM_CCFLAGS += -fsanitize=thread -fPIC -DFOLLY_SANITIZE_THREAD
 	PLATFORM_CXXFLAGS += -fsanitize=thread -fPIC -DFOLLY_SANITIZE_THREAD
@@ -481,9 +488,6 @@ ifdef ROCKSDB_MODIFY_NPHASH
   PLATFORM_CXXFLAGS += -DROCKSDB_MODIFY_NPHASH=1
 endif
 
-# This (the first rule) must depend on "all".
-default: all
-
 WARNING_FLAGS = -W -Wextra -Wall -Wsign-compare -Wshadow \
   -Wunused-parameter
 
@@ -638,6 +642,7 @@ PARALLEL_TEST = $(filter-out $(NON_PARALLEL_TEST), $(ROCKSDBTESTS_SUBSET))
 TESTS_PLATFORM_DEPENDENT := \
 	db_basic_test \
 	db_blob_basic_test \
+	db_blob_direct_write_test \
 	db_encryption_test \
 	external_sst_file_basic_test \
 	auto_roll_logger_test \
@@ -645,6 +650,7 @@ TESTS_PLATFORM_DEPENDENT := \
 	dynamic_bloom_test \
 	c_test \
 	checkpoint_test \
+	sorted_run_builder_test \
 	crc32c_test \
 	coding_test \
 	inlineskiplist_test \
@@ -758,7 +764,6 @@ util/build_version.cc: $(filter-out $(OBJ_DIR)/util/build_version.o, $(LIB_OBJEC
 	$(AM_V_at)$(gen_build_version) > $@
 endif
 CLEAN_FILES += util/build_version.cc
-
 default: all
 
 #-----------------------------------------------
@@ -803,9 +808,21 @@ endif  # PLATFORM_SHARED_EXT
 .PHONY: check clean coverage ldb_tests package dbg gen-pc build_size \
 	release tags tags0 valgrind_check format static_lib shared_lib all \
 	rocksdbjavastatic rocksdbjava install install-static install-shared \
-	uninstall analyze tools tools_lib check-headers checkout_folly clang-tidy
+	uninstall analyze tools tools_lib check-headers checkout_folly clang-tidy \
+	check-c-api-c_test
 
-all: $(LIBRARY) $(BENCHMARKS) tools tools_lib test_libs $(TESTS)
+# Auto-configure git hooks on first build so developers do not need to run
+# "make install-hooks" manually. This is a no-op if already set.
+setup-hooks:
+	@if [ -d .git ] && [ -d githooks ]; then \
+		cur=$$(git config core.hooksPath 2>/dev/null); \
+		if [ "$$cur" != "githooks" ]; then \
+			git config core.hooksPath githooks; \
+			echo "git hooks: configured core.hooksPath = githooks"; \
+		fi; \
+	fi
+
+all: setup-hooks $(LIBRARY) $(BENCHMARKS) tools tools_lib test_libs $(TESTS)
 
 all_but_some_tests: $(LIBRARY) $(BENCHMARKS) tools tools_lib test_libs $(ROCKSDBTESTS_SUBSET)
 
@@ -870,23 +887,43 @@ coverage: clean
 parallel_tests = $(patsubst %,parallel_%,$(PARALLEL_TEST))
 .PHONY: gen_parallel_tests $(parallel_tests)
 # Shard size controls how many test cases run per process. The actual number
-# of shards per binary is: min(ceil(test_count / GTEST_SHARD_SIZE), NCORES * 8)
+# of shards per binary is: min(ceil(test_count / shard_size), NCORES * 8)
 # This adapts to machine size: many small shards on beefy machines, fewer
 # larger shards on CI (typically 2-4 cores).
 GTEST_SHARD_SIZE ?= 10
 NCORES ?= $(shell nproc 2>/dev/null || sysctl -n hw.ncpu 2>/dev/null || echo 4)
 
+# Per-binary overrides for GTEST_SHARD_SIZE, used to chop up binaries whose
+# individual test cases are slow enough that the default of 10 leaves one
+# shard on the critical path of "make check". Format: whitespace-separated
+# tokens "BINARY:SIZE". Sizes were tuned from `make suggest-slow-tests`
+# output to keep max shard time under ~20s on a beefy box. Reducing too far
+# adds per-shard fixed overhead, so don't use 1 unless individual tests are
+# >5s.
+SHARD_SIZE_OVERRIDES ?= \
+  point_lock_manager_stress_test:1 \
+  external_sst_file_test:1 \
+  external_sst_file_basic_test:2 \
+  db_test:3 \
+  compaction_service_test:1
+
 $(parallel_tests):
 	$(AM_V_at)TEST_BINARY=$(patsubst parallel_%,%,$@); \
   TEST_COUNT=` \
     (./$$TEST_BINARY --gtest_list_tests 2>/dev/null || echo "  list_failure") \
     | grep -c '^ '`; \
 	if [ "$$TEST_COUNT" -le 0 ]; then TEST_COUNT=1; fi; \
+	SHARD_SIZE=$(GTEST_SHARD_SIZE); \
+	for o in $(SHARD_SIZE_OVERRIDES); do \
+		case "$$o" in \
+			"$$TEST_BINARY":*) SHARD_SIZE=$${o#*:} ;; \
+		esac; \
+	done; \
 	MAX_SHARDS=$$(( $(NCORES) * 8 )); \
-	NUM_SHARDS=$$(( (TEST_COUNT + $(GTEST_SHARD_SIZE) - 1) / $(GTEST_SHARD_SIZE) )); \
+	NUM_SHARDS=$$(( (TEST_COUNT + SHARD_SIZE - 1) / SHARD_SIZE )); \
 	if [ "$$NUM_SHARDS" -gt "$$MAX_SHARDS" ]; then NUM_SHARDS=$$MAX_SHARDS; fi; \
 	if [ "$$NUM_SHARDS" -le 0 ]; then NUM_SHARDS=1; fi; \
-	echo "  Generating $$NUM_SHARDS shards for $$TEST_BINARY ($$TEST_COUNT tests)"; \
+	echo "  Generating $$NUM_SHARDS shards for $$TEST_BINARY ($$TEST_COUNT tests, shard_size=$$SHARD_SIZE)"; \
 	SHARD_IDX=0; \
 	while [ "$$SHARD_IDX" -lt "$$NUM_SHARDS" ]; do \
 		if [ -n "$(CI_TOTAL_SHARDS)" ] && [ $$(( $$SHARD_IDX % $(CI_TOTAL_SHARDS) )) -ne $(CI_SHARD_INDEX) ]; then \
@@ -912,31 +949,64 @@ gen_parallel_tests:
 	$(AM_V_at)$(FIND) t -type f -name 'run-*' -exec rm -f {} \;
 	$(MAKE) $(parallel_tests)
 
-# Reorder input lines (which are one per test) so that the
-# longest-running tests appear first in the output.
-# Do this by prefixing each selected name with its duration,
-# sort the resulting names, and remove the leading numbers.
-# FIXME: the "100" we prepend is a fake time, for now.
-# FIXME: squirrel away timings from each run and use them
-# (when present) on subsequent runs to order these tests.
+# Reorder input lines (one per test or per shard) so the longest-running
+# tests/shards start first under "make check" parallel scheduling. The trick
+# is to prefix matched names with a fake duration (100), sort numerically
+# descending, then strip the prefix. Slow-matched tests therefore come ahead
+# of all unmatched tests; ordering within each group is unspecified.
 #
-# Without this reordering, these two tests would happen to start only
-# after almost all other tests had completed, thus adding 100 seconds
-# to the duration of parallel "make check".  That's the difference
-# between 4 minutes (old) and 2m20s (new).
+# Why this matters: gnu_parallel hands out jobs in input order. If a slow
+# test starts only after most others have finished, the run is bottlenecked
+# on its tail. Front-loading slow tests overlaps them with the bulk of
+# faster ones for much better wall-clock time.
 #
-# 152.120 PASS t/DBTest.FileCreationRandomFailure
-# 107.816 PASS t/DBTest.EncodeDecompressedBlockSizeTest
+# Maintenance:
+#   1) Run `make check` so LOG is up-to-date.
+#   2) Run `make suggest-slow-tests` to see candidate slow binaries.
+#   3) Add binaries with high per-shard time (slow critical path) or high
+#      total time across many shards (need early queueing for fan-out).
+# Each alternative is wrapped in `^.*NAME.*$$` so the *whole input line* is
+# captured into $$1; then `s,(...),100 $$1,` prepends "100 " to the line.
 #
-# With sharded test execution, prioritize binaries known to be slow.
-# These generate many shards and should start early for good load balancing.
+# Tiers below are based on observed timings (see suggest-slow-tests).
+# Tier 1: max single-shard time >= 30s (critical-path bottlenecks).
+# Tier 2: max single-shard time 15-30s.
+# Tier 3: huge total time across many tiny shards; front-loading them keeps
+#         the tail of the run busy while big shards finish.
 slow_test_regexp = \
-	^.*block_based_table_reader_test.*$$|^.*table_test.*$$|^.*block_test.*$$|^.*write_prepared_transaction_test.*$$|^.*transaction_test.*$$|^.*external_sst_file_test.*$$|^.*db_wal_test.*$$|^.*db_with_timestamp_basic_test.*$$|^.*db_test-.*$$
+	^.*point_lock_manager_stress_test.*$$|^.*db_test.*$$|^.*external_sst_file_test.*$$|^.*compaction_service_test.*$$|^.*corruption_test.*$$|^.*comparator_db_test.*$$|^.*external_sst_file_basic_test.*$$|^.*rate_limiter_test.*$$|^.*db_compaction_test.*$$|^.*write_prepared_transaction_test.*$$|^.*db_merge_operator_test.*$$|\
+	^.*db_dynamic_level_test.*$$|^.*db_bloom_filter_test.*$$|^.*error_handler_fs_test.*$$|^.*merge_helper_test.*$$|^.*transaction_test.*$$|^.*db_kv_checksum_test.*$$|^.*inlineskiplist_test.*$$|\
+	^.*db_with_timestamp_basic_test.*$$|^.*table_test.*$$|^.*db_wal_test.*$$|^.*block_based_table_reader_test.*$$|^.*block_test.*$$
 prioritize_long_running_tests =						\
   perl -pe 's,($(slow_test_regexp)),100 $$1,'				\
     | sort -k1,1gr							\
     | sed 's/^[.0-9]* //'
 
+# Helper: print binaries observed to be slow in the last `make check` run.
+# Use this to decide whether `slow_test_regexp` above needs updating.
+# Aggregates per-binary across shards; flags any binary whose max single-shard
+# time is >= 20s OR whose total time is >= 200s.
+.PHONY: suggest-slow-tests
+suggest-slow-tests:
+	@if [ ! -s LOG ]; then \
+	  echo "No (or empty) LOG file. Run 'make check' first." >&2; exit 1; \
+	fi
+	@bash -c '$(quoted_perl_command)' < LOG | awk '\
+	  { \
+	    t=$$1; name=$$3; \
+	    sub(/^t\/run-/, "", name); \
+	    sub(/-shard-[0-9]+$$/, "", name); \
+	    total[name] += t; \
+	    if (t > max[name]) max[name] = t; \
+	    count[name]++; \
+	  } \
+	  END { \
+	    printf "%8s %8s %5s  %s\n", "TOTAL_s", "MAX_s", "SHRDS", "BINARY"; \
+	    for (n in total) \
+	      if (max[n] >= 20 || total[n] >= 200) \
+	        printf "%8.1f %8.1f %5d  %s\n", total[n], max[n], count[n], n; \
+	  }' | (read header; echo "$$header"; sort -k2,2gr)
+
 # "make check" uses
 # Run with "make J=1 check" to disable parallelism in "make check".
 # Run with "make J=200% check" to run two parallel jobs per core.
@@ -973,7 +1043,7 @@ check_0:
 		        awk -v s=$(CI_SHARD_INDEX) -v n=$(CI_TOTAL_SHARDS) '(NR-1)%n==s'; \
 		      else cat; fi; \
 		fi; \
-		find t -name 'run-*' -print; \
+		$(FIND) t -name 'run-*' -print; \
 	} \
 	  | $(prioritize_long_running_tests)				\
 	  | grep -E '$(tests-regexp)'					\
@@ -995,7 +1065,7 @@ valgrind_check_0:
 	  '  run "make watch-log" in a separate window' '';		\
 	{								\
 	  printf './%s\n' $(filter-out $(PARALLEL_TEST) %skiplist_test options_settable_test, $(TESTS));		\
-	  find t -name 'run-*' -print; \
+	  $(FIND) t -name 'run-*' -print; \
 	}								\
 	  | $(prioritize_long_running_tests)				\
 	  | grep -E '$(tests-regexp)'					\
@@ -1048,6 +1118,7 @@ ifneq ($(PLATFORM), OS_AIX)
 	$(PYTHON) tools/check_all_python.py
 ifndef ASSERT_STATUS_CHECKED # not yet working with these tests
 	$(PYTHON) tools/ldb_test.py
+	$(PYTHON) tools/db_crashtest_test.py
 	sh tools/rocksdb_dump_test.sh
 endif
 endif
@@ -1055,8 +1126,90 @@ ifndef SKIP_FORMAT_BUCK_CHECKS
 	$(MAKE) check-format
 	$(MAKE) check-buck-targets
 	$(MAKE) check-sources
+	$(MAKE) check-workflow-yaml
+	$(MAKE) check-c-api-gen
 endif
 
+# Check that the auto-generated C API files are up to date. It regenerates the
+# fragments and the inlined c.h/c.cc and compares them against a snapshot of the
+# checked-in copies (no net change when everything is up to date). It requires
+# clang++ (libclang, used to parse the C++ headers) and clang-format. When those
+# are unavailable the staleness/compat sub-checks are SKIPPED with a message so
+# `make check` still works without the codegen toolchain; the link-completeness
+# sub-check always runs (it needs no toolchain).
+#
+# Pin the formatter to match CI by setting CLANG_FORMAT_BINARY, e.g.:
+#   make check-c-api-gen CLANG_FORMAT_BINARY=clang-format-21
+# This target is part of `make check` and is skipped by SKIP_FORMAT_BUCK_CHECKS.
+#
+# Set CHECK_C_API_GEN_STRICT=1 to turn every "skip" below into a hard error, so a
+# core CI job that runs this target cannot silently regress to a no-op if a
+# prerequisite (clang++, or the compat baseline ref) goes missing. The dedicated
+# build-linux-clang-21-no_test_run CI job runs this target with the flag set.
+# Any non-empty value other than 0/no/false enables strict mode.
+CLANG_FORMAT_BINARY ?=
+# Backward-compatibility baseline for the C API (signature-level) check. CI
+# overrides this with the PR's merge target; locally it falls back to main /
+# origin/main and is skipped if neither resolves.
+API_COMPAT_REF ?= main
+CHECK_C_API_GEN_STRICT ?=
+check-c-api-gen:
+	# Link-completeness is a property of the checked-in c.h/c.cc and needs no
+	# clang toolchain, so it always runs: every declared public C API function
+	# must have exactly one definition (guards against dropped wrappers that
+	# would break downstream language bindings at link time).
+	$(PYTHON) tools/c_api_gen/check_api_completeness.py
+	# Backward-compatibility: no public C function may be removed or have its
+	# signature changed vs the baseline. Skipped if the baseline ref is not
+	# resolvable locally (CI passes an explicit ref), unless CHECK_C_API_GEN_STRICT.
+	@strict=""; case "$(CHECK_C_API_GEN_STRICT)" in ""|0|no|NO|false|FALSE) ;; *) strict=1 ;; esac; \
+	ref=""; \
+	if git rev-parse --verify --quiet "$(API_COMPAT_REF)^{commit}" >/dev/null; then ref="$(API_COMPAT_REF)"; \
+	elif git rev-parse --verify --quiet "origin/$(API_COMPAT_REF)^{commit}" >/dev/null; then ref="origin/$(API_COMPAT_REF)"; fi; \
+	if [ -n "$$ref" ]; then \
+	  $(PYTHON) tools/c_api_gen/check_api_compatibility.py --ref "$$ref"; \
+	elif [ -n "$$strict" ]; then \
+	  echo "ERROR: C API compat baseline '$(API_COMPAT_REF)' not resolvable and CHECK_C_API_GEN_STRICT is set" >&2; exit 1; \
+	else \
+	  echo "Skipping C API backward-compatibility check ($(API_COMPAT_REF) not found; set API_COMPAT_REF)"; \
+	fi
+	# Staleness: regenerate and confirm the checked-in output is current. Needs a
+	# clang++ (the generator parses C++ ASTs); detect one the way the generator
+	# does (a clang in $(CXX) -- which may be ccache-prefixed/versioned -- else a
+	# bare/versioned clang++ on PATH) rather than testing $(CXX) verbatim.
+	@strict=""; case "$(CHECK_C_API_GEN_STRICT)" in ""|0|no|NO|false|FALSE) ;; *) strict=1 ;; esac; \
+	cf_arg=""; \
+	if [ -n "$(CLANG_FORMAT_BINARY)" ]; then cf_arg="--clang-format $(CLANG_FORMAT_BINARY)"; fi; \
+	have_clang=""; \
+	for c in $$(printf '%s\n' $(CXX) | grep -i clang) clang++ clang++-21 clang++-20 clang++-19 clang++-18 clang++-17 clang++-16 clang++-15 clang++-14 clang++-13; do \
+	  if command -v "$$c" >/dev/null 2>&1; then have_clang=1; break; fi; \
+	done; \
+	if [ -n "$$have_clang" ]; then \
+	  $(PYTHON) tools/c_api_gen/verify_generated_up_to_date.py $$cf_arg; \
+	elif [ -n "$$strict" ]; then \
+	  echo "ERROR: no clang++ found and CHECK_C_API_GEN_STRICT is set; cannot run the C API staleness check" >&2; exit 1; \
+	else \
+	  echo "Skipping C API codegen staleness check (no clang++ found; install clang++ or set CXX to a clang to enable)"; \
+	fi
+
+# Quick local validation for C API generation plus the focused C API test.
+# This verifies the checked-in generated fragments as well as the inlined
+# include/rocksdb/c.h and db/c.cc outputs, then runs c_test in an isolated
+# TEST_TMPDIR to avoid stale-state failures.
+check-c-api-c_test:
+	$(PYTHON) tools/c_api_gen/verify_generated_up_to_date.py
+	$(MAKE) c_test
+	@tmpdir=$$(mktemp -d); \
+	  trap 'rm -rf "$$tmpdir"' EXIT; \
+	  echo "===== Running c_test with TEST_TMPDIR=$$tmpdir"; \
+	  if command -v timeout >/dev/null 2>&1; then \
+	    TEST_TMPDIR="$$tmpdir" timeout 60 ./c_test; \
+	  elif command -v gtimeout >/dev/null 2>&1; then \
+	    TEST_TMPDIR="$$tmpdir" gtimeout 60 ./c_test; \
+	  else \
+	    TEST_TMPDIR="$$tmpdir" ./c_test; \
+	  fi
+
 # TODO add ldb_tests
 check_some: $(ROCKSDBTESTS_SUBSET)
 	for t in $(ROCKSDBTESTS_SUBSET); do echo "===== Running $$t (`date`)"; ./$$t || exit 1; done
@@ -1065,6 +1218,10 @@ check_some: $(ROCKSDBTESTS_SUBSET)
 ldb_tests: ldb
 	$(PYTHON) tools/ldb_test.py
 
+.PHONY: db_crashtest_tests
+db_crashtest_tests:
+	$(PYTHON) tools/db_crashtest_test.py
+
 include crash_test.mk
 
 asan_check: clean
@@ -1202,16 +1359,32 @@ rocksdb.h rocksdb.cc: build_tools/amalgamate.py Makefile $(LIB_SOURCES) unity.cc
 	build_tools/amalgamate.py -I. -i./include unity.cc -x include/rocksdb/c.h -H rocksdb.h -o rocksdb.cc
 
 clean: clean-ext-libraries-all clean-rocks clean-rocksjava
+# Removed here rather than in clean-rocks (via CLEAN_FILES) so the build-parameter
+# auto-clean, which runs clean-rocks, doesn't delete the make_config.mk already
+# included by the in-progress make.
+	@rm -f make_config.mk
 
 clean-not-downloaded: clean-ext-libraries-bin clean-rocks clean-not-downloaded-rocksjava
+	@rm -f make_config.mk
 
 clean-rocks:
 # Not practical to exactly match all versions/variants in naming (e.g. debug or not)
-	rm -f ${LIBNAME}*.so* ${LIBNAME}*.a
-	rm -f $(BENCHMARKS) $(TOOLS) $(TESTS) $(PARALLEL_TEST) $(MICROBENCHS)
-	rm -rf $(CLEAN_FILES) ios-x86 ios-arm scan_build_report
-	$(FIND) . -name "*.[oda]" -exec rm -f {} \;
-	$(FIND) . -type f \( -name "*.gcda" -o -name "*.gcno" \) -exec rm -f {} \;
+	@echo Removing link targets
+	@rm -f ${LIBNAME}*.so* ${LIBNAME}*.a $(BENCHMARKS) $(TOOLS) $(TESTS) $(PARALLEL_TEST) $(MICROBENCHS)
+	@echo Finding and cleaning other files
+	@rm -rf $(CLEAN_FILES) ios-x86 ios-arm scan_build_report
+	@if $(FIND) Makefile -regextype awk &> /dev/null; then \
+		$(FIND) . -regextype awk \
+			-regex '.*/([.].*|third-party)' -prune -o \
+			-type f -regex '.*[.]([oda]|gcda|gcno)' -exec rm -f {} \; ; \
+	else \
+		$(FIND) . -name "*.[oda]" -exec rm -f {} \; ; \
+	fi
+# Remove the build signature(s) LAST. Done after the object files are gone so
+# that a Ctrl+C partway through clean leaves the old signature in place: it
+# still matches the leftover objects, so a later build with different
+# parameters is still detected (signature usefulness is preserved).
+	@rm -f $(BUILD_SIG_FILE) jl/.build_signature jls/.build_signature
 
 clean-rocksjava: clean-rocks
 	rm -rf jl jls
@@ -1224,7 +1397,7 @@ clean-ext-libraries-all:
 	rm -rf bzip2* snappy* zlib* lz4* zstd*
 
 clean-ext-libraries-bin:
-	find . -maxdepth 1 -type d \( -name bzip2\* -or -name snappy\* -or -name zlib\* -or -name lz4\* -or -name zstd\* \) -prune -exec rm -rf {} \;
+	$(FIND) . -maxdepth 1 -type d \( -name bzip2\* -or -name snappy\* -or -name zlib\* -or -name lz4\* -or -name zstd\* \) -prune -exec rm -rf {} \;
 
 tags:
 	ctags -R .
@@ -1248,12 +1421,43 @@ format-auto:
 check-format:
 	build_tools/format-diff.sh -c
 
+
+# Crude alternative to setup-hooks: copies hooks into .git/hooks/ instead of
+# using core.hooksPath. The copies won't track changes to githooks/.
+install-hooks:
+	@echo "Installing git hooks from githooks/..."
+	@if [ -d githooks ]; then \
+		for hook in githooks/*; do \
+			hook_name=$$(basename "$$hook"); \
+			cp "$$hook" .git/hooks/"$$hook_name"; \
+			chmod +x .git/hooks/"$$hook_name"; \
+			echo "  Installed $$hook_name"; \
+		done; \
+		echo "Done. Hooks installed to .git/hooks/"; \
+	else \
+		echo "Error: githooks/ directory not found"; \
+		exit 1; \
+	fi
+
+# Reverse of install-hooks (not needed if using setup-hooks / core.hooksPath).
+uninstall-hooks:
+	@echo "Removing installed git hooks..."
+	@for hook in githooks/*; do \
+		hook_name=$$(basename "$$hook"); \
+		rm -f .git/hooks/"$$hook_name"; \
+		echo "  Removed $$hook_name"; \
+	done
+	@echo "Done."
+
 check-buck-targets:
 	buckifier/check_buck_targets.sh
 
 check-sources:
 	build_tools/check-sources.sh
 
+check-workflow-yaml:
+	build_tools/check-workflow-yaml.sh
+
 # Run clang-tidy on locally changed files, filtered to changed lines only.
 # Requires compile_commands.json (generate with cmake -DCMAKE_EXPORT_COMPILE_COMMANDS=ON).
 # Override CLANG_TIDY_BINARY and CLANG_TIDY_JOBS as needed:
@@ -1441,6 +1645,9 @@ db_basic_test: $(OBJ_DIR)/db/db_basic_test.o $(TEST_LIBRARY) $(LIBRARY)
 db_blob_basic_test: $(OBJ_DIR)/db/blob/db_blob_basic_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)
 
+db_blob_direct_write_test: $(OBJ_DIR)/db/blob/db_blob_direct_write_test.o $(TEST_LIBRARY) $(LIBRARY)
+	$(AM_LINK)
+
 db_blob_compaction_test: $(OBJ_DIR)/db/blob/db_blob_compaction_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)
 
@@ -1450,6 +1657,9 @@ db_readonly_with_timestamp_test: $(OBJ_DIR)/db/db_readonly_with_timestamp_test.o
 db_wide_basic_test: $(OBJ_DIR)/db/wide/db_wide_basic_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)
 
+db_wide_blob_direct_write_test: $(OBJ_DIR)/db/wide/db_wide_blob_direct_write_test.o $(TEST_LIBRARY) $(LIBRARY)
+	$(AM_LINK)
+
 db_with_timestamp_basic_test: $(OBJ_DIR)/db/db_with_timestamp_basic_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)
 
@@ -1459,10 +1669,13 @@ db_with_timestamp_compaction_test: db/db_with_timestamp_compaction_test.o $(TEST
 db_encryption_test: $(OBJ_DIR)/db/db_encryption_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)
 
+db_open_with_config_test: $(OBJ_DIR)/db/db_open_with_config_test.o $(TEST_LIBRARY) $(LIBRARY)
+	$(AM_LINK)
+
 db_test: $(OBJ_DIR)/db/db_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)
 
-db_test2: $(OBJ_DIR)/db/db_test2.o $(TEST_LIBRARY) $(LIBRARY)
+db_etc2_test: $(OBJ_DIR)/db/db_etc2_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)
 
 db_etc3_test: $(OBJ_DIR)/db/db_etc3_test.o $(TEST_LIBRARY) $(LIBRARY)
@@ -1603,6 +1816,9 @@ backup_engine_test: $(OBJ_DIR)/utilities/backup/backup_engine_test.o $(TEST_LIBR
 checkpoint_test: $(OBJ_DIR)/utilities/checkpoint/checkpoint_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)
 
+sorted_run_builder_test: $(OBJ_DIR)/utilities/sorted_run_builder/sorted_run_builder_test.o $(TEST_LIBRARY) $(LIBRARY)
+	$(AM_LINK)
+
 cache_simulator_test: $(OBJ_DIR)/utilities/simulator_cache/cache_simulator_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)
 
@@ -2202,6 +2418,101 @@ ifeq ($(PLATFORM), OS_OPENBSD)
 endif
 export SHA256_CMD
 
+# ----------------------------------------------------------------------------
+# Build parameter change detection
+# ----------------------------------------------------------------------------
+# RocksDB writes object files to the same $(OBJ_DIR) paths regardless of
+# DEBUG_LEVEL, sanitizers (ASAN/TSAN/UBSAN), ASSERT_STATUS_CHECKED, RTTI, LTO,
+# COERCE_CONTEXT_SWITCH, etc.  Most of those are applied in this Makefile after
+# `include make_config.mk` and are therefore NOT reflected in make_config.mk,
+# so switching them and rebuilding silently mixes incompatible object files.
+# To guard against that, we hash the fully-resolved compile/link parameters
+# (which already embed make_config.mk via PLATFORM_*) and compare against the
+# value recorded from the previous build.  On mismatch the build stops so the
+# user can `make clean`.  Knobs:
+#   AUTO_CLEAN=1
+#     run `make clean-rocks` automatically on mismatch
+#   ALLOW_BUILD_PARAMETER_CHANGE=1
+#     skip the check entirely (e.g. intentionally mixing DEBUG_LEVEL=1 and
+#     DEBUG_LEVEL=2)
+# The signature is stored per-$(OBJ_DIR), so Java (jl/jls) builds are tracked
+# independently from the default build.
+BUILD_SIG_FILE := $(OBJ_DIR)/.build_signature
+# NOTE: deliberately NOT added to CLEAN_FILES. The signature is removed as the
+# very last step of clean-rocks so that an interrupted clean keeps it (see
+# clean-rocks), preserving change detection across a Ctrl+C'd clean.
+
+# Goals that do not compile object files into the tree (pure utilities,
+# informational, source/config checks): the change check is irrelevant for them.
+BUILD_SIG_NONBUILD_GOALS := \
+	clean clean-rocks clean-not-downloaded clean-rocksjava \
+	clean-not-downloaded-rocksjava clean-ext-libraries-all \
+	clean-ext-libraries-bin \
+	format format-auto check-format check-buck-targets check-headers \
+	check-sources check-workflow-yaml check-progress clang-tidy \
+	tags tags0 package jclean checkout_folly build_folly \
+	watch-log dump-log suggest-slow-tests list_all_tests gen-pc \
+	gen_parallel_tests check-c-api-gen \
+	setup-hooks install-hooks uninstall-hooks uninstall db_crashtest_tests
+# Goals that clean before building (depend on or invoke `clean`): they manage
+# their own freshness, so the check must not block them (it would error before
+# their built-in clean runs).
+BUILD_SIG_NONBUILD_GOALS += \
+	release coverage build_size analyze analyze_incremental \
+	asan_check asan_crash_test asan_crash_test_with_atomic_flush \
+	asan_crash_test_with_txn asan_crash_test_with_best_efforts_recovery \
+	whitebox_asan_crash_test blackbox_asan_crash_test \
+	ubsan_check ubsan_crash_test ubsan_crash_test_with_atomic_flush \
+	ubsan_crash_test_with_txn ubsan_crash_test_with_best_efforts_recovery \
+	whitebox_ubsan_crash_test blackbox_ubsan_crash_test
+
+# Goals that explicitly clean; never trip the check when one is requested.
+BUILD_SIG_CLEAN_GOALS := \
+	clean clean-rocks clean-not-downloaded clean-rocksjava \
+	clean-not-downloaded-rocksjava clean-ext-libraries-all \
+	clean-ext-libraries-bin
+
+BUILD_SIG_GOALS := $(if $(MAKECMDGOALS),$(MAKECMDGOALS),all)
+BUILD_SIG_DO_BUILD := $(filter-out $(BUILD_SIG_NONBUILD_GOALS),$(BUILD_SIG_GOALS))
+BUILD_SIG_CLEANING := $(filter $(BUILD_SIG_CLEAN_GOALS),$(MAKECMDGOALS))
+# Dry runs (-n/--just-print) must not write the stamp, clean, or error; the
+# parse-time $(shell) below would otherwise run even under -n. Note that
+# actual options are canonicalized and shorted as possible such that all
+# short options are in the first word of MAKEFLAGS.
+BUILD_SIG_DRYRUN := $(findstring n,$(firstword -$(MAKEFLAGS)))
+
+# Only enforce at the top level. Sub-makes (e.g. `check` invokes
+# $(MAKE) gen_parallel_tests / check-c-api-gen / check_0) inherit the parent's
+# build flags, so the top-level check already covers them; re-checking inside a
+# sub-make only causes spurious failures.
+ifneq ($(ALLOW_BUILD_PARAMETER_CHANGE),1)
+ifeq ($(MAKELEVEL),0)
+ifeq ($(BUILD_SIG_DRYRUN),)
+ifeq ($(BUILD_SIG_CLEANING),)
+ifneq ($(BUILD_SIG_DO_BUILD),)
+  BUILD_SIG := $(shell printf '%s' '$(CC)|$(CXX)|$(CFLAGS)|$(CXXFLAGS)|$(LDFLAGS)|$(EXEC_LDFLAGS)' | $(SHA256_CMD) | cut -d ' ' -f 1)
+  BUILD_SIG_OLD := $(shell cat $(BUILD_SIG_FILE) 2>/dev/null)
+  ifneq ($(BUILD_SIG_OLD),)
+  ifneq ($(BUILD_SIG_OLD),$(BUILD_SIG))
+  ifeq ($(AUTO_CLEAN),1)
+    $(info *** Build parameters changed since last build (OBJ_DIR=$(OBJ_DIR)); running 'make clean-rocks' because AUTO_CLEAN=1)
+    BUILD_SIG_CLEAN_OUTPUT := $(shell $(MAKE) clean-rocks 1>&2)
+    ifneq ($(.SHELLSTATUS),0)
+      $(error AUTO_CLEAN: 'make clean-rocks' failed (exit $(.SHELLSTATUS)); not building against a partially-cleaned tree)
+    endif
+  else
+    $(error Build parameters changed since the last build (OBJ_DIR=$(OBJ_DIR)). Existing object files are stale and must be removed. Run 'make clean', or set AUTO_CLEAN=1 to clean automatically, or ALLOW_BUILD_PARAMETER_CHANGE=1 to build anyway)
+  endif
+  endif
+  endif
+  # Record the current signature for the next build.
+  BUILD_SIG_WRITE := $(shell mkdir -p $(OBJ_DIR) 2>/dev/null; printf '%s\n' '$(BUILD_SIG)' > $(BUILD_SIG_FILE))
+endif
+endif
+endif
+endif
+endif
+
 zlib-$(ZLIB_VER).tar.gz:
 	curl --fail --output zlib-$(ZLIB_VER).tar.gz --location ${ZLIB_DOWNLOAD_BASE}/zlib-$(ZLIB_VER).tar.gz
 	ZLIB_SHA256_ACTUAL=`$(SHA256_CMD) zlib-$(ZLIB_VER).tar.gz | cut -d ' ' -f 1`; \
@@ -2602,7 +2913,7 @@ list_all_tests:
 
 # Remove the rules for which dependencies should not be generated and see if any are left.
 #If so, include the dependencies; if not, do not include the dependency files
-ROCKS_DEP_RULES=$(filter-out clean format check-format check-buck-targets check-headers check-sources clang-tidy jclean jtest package analyze tags rocksdbjavastatic% unity.% unity_test checkout_folly, $(MAKECMDGOALS))
+ROCKS_DEP_RULES=$(filter-out clean format check-format check-buck-targets check-headers check-sources check-workflow-yaml clang-tidy jclean jtest package analyze tags rocksdbjavastatic% unity.% unity_test checkout_folly, $(MAKECMDGOALS))
 ifneq ("$(ROCKS_DEP_RULES)", "")
 -include $(DEPFILES)
 endif

buckifier/buckify_rocksdb.py

@@ -146,10 +146,10 @@ def generate_buck(repo_path, deps_map):
         src_mk["RANGE_TREE_SOURCES"] + src_mk["TOOL_LIB_SOURCES"],
         deps=[
             "//folly/container:f14_hash",
-            "//folly/experimental/coro:blocking_wait",
-            "//folly/experimental/coro:collect",
-            "//folly/experimental/coro:coroutine",
-            "//folly/experimental/coro:task",
+            "//folly/coro:blocking_wait",
+            "//folly/coro:collect",
+            "//folly/coro:coroutine",
+            "//folly/coro:task",
             "//folly/synchronization:distributed_mutex",
         ],
         headers=LiteralValue("glob([\"**/*.h\"])")
@@ -360,6 +360,7 @@ def generate_buck(repo_path, deps_map):
                         extra_compiler_flags=json.dumps(deps["extra_compiler_flags"]),
                     )
     BUCK.export_file("tools/db_crashtest.py")
+    BUCK.export_file("tools/fault_injection_log_parser.py")
 
     print(ColorString.info("Generated BUCK Summary:"))
     print(ColorString.info("- %d libs" % BUCK.total_lib))

buckifier/targets_builder.py

@@ -118,6 +118,12 @@ class TARGETSBuilder:
         self.total_bin = self.total_bin + 1
 
     def add_c_test(self):
+        # The actual c_test_bin target is defined by add_c_test_wrapper in the
+        # internal //rocks/buckifier:defs.bzl (not in this OSS repo). db/c_test.c
+        # #includes the generated c_api_gen/*.inc fragments, so under Buck's
+        # hermetic sandbox that wrapper must expose them as headers, e.g.
+        #   headers = native.glob(["c_api_gen/**/*.inc"])
+        # (Make/CMake resolve the include via -I. / PROJECT_SOURCE_DIR.)
         with open(self.path, "ab") as targets_file:
             targets_file.write(
                 b"""

build_tools/build_detect_platform

@@ -27,7 +27,6 @@
 #       -DLZ4                       if the LZ4 library is present
 #       -DZSTD                      if the ZSTD library is present
 #       -DNUMA                      if the NUMA library is present
-#       -DTBB                       if the TBB library is present
 #       -DMEMKIND                   if the memkind library is present
 #
 # Using gflags in rocksdb:
@@ -424,19 +423,6 @@ EOF
         fi
     fi
 
-    if ! test $ROCKSDB_DISABLE_TBB; then
-        # Test whether tbb is available
-        $CXX $PLATFORM_CXXFLAGS $LDFLAGS -x c++ - -o test.o -ltbb 2>/dev/null  <<EOF
-          #include <tbb/tbb.h>
-          int main() {}
-EOF
-        if [ "$?" = 0 ]; then
-            COMMON_FLAGS="$COMMON_FLAGS -DTBB"
-            PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -ltbb"
-            JAVA_LDFLAGS="$JAVA_LDFLAGS -ltbb"
-        fi
-    fi
-
     if ! test $ROCKSDB_DISABLE_JEMALLOC; then
         # Test whether jemalloc is available
         if echo 'int main() {}' | $CXX $PLATFORM_CXXFLAGS $LDFLAGS -x c++ - -o test.o -ljemalloc \

build_tools/check-public-header.sh

@@ -1,5 +1,8 @@
 #!/usr/bin/env bash
-# Copyright (c) Meta Platforms, Inc. and affiliates. All Rights Reserved.
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# This source code is licensed under both the GPLv2 (found in the
+# COPYING file in the root directory) and Apache 2.0 License
+# (found in the LICENSE.Apache file in the root directory).
 #
 # Check for some simple mistakes in public headers (on the command line)
 # that should prevent commit or push

build_tools/check-sources.sh

@@ -37,7 +37,7 @@ if [ "$?" != "1" ]; then
   BAD=1
 fi
 
-git grep -n -P "[\x80-\xFF]" -- ':!docs' ':!*.md'
+LC_ALL=C git grep -n $'[\x80-\xff]' -- ':!docs' ':!*.md' ':!.github'
 if [ "$?" != "1" ]; then
   echo '^^^^ Use only ASCII characters in source files'
   BAD=1

build_tools/check-workflow-yaml.sh

@@ -0,0 +1,42 @@
+#!/usr/bin/env bash
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
+#
+# Validate GitHub Actions workflow YAML before it reaches CI runtime.
+
+set -euo pipefail
+
+if ! command -v ruby >/dev/null 2>&1; then
+  echo "ruby is required to validate GitHub Actions workflow YAML"
+  echo "On CentOS Stream: sudo dnf install ruby rubygems rubygem-psych"
+  exit 1
+fi
+
+if ! ruby -e 'require "psych"' 2>/dev/null; then
+  echo "ruby is installed but cannot load required library 'psych'"
+  echo "On CentOS Stream: sudo dnf install rubygems rubygem-psych"
+  exit 1
+fi
+
+ruby <<'RUBY'
+require "psych"
+
+bad = false
+workflow_files = Dir[".github/workflows/*.{yml,yaml}"].sort
+
+if workflow_files.empty?
+  warn "No workflow YAML files found under .github/workflows"
+  exit 1
+end
+
+workflow_files.each do |path|
+  begin
+    Psych.parse_file(path)
+    puts "OK #{path}"
+  rescue Psych::Exception => e
+    warn "Invalid YAML in #{path}: #{e.message}"
+    bad = true
+  end
+end
+
+exit(bad ? 1 : 0)
+RUBY

build_tools/dependencies_platform010.sh

@@ -1,21 +1,20 @@
 # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
 # The file is generated using update_dependencies.sh.
-GCC_BASE=/mnt/gvfs/third-party2/gcc/62de5a92e5f23c661c3d4b9f322e04eb14e7a5bd/11.x/centos8-native/886b5eb
-CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/1f6edd1ff15c99c861afc8f3cd69054cd974dd64/15/platform010/72a2ff8
-LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/d1129753c8361ac8e9453c0f4291337a4507ebe6/11.x/platform010/5684a5a
-GLIBC_BASE=/mnt/gvfs/third-party2/glibc/fed6e93d87571fb162734c86636119d45a398963/2.34/platform010/f259413
-SNAPPY_BASE=/mnt/gvfs/third-party2/snappy/31a346126a1f3b64812c362511cb04cc1bd40855/1.1.8/platform010/76ebdda
-ZLIB_BASE=/mnt/gvfs/third-party2/zlib/0c65c05468b5a38cef1a106a1f526463e120c8dd/1.2.8/platform010/76ebdda
+GCC_BASE=/mnt/gvfs/third-party2/gcc/e9cb2d4550b5a95024125084420342d256b349cd/11.x/centos9-native/3bed279
+CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/4e46e6967e15412702e51960c69790a78933856f/21/platform010/72a2ff8
+LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/63c5179edc90fec473e53c1b2f92269c65c39e07/11.x/platform010/5684a5a
+GLIBC_BASE=/mnt/gvfs/third-party2/glibc/b4a6248a76a163bcccb1e4751e247712c535846e/2.34/platform010/f259413
+SNAPPY_BASE=/mnt/gvfs/third-party2/snappy/ef895164499fdfedb9d9b9c85520de10aa8ff9ca/1.1.8/platform010/76ebdda
+ZLIB_BASE=/mnt/gvfs/third-party2/zlib/643dbf0136f869562485cfcb2953f5cab60447ef/1.2.8/platform010/76ebdda
 BZIP2_BASE=/mnt/gvfs/third-party2/bzip2/09703139cfc376bd8a82642385a0e97726b28287/1.0.6/platform010/76ebdda
-LZ4_BASE=/mnt/gvfs/third-party2/lz4/ff23d17b932725cc1734a14896a8b67c518ba169/1.9.4/platform010/76ebdda
-ZSTD_BASE=/mnt/gvfs/third-party2/zstd/576397d8b1d9cea7306ad1e454d5e55caaa2ff1c/1.4.x/platform010/64091f4
-GFLAGS_BASE=/mnt/gvfs/third-party2/gflags/fecac07861cb829f5e60dbeff0503d3272db73c0/2.2.0/platform010/76ebdda
-JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/0bb3f5756788ce26e2e16a1cb2f2af2c59b51abe/master/platform010/f57cc4a
+LZ4_BASE=/mnt/gvfs/third-party2/lz4/823bf8e000654abef784311296ffc8166d9d3986/1.9.4/platform010/76ebdda
+ZSTD_BASE=/mnt/gvfs/third-party2/zstd/0cda78647b5712bec7ec6d9ba2f0dcdabfb6f44f/1.4.x/platform010/64091f4
+GFLAGS_BASE=/mnt/gvfs/third-party2/gflags/a02e0de00945729ad553c187c64e86b57f45a33b/2.2.0/platform010/76ebdda
+JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/a1a50742bec7195bfafb594e174ea45dc197a420/master/platform010/779a252
 NUMA_BASE=/mnt/gvfs/third-party2/numa/5b602edd46fda54cdd7ea45f77dbe4061206e174/2.0.11/platform010/76ebdda
-LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/97cac22a149c2e202917e05d44e87e516b68216f/1.4/platform010/5074a48
-TBB_BASE=/mnt/gvfs/third-party2/tbb/53953ebc4e3eda85ad6fc3e429ba146035e97b90/2018_U5/platform010/76ebdda
+LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/f3cce2e1a79ac8d64e102886603a6997d6e88640/1.8/platform010/76ebdda
 LIBURING_BASE=/mnt/gvfs/third-party2/liburing/a98e2d137007e3ebf7f33bd6f99c2c56bdaf8488/20210212/platform010/76ebdda
 BENCHMARK_BASE=/mnt/gvfs/third-party2/benchmark/780c7a0f9cf0967961e69ad08e61cddd85d61821/trunk/platform010/76ebdda
-KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/624a2f8f6c93c3c1df8aa4a6255d8202631a6c80/fb/platform010/da39a3e
-BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/39579e8603b48b3540f8b0633f43adf29acccb8b/2.37/centos8-native/da39a3e
-VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/cd9cc656d49ecb53797ce4d055e49fde29fd57ff/3.19.0/platform010/76ebdda
+KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/0883cccda758e68b47e4c04e4fa01142e1f60f32/fb/platform010/da39a3e
+BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/1d5d19135a406a941d3a441eb2bf506bb9d19210/2.43/centos9-native/be7139e
+VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/3e2b7ccc315c9a0f2b678ce6fd774f4970f9bf3b/3.22.0/platform010/76ebdda

build_tools/fbcode_config.sh

@@ -87,15 +87,6 @@ if test -z $PIC_BUILD; then
   LIBUNWIND="$LIBUNWIND_BASE/lib/libunwind.a"
 fi
 
-# location of TBB
-TBB_INCLUDE=" -isystem $TBB_BASE/include/"
-if test -z $PIC_BUILD; then
-  TBB_LIBS="$TBB_BASE/lib/libtbb.a"
-else
-  TBB_LIBS="$TBB_BASE/lib/libtbb_pic.a"
-fi
-CFLAGS+=" -DTBB"
-
 test "$USE_SSE" || USE_SSE=1
 export USE_SSE
 test "$PORTABLE" || PORTABLE=1
@@ -104,7 +95,7 @@ export PORTABLE
 BINUTILS="$BINUTILS_BASE/bin"
 AR="$BINUTILS/ar"
 
-DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE $TBB_INCLUDE"
+DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE"
 
 STDLIBS="-L $GCC_BASE/lib64"
 
@@ -150,17 +141,17 @@ CFLAGS+=" $DEPS_INCLUDE"
 CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE -DROCKSDB_RANGESYNC_PRESENT -DROCKSDB_SCHED_GETCPU_PRESENT"
 CXXFLAGS+=" $CFLAGS"
 
-EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB $TBB_LIBS"
+EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB"
 EXEC_LDFLAGS+=" -B$BINUTILS/gold"
 EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/gcc-5-glibc-2.23/lib/ld.so"
 EXEC_LDFLAGS+=" $LIBUNWIND"
 EXEC_LDFLAGS+=" -Wl,-rpath=/usr/local/fbcode/gcc-5-glibc-2.23/lib"
-# required by libtbb
+# dynamic loading
 EXEC_LDFLAGS+=" -ldl"
 
 PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS $STDLIBS -lgcc -lstdc++"
 
-EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $TBB_LIBS"
+EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS"
 
 VALGRIND_VER="$VALGRIND_BASE/bin/"
 

build_tools/fbcode_config_platform010.sh

@@ -82,11 +82,6 @@ CFLAGS+=" -DNUMA"
 # location of libunwind
 LIBUNWIND="$LIBUNWIND_BASE/lib/libunwind${MAYBE_PIC}.a"
 
-# location of TBB
-TBB_INCLUDE=" -isystem $TBB_BASE/include/"
-TBB_LIBS="$TBB_BASE/lib/libtbb${MAYBE_PIC}.a"
-CFLAGS+=" -DTBB"
-
 # location of LIBURING
 LIBURING_INCLUDE=" -isystem $LIBURING_BASE/include/"
 LIBURING_LIBS="$LIBURING_BASE/lib/liburing${MAYBE_PIC}.a"
@@ -101,7 +96,7 @@ BINUTILS="$BINUTILS_BASE/bin"
 AR="$BINUTILS/ar"
 AS="$BINUTILS/as"
 
-DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE $TBB_INCLUDE $LIBURING_INCLUDE $BENCHMARK_INCLUDE"
+DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE $LIBURING_INCLUDE $BENCHMARK_INCLUDE"
 
 STDLIBS="-L $GCC_BASE/lib64"
 
@@ -157,18 +152,18 @@ CFLAGS+=" $DEPS_INCLUDE"
 CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE -DROCKSDB_RANGESYNC_PRESENT -DROCKSDB_SCHED_GETCPU_PRESENT -DROCKSDB_IOURING_PRESENT"
 CXXFLAGS+=" $CFLAGS"
 
-EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB $TBB_LIBS $LIBURING_LIBS $BENCHMARK_LIBS"
+EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB $LIBURING_LIBS $BENCHMARK_LIBS"
 EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/platform010/lib/ld.so"
 EXEC_LDFLAGS+=" $LIBUNWIND"
 EXEC_LDFLAGS+=" -Wl,-rpath=/usr/local/fbcode/platform010/lib"
 EXEC_LDFLAGS+=" -Wl,-rpath=$GCC_BASE/lib64"
-# required by libtbb
+# dynamic loading
 EXEC_LDFLAGS+=" -ldl"
 
 PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS $STDLIBS -lgcc -lstdc++"
 PLATFORM_LDFLAGS+=" -B$BINUTILS"
 
-EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $TBB_LIBS $LIBURING_LIBS $BENCHMARK_LIBS"
+EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $LIBURING_LIBS $BENCHMARK_LIBS"
 
 VALGRIND_VER="$VALGRIND_BASE/bin/"
 

build_tools/getdeps_fallback_mirror.py

@@ -1,62 +1,188 @@
 #!/usr/bin/env python3
+#  Copyright (c) Meta Platforms, Inc. and affiliates.
+#  This source code is licensed under both the GPLv2 (found in the COPYING file in the root directory)
+#  and the Apache 2.0 License (found in the LICENSE.Apache file in the root directory).
 """
 Pre-download packages with unreliable mirrors using fallback mirrors.
 Reads package info from folly's getdeps manifest files.
 """
-import sys
-import os
-import hashlib
-import subprocess
 import configparser
+import hashlib
+import os
+import shutil
+import sys
+import urllib.request
+
+DOWNLOAD_TIMEOUT_SECONDS = 120
+DOWNLOAD_CHUNK_BYTES = 64 * 1024
+MAX_DOWNLOAD_BYTES = 50 * 1024 * 1024
+
+MIRROR_FALLBACKS = {
+    "ftpmirror.gnu.org/gnu/": [
+        "https://mirrors.kernel.org/gnu/",
+        "https://ftpmirror.gnu.org/gnu/",
+        "https://ftp.gnu.org/gnu/",
+    ],
+    "ftp.gnu.org/gnu/": [
+        "https://mirrors.kernel.org/gnu/",
+        "https://ftpmirror.gnu.org/gnu/",
+        "https://ftp.gnu.org/gnu/",
+    ],
+}
+
+# These packages must have URLs matching MIRROR_FALLBACKS; other packages are
+# left for getdeps.py's normal download path.
+PACKAGES_TO_CHECK = ("autoconf", "automake", "libtool", "libiberty")
+
 
 def sha256_file(path):
     """Calculate SHA256 hash of a file."""
     h = hashlib.sha256()
     try:
-        with open(path, 'rb') as f:
-            for chunk in iter(lambda: f.read(65536), b''):
+        with open(path, "rb") as f:
+            for chunk in iter(lambda: f.read(65536), b""):
                 h.update(chunk)
         return h.hexdigest()
     except Exception:
         return None
 
+
 def parse_manifest(manifest_path):
     """Parse a getdeps manifest file to extract download info."""
-    config = configparser.ConfigParser()
+    # folly manifests can contain bare keys in sections unrelated to downloads.
+    config = configparser.ConfigParser(allow_no_value=True, interpolation=None)
     try:
-        config.read(manifest_path)
-        if 'download' in config:
-            return {
-                'url': config['download'].get('url', ''),
-                'sha256': config['download'].get('sha256', ''),
-            }
-    except Exception:
-        pass
+        with open(manifest_path, encoding="utf-8") as manifest_file:
+            config.read_file(manifest_file)
+    except Exception as ex:
+        print(f"  {os.path.basename(manifest_path)}: WARNING - parse failed: {ex}")
+        return None
+
+    if "download" in config:
+        return {
+            "url": config["download"].get("url", ""),
+            "sha256": config["download"].get("sha256", ""),
+        }
     return None
 
+
+def file_size(path):
+    try:
+        return os.path.getsize(path)
+    except Exception:
+        return None
+
+
 def get_fallback_mirrors(url):
     """Get fallback mirror URLs for a given URL."""
-    # Fallback mirror patterns for known unreliable hosts
-    mirror_fallbacks = {
-        "ftp.gnu.org/gnu/": [
-            "https://mirrors.kernel.org/gnu/",
-            "https://ftpmirror.gnu.org/gnu/",
-            "https://ftp.gnu.org/gnu/",
-        ],
-        "ftpmirror.gnu.org/gnu/": [
-            "https://mirrors.kernel.org/gnu/",
-            "https://ftpmirror.gnu.org/gnu/",
-            "https://ftp.gnu.org/gnu/",
-        ],
-    }
-
-    for pattern, mirrors in mirror_fallbacks.items():
+    for pattern, mirrors in MIRROR_FALLBACKS.items():
         if pattern in url:
             # Extract the path after the pattern
             path_start = url.find(pattern) + len(pattern)
             path = url[path_start:]
             return [mirror + path for mirror in mirrors]
-    return [url]  # No fallback, use original
+    return []
+
+
+def download_url(url, filepath):
+    """Download URL to filepath without leaving partial files behind."""
+    tmp_filepath = filepath + ".tmp"
+    if os.path.exists(tmp_filepath):
+        os.remove(tmp_filepath)
+
+    request = urllib.request.Request(
+        url, headers={"User-Agent": "rocksdb-getdeps-fallback/1.0"}
+    )
+    try:
+        with urllib.request.urlopen(
+            request, timeout=DOWNLOAD_TIMEOUT_SECONDS
+        ) as response, open(tmp_filepath, "wb") as output:
+            copied = 0
+            while True:
+                chunk = response.read(DOWNLOAD_CHUNK_BYTES)
+                if not chunk:
+                    break
+
+                copied += len(chunk)
+                if copied > MAX_DOWNLOAD_BYTES:
+                    raise Exception(
+                        f"download exceeds {MAX_DOWNLOAD_BYTES} bytes"
+                    )
+                output.write(chunk)
+        os.replace(tmp_filepath, filepath)
+    finally:
+        if os.path.exists(tmp_filepath):
+            os.remove(tmp_filepath)
+
+
+def prepare_download(package, info, download_dir, cache_dir):
+    url = info["url"]
+    expected_sha256 = info["sha256"]
+    mirrors = get_fallback_mirrors(url)
+    if not mirrors:
+        return False
+
+    if not expected_sha256:
+        print(f"  {package}: WARNING - skipped fallback without sha256")
+        return False
+
+    # getdeps uses format: {package}-{filename}
+    filename = f"{package}-{os.path.basename(url)}"
+    filepath = os.path.join(download_dir, filename)
+    cache_path = os.path.join(cache_dir, filename)
+
+    # Check if already valid.
+    actual_sha256 = sha256_file(filepath) if os.path.exists(filepath) else None
+    if actual_sha256 == expected_sha256:
+        print(f"  {filename}: OK (already downloaded)")
+        return True
+    if actual_sha256 is not None:
+        print(
+            f"  {filename}: WARNING - removing invalid download "
+            f"sha256={actual_sha256}"
+        )
+        os.remove(filepath)
+
+    # The cache is only an opportunistic single-build accelerator; callers
+    # should not share it across concurrent builds without external locking.
+    actual_sha256 = sha256_file(cache_path) if os.path.exists(cache_path) else None
+    if actual_sha256 == expected_sha256:
+        print(f"  {filename}: OK (from cache)")
+        shutil.copy2(cache_path, filepath)
+        return True
+    if actual_sha256 is not None:
+        print(
+            f"  {filename}: WARNING - removing invalid cache "
+            f"sha256={actual_sha256}"
+        )
+        os.remove(cache_path)
+
+    # Try fallback mirrors.
+    for mirror_url in mirrors:
+        print(f"  {filename}: trying {mirror_url}...")
+        try:
+            download_url(mirror_url, filepath)
+        except Exception as ex:
+            print(f"  {filename}: WARNING - download failed: {ex}")
+            continue
+
+        actual_sha256 = sha256_file(filepath)
+        if actual_sha256 == expected_sha256:
+            size = file_size(filepath)
+            print(f"  {filename}: OK (downloaded, {size} bytes)")
+            shutil.copy2(filepath, cache_path)
+            return True
+
+        size = file_size(filepath)
+        print(
+            f"  {filename}: WARNING - sha256 mismatch from {mirror_url}: "
+            f"expected={expected_sha256} actual={actual_sha256} size={size}"
+        )
+        os.remove(filepath)
+
+    print(f"  {filename}: WARNING - all mirrors failed")
+    return False
+
 
 def main():
     if len(sys.argv) != 4:
@@ -64,60 +190,39 @@ def main():
         sys.exit(1)
 
     download_dir, cache_dir, manifests_dir = sys.argv[1], sys.argv[2], sys.argv[3]
+    os.makedirs(download_dir, exist_ok=True)
+    os.makedirs(cache_dir, exist_ok=True)
 
-    # Packages known to have unreliable mirrors
-    packages_to_check = ["autoconf", "automake", "libtool"]
-
-    for package in packages_to_check:
+    checked = 0
+    ready = 0
+    for package in PACKAGES_TO_CHECK:
         manifest_path = os.path.join(manifests_dir, package)
-        if not os.path.exists(manifest_path):
+        if not os.path.isfile(manifest_path):
             continue
 
         info = parse_manifest(manifest_path)
-        if not info or not info['url'] or not info['sha256']:
+        if not info or not info["url"]:
             continue
 
-        # Determine filename from URL
-        url = info['url']
-        expected_sha256 = info['sha256']
-        url_filename = os.path.basename(url)
-
-        # getdeps uses format: {package}-{filename}
-        filename = f"{package}-{url_filename}"
-        filepath = os.path.join(download_dir, filename)
-        cache_path = os.path.join(cache_dir, filename)
-
-        # Check if already valid
-        if os.path.exists(filepath) and sha256_file(filepath) == expected_sha256:
-            print(f"  {filename}: OK (already downloaded)")
+        if not info["sha256"]:
+            print(f"  {package}: WARNING - skipped fallback without sha256")
             continue
 
-        # Check cache
-        if os.path.exists(cache_path) and sha256_file(cache_path) == expected_sha256:
-            print(f"  {filename}: OK (from cache)")
-            subprocess.run(['cp', cache_path, filepath], check=True)
+        if not get_fallback_mirrors(info["url"]):
+            print(
+                f"  {package}: WARNING - skipped fallback without known mirror "
+                f"for {info['url']}"
+            )
             continue
 
-        # Try fallback mirrors
-        mirrors = get_fallback_mirrors(url)
-        downloaded = False
-        for mirror_url in mirrors:
-            print(f"  {filename}: trying {mirror_url}...")
-            try:
-                subprocess.run(['wget', '-q', '-O', filepath, mirror_url], check=True, timeout=120)
-                if sha256_file(filepath) == expected_sha256:
-                    print(f"  {filename}: OK (downloaded)")
-                    subprocess.run(['cp', filepath, cache_path], check=False)
-                    downloaded = True
-                    break
-                else:
-                    os.remove(filepath)
-            except Exception:
-                if os.path.exists(filepath):
-                    os.remove(filepath)
+        checked += 1
+        try:
+            if prepare_download(package, info, download_dir, cache_dir):
+                ready += 1
+        except Exception as ex:
+            print(f"  {package}: WARNING - fallback preparation failed: {ex}")
 
-        if not downloaded:
-            print(f"  {filename}: WARNING - all mirrors failed")
+    print(f"  fallback mirror downloads ready: {ready}/{checked}")
 
 if __name__ == "__main__":
     main()

build_tools/gtest-parallel

@@ -0,0 +1,18 @@
+#!/usr/bin/env python3
+# Copyright 2017 Google Inc. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import gtest_parallel
+import sys
+
+sys.exit(gtest_parallel.main())

build_tools/gtest_parallel.py

@@ -0,0 +1,959 @@
+#  Copyright (c) Meta Platforms, Inc. and affiliates.
+#  This source code is licensed under both the GPLv2 (found in the COPYING file in the root directory)
+#  and the Apache 2.0 License (found in the LICENSE.Apache file in the root directory).
+
+# Copyright 2013 Google Inc. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import errno
+from functools import total_ordering
+import gzip
+import io
+import json
+import multiprocessing
+import optparse
+import os
+import re
+import shutil
+import signal
+import subprocess
+import sys
+import tempfile
+import threading
+import time
+
+if sys.version_info.major >= 3:
+  long = int
+  import _pickle as cPickle
+  import _thread as thread
+else:
+  import cPickle
+  import thread
+
+from pickle import HIGHEST_PROTOCOL as PICKLE_HIGHEST_PROTOCOL
+
+if sys.platform == 'win32':
+  import msvcrt
+else:
+  import fcntl
+
+
+# An object that catches SIGINT sent to the Python process and notices
+# if processes passed to wait() die by SIGINT (we need to look for
+# both of those cases, because pressing Ctrl+C can result in either
+# the main process or one of the subprocesses getting the signal).
+#
+# Before a SIGINT is seen, wait(p) will simply call p.wait() and
+# return the result. Once a SIGINT has been seen (in the main process
+# or a subprocess, including the one the current call is waiting for),
+# wait(p) will call p.terminate() and raise ProcessWasInterrupted.
+class SigintHandler(object):
+  class ProcessWasInterrupted(Exception):
+    pass
+
+  sigint_returncodes = {
+      -signal.SIGINT,  # Unix
+      -1073741510,  # Windows
+  }
+
+  def __init__(self):
+    self.__lock = threading.Lock()
+    self.__processes = set()
+    self.__got_sigint = False
+    signal.signal(signal.SIGINT, lambda signal_num, frame: self.interrupt())
+
+  def __on_sigint(self):
+    self.__got_sigint = True
+    while self.__processes:
+      try:
+        self.__processes.pop().terminate()
+      except OSError:
+        pass
+
+  def interrupt(self):
+    with self.__lock:
+      self.__on_sigint()
+
+  def got_sigint(self):
+    with self.__lock:
+      return self.__got_sigint
+
+  def wait(self, p, timeout_per_test):
+    with self.__lock:
+      if self.__got_sigint:
+        p.terminate()
+      self.__processes.add(p)
+    try:
+      code = p.wait(timeout_per_test)
+    except subprocess.TimeoutExpired :
+      p.terminate()
+      self.__processes.remove(p)
+      code = -errno.ETIME
+    with self.__lock:
+      self.__processes.discard(p)
+      if code in self.sigint_returncodes:
+        self.__on_sigint()
+      if self.__got_sigint:
+        raise self.ProcessWasInterrupted
+    return code
+
+
+sigint_handler = SigintHandler()
+
+
+# Return the width of the terminal, or None if it couldn't be
+# determined (e.g. because we're not being run interactively).
+def term_width(out):
+  if not out.isatty():
+    return None
+  try:
+    p = subprocess.Popen(["stty", "size"],
+                         stdout=subprocess.PIPE,
+                         stderr=subprocess.PIPE)
+    (out, err) = p.communicate()
+    if p.returncode != 0 or err:
+      return None
+    return int(out.split()[1])
+  except (IndexError, OSError, ValueError):
+    return None
+
+
+# Output transient and permanent lines of text. If several transient
+# lines are written in sequence, the new will overwrite the old. We
+# use this to ensure that lots of unimportant info (tests passing)
+# won't drown out important info (tests failing).
+class Outputter(object):
+  def __init__(self, out_file):
+    self.__out_file = out_file
+    self.__previous_line_was_transient = False
+    self.__width = term_width(out_file)  # Line width, or None if not a tty.
+
+  def transient_line(self, msg):
+    if self.__width is None:
+      self.__out_file.write(msg + "\n")
+      self.__out_file.flush()
+    else:
+      self.__out_file.write("\r" + msg[:self.__width].ljust(self.__width))
+      self.__previous_line_was_transient = True
+
+  def flush_transient_output(self):
+    if self.__previous_line_was_transient:
+      self.__out_file.write("\n")
+      self.__previous_line_was_transient = False
+
+  def permanent_line(self, msg):
+    self.flush_transient_output()
+    self.__out_file.write(msg + "\n")
+    if self.__width is None:
+      self.__out_file.flush()
+
+
+def get_save_file_path():
+  """Return path to file for saving transient data."""
+  if sys.platform == 'win32':
+    default_cache_path = os.path.join(os.path.expanduser('~'), 'AppData',
+                                      'Local')
+    cache_path = os.environ.get('LOCALAPPDATA', default_cache_path)
+  else:
+    # We don't use xdg module since it's not a standard.
+    default_cache_path = os.path.join(os.path.expanduser('~'), '.cache')
+    cache_path = os.environ.get('XDG_CACHE_HOME', default_cache_path)
+
+  if os.path.isdir(cache_path):
+    return os.path.join(cache_path, 'gtest-parallel')
+  else:
+    sys.stderr.write('Directory {} does not exist'.format(cache_path))
+    return os.path.join(os.path.expanduser('~'), '.gtest-parallel-times')
+
+
+@total_ordering
+class Task(object):
+  """Stores information about a task (single execution of a test).
+
+  This class stores information about the test to be executed (gtest binary and
+  test name), and its result (log file, exit code and runtime).
+  Each task is uniquely identified by the gtest binary, the test name and an
+  execution number that increases each time the test is executed.
+  Additionaly we store the last execution time, so that next time the test is
+  executed, the slowest tests are run first.
+  """
+
+  def __init__(self, test_binary, test_name, test_command, execution_number,
+               last_execution_time, output_dir):
+    self.test_name = test_name
+    self.output_dir = output_dir
+    self.test_binary = test_binary
+    self.test_command = test_command
+    self.execution_number = execution_number
+    self.last_execution_time = last_execution_time
+
+    self.exit_code = None
+    self.runtime_ms = None
+
+    self.test_id = (test_binary, test_name)
+    self.task_id = (test_binary, test_name, self.execution_number)
+
+    self.log_file = Task._logname(self.output_dir, self.test_binary, test_name,
+                                  self.execution_number)
+
+  def __sorting_key(self):
+    # Unseen or failing tests (both missing execution time) take precedence over
+    # execution time. Tests are greater (seen as slower) when missing times so
+    # that they are executed first.
+    return (1 if self.last_execution_time is None else 0,
+            self.last_execution_time)
+
+  def __eq__(self, other):
+    return self.__sorting_key() == other.__sorting_key()
+
+  def __ne__(self, other):
+    return not (self == other)
+
+  def __lt__(self, other):
+    return self.__sorting_key() < other.__sorting_key()
+
+  @staticmethod
+  def _normalize(string):
+    return re.sub('[^A-Za-z0-9]', '_', string)
+
+  @staticmethod
+  def _logname(output_dir, test_binary, test_name, execution_number):
+    # Store logs to temporary files if there is no output_dir.
+    if output_dir is None:
+      (log_handle, log_name) = tempfile.mkstemp(prefix='gtest_parallel_',
+                                                suffix=".log")
+      os.close(log_handle)
+      return log_name
+
+    log_name = '%s-%s-%d.log' % (Task._normalize(os.path.basename(test_binary)),
+                                 Task._normalize(test_name), execution_number)
+
+    return os.path.join(output_dir, log_name)
+
+  def run(self, timeout_per_test):
+    begin = time.time()
+    with open(self.log_file, 'w') as log:
+      task = subprocess.Popen(self.test_command, stdout=log, stderr=log)
+      try:
+        self.exit_code = sigint_handler.wait(task, timeout_per_test)
+      except sigint_handler.ProcessWasInterrupted:
+        thread.exit()
+    self.runtime_ms = int(1000 * (time.time() - begin))
+    self.last_execution_time = None if self.exit_code else self.runtime_ms
+
+
+class TaskManager(object):
+  """Executes the tasks and stores the passed, failed and interrupted tasks.
+
+  When a task is run, this class keeps track if it passed, failed or was
+  interrupted. After a task finishes it calls the relevant functions of the
+  Logger, TestResults and TestTimes classes, and in case of failure, retries the
+  test as specified by the --retry_failed flag.
+  """
+
+  def __init__(self, times, logger, test_results, task_factory, times_to_retry,
+               initial_execution_number):
+    self.times = times
+    self.logger = logger
+    self.test_results = test_results
+    self.task_factory = task_factory
+    self.times_to_retry = times_to_retry
+    self.initial_execution_number = initial_execution_number
+
+    self.global_exit_code = 0
+
+    self.passed = []
+    self.failed = []
+    self.started = {}
+    self.timed_out = []
+    self.execution_number = {}
+
+    self.lock = threading.Lock()
+
+  def __get_next_execution_number(self, test_id):
+    with self.lock:
+      next_execution_number = self.execution_number.setdefault(
+          test_id, self.initial_execution_number)
+      self.execution_number[test_id] += 1
+    return next_execution_number
+
+  def __register_start(self, task):
+    with self.lock:
+      self.started[task.task_id] = task
+
+  def register_exit(self, task):
+    self.logger.log_exit(task)
+    self.times.record_test_time(task.test_binary, task.test_name,
+                                task.last_execution_time)
+    if self.test_results:
+      self.test_results.log(task.test_name, task.runtime_ms / 1000.0,
+                            task.exit_code)
+
+    with self.lock:
+      self.started.pop(task.task_id)
+      if task.exit_code == 0:
+        self.passed.append(task)
+      elif task.exit_code == -errno.ETIME:
+        self.timed_out.append(task)
+      else:
+        self.failed.append(task)
+
+  def run_task(self, task, timeout_per_test):
+    for try_number in range(self.times_to_retry + 1):
+      self.__register_start(task)
+      task.run(timeout_per_test)
+      self.register_exit(task)
+
+      if task.exit_code == 0:
+        break
+
+      if try_number < self.times_to_retry:
+        execution_number = self.__get_next_execution_number(task.test_id)
+        # We need create a new Task instance. Each task represents a single test
+        # execution, with its own runtime, exit code and log file.
+        task = self.task_factory(task.test_binary, task.test_name,
+                                 task.test_command, execution_number,
+                                 task.last_execution_time, task.output_dir)
+
+    with self.lock:
+      if task.exit_code != 0:
+        self.global_exit_code = task.exit_code
+
+
+class FilterFormat(object):
+  def __init__(self, output_dir):
+    if sys.stdout.isatty():
+      # stdout needs to be unbuffered since the output is interactive.
+      if isinstance(sys.stdout, io.TextIOWrapper):
+        # workaround for https://bugs.python.org/issue17404
+        sys.stdout = io.TextIOWrapper(sys.stdout.detach(),
+                                      line_buffering=True,
+                                      write_through=True,
+                                      newline='\n')
+      else:
+        sys.stdout = os.fdopen(sys.stdout.fileno(), 'w', 0)
+
+    self.output_dir = output_dir
+
+    self.total_tasks = 0
+    self.finished_tasks = 0
+    self.out = Outputter(sys.stdout)
+    self.stdout_lock = threading.Lock()
+
+  def move_to(self, destination_dir, tasks):
+    if self.output_dir is None:
+      return
+
+    destination_dir = os.path.join(self.output_dir, destination_dir)
+    os.makedirs(destination_dir)
+    for task in tasks:
+      shutil.move(task.log_file, destination_dir)
+
+  def print_tests(self, message, tasks, print_try_number, print_test_command):
+    self.out.permanent_line("%s (%s/%s):" %
+                            (message, len(tasks), self.total_tasks))
+    for task in sorted(tasks):
+      runtime_ms = 'Interrupted'
+      if task.runtime_ms is not None:
+        runtime_ms = '%d ms' % task.runtime_ms
+      if print_test_command:
+        try:
+          cmd_str = " ".join(task.test_command)
+        except TypeError:
+          cmd_str = task.test_command
+        self.out.permanent_line(
+            "%11s: %s%s" %
+            (runtime_ms, cmd_str,
+            (" (try #%d)" % task.execution_number) if print_try_number else ""))
+      else:
+        self.out.permanent_line(
+            "%11s: %s %s%s" %
+            (runtime_ms, task.test_binary, task.test_name,
+            (" (try #%d)" % task.execution_number) if print_try_number else ""))
+
+  def log_exit(self, task):
+    with self.stdout_lock:
+      self.finished_tasks += 1
+      self.out.transient_line("[%d/%d] %s (%d ms)" %
+                              (self.finished_tasks, self.total_tasks,
+                               task.test_name, task.runtime_ms))
+      if task.exit_code != 0:
+        signal_name = None
+        if task.exit_code < 0:
+          try:
+            signal_name = signal.Signals(-task.exit_code).name
+          except ValueError:
+            pass
+
+        with open(task.log_file) as f:
+          for line in f.readlines():
+            self.out.permanent_line(line.rstrip())
+        if task.exit_code is None:
+          self.out.permanent_line("[%d/%d] %s aborted after %d ms" %
+                                  (self.finished_tasks, self.total_tasks,
+                                   task.test_name, task.runtime_ms))
+        elif task.exit_code == -errno.ETIME:
+          self.out.permanent_line(
+            "\033[31m[  TIMEOUT ]\033[0m %s timed out after %d s"
+            % (task.test_name, task.runtime_ms/1000))
+        elif signal_name is not None:
+          self.out.permanent_line(
+              "[%d/%d] %s killed by signal %s (%d ms)" %
+              (self.finished_tasks, self.total_tasks, task.test_name,
+               signal_name, task.runtime_ms))
+        else:
+          self.out.permanent_line(
+              "[%d/%d] %s returned with exit code %d (%d ms)" %
+              (self.finished_tasks, self.total_tasks, task.test_name,
+               task.exit_code, task.runtime_ms))
+
+    if self.output_dir is None:
+      # Try to remove the file 100 times (sleeping for 0.1 second in between).
+      # This is a workaround for a process handle seemingly holding on to the
+      # file for too long inside os.subprocess. This workaround is in place
+      # until we figure out a minimal repro to report upstream (or a better
+      # suspect) to prevent os.remove exceptions.
+      num_tries = 100
+      for i in range(num_tries):
+        try:
+          os.remove(task.log_file)
+        except OSError as e:
+          if e.errno is not errno.ENOENT:
+            if i is num_tries - 1:
+              self.out.permanent_line('Could not remove temporary log file: ' +
+                                      str(e))
+            else:
+              time.sleep(0.1)
+            continue
+        break
+
+  def log_tasks(self, total_tasks):
+    self.total_tasks += total_tasks
+    self.out.transient_line("[0/%d] Running tests..." % self.total_tasks)
+
+  def summarize(self, passed_tasks, failed_tasks, interrupted_tasks):
+    stats = {}
+
+    def add_stats(stats, task, idx):
+      task_key = (task.test_binary, task.test_name)
+      if not task_key in stats:
+        # (passed, failed, interrupted) task_key is added as tie breaker to get
+        # alphabetic sorting on equally-stable tests
+        stats[task_key] = [0, 0, 0, task_key]
+      stats[task_key][idx] += 1
+
+    for task in passed_tasks:
+      add_stats(stats, task, 0)
+    for task in failed_tasks:
+      add_stats(stats, task, 1)
+    for task in interrupted_tasks:
+      add_stats(stats, task, 2)
+
+    self.out.permanent_line("SUMMARY:")
+    for task_key in sorted(stats, key=stats.__getitem__):
+      (num_passed, num_failed, num_interrupted, _) = stats[task_key]
+      (test_binary, task_name) = task_key
+      total_runs = num_passed + num_failed + num_interrupted
+      if num_passed == total_runs:
+        continue
+      self.out.permanent_line("  %s %s passed %d / %d times%s." %
+                              (test_binary, task_name, num_passed, total_runs,
+                               "" if num_interrupted == 0 else
+                               (" (%d interrupted)" % num_interrupted)))
+
+  def flush(self):
+    self.out.flush_transient_output()
+
+
+class CollectTestResults(object):
+  def __init__(self, json_dump_filepath):
+    self.test_results_lock = threading.Lock()
+    self.json_dump_file = open(json_dump_filepath, 'w')
+    self.test_results = {
+        "interrupted": False,
+        "path_delimiter": ".",
+        # Third version of the file format. See the link in the flag description
+        # for details.
+        "version": 3,
+        "seconds_since_epoch": int(time.time()),
+        "num_failures_by_type": {
+            "PASS": 0,
+            "FAIL": 0,
+            "TIMEOUT": 0,
+        },
+        "tests": {},
+    }
+
+  def log(self, test, runtime_seconds, exit_code):
+    if exit_code is None:
+      actual_result = "TIMEOUT"
+    elif exit_code == 0:
+      actual_result = "PASS"
+    else:
+      actual_result = "FAIL"
+    with self.test_results_lock:
+      self.test_results['num_failures_by_type'][actual_result] += 1
+      results = self.test_results['tests']
+      for name in test.split('.'):
+        results = results.setdefault(name, {})
+
+      if results:
+        results['actual'] += ' ' + actual_result
+        results['times'].append(runtime_seconds)
+      else:  # This is the first invocation of the test
+        results['actual'] = actual_result
+        results['times'] = [runtime_seconds]
+        results['time'] = runtime_seconds
+        results['expected'] = 'PASS'
+
+  def dump_to_file_and_close(self):
+    json.dump(self.test_results, self.json_dump_file)
+    self.json_dump_file.close()
+
+
+# Record of test runtimes. Has built-in locking.
+class TestTimes(object):
+  class LockedFile(object):
+    def __init__(self, filename, mode):
+      self._filename = filename
+      self._mode = mode
+      self._fo = None
+
+    def __enter__(self):
+      self._fo = open(self._filename, self._mode)
+
+      # Regardless of opening mode we always seek to the beginning of file.
+      # This simplifies code working with LockedFile and also ensures that
+      # we lock (and unlock below) always the same region in file on win32.
+      self._fo.seek(0)
+
+      try:
+        if sys.platform == 'win32':
+          # We are locking here fixed location in file to use it as
+          # an exclusive lock on entire file.
+          msvcrt.locking(self._fo.fileno(), msvcrt.LK_LOCK, 1)
+        else:
+          fcntl.flock(self._fo.fileno(), fcntl.LOCK_EX)
+      except IOError:
+        self._fo.close()
+        raise
+
+      return self._fo
+
+    def __exit__(self, exc_type, exc_value, traceback):
+      # Flush any buffered data to disk. This is needed to prevent race
+      # condition which happens from the moment of releasing file lock
+      # till closing the file.
+      self._fo.flush()
+
+      try:
+        if sys.platform == 'win32':
+          self._fo.seek(0)
+          msvcrt.locking(self._fo.fileno(), msvcrt.LK_UNLCK, 1)
+        else:
+          fcntl.flock(self._fo.fileno(), fcntl.LOCK_UN)
+      finally:
+        self._fo.close()
+
+      return exc_value is None
+
+  def __init__(self, save_file):
+    "Create new object seeded with saved test times from the given file."
+    self.__times = {}  # (test binary, test name) -> runtime in ms
+
+    # Protects calls to record_test_time(); other calls are not
+    # expected to be made concurrently.
+    self.__lock = threading.Lock()
+
+    try:
+      with TestTimes.LockedFile(save_file, 'rb') as fd:
+        times = TestTimes.__read_test_times_file(fd)
+    except IOError:
+      # We couldn't obtain the lock.
+      return
+
+    # Discard saved times if the format isn't right.
+    if type(times) is not dict:
+      return
+    for ((test_binary, test_name), runtime) in times.items():
+      if (type(test_binary) is not str or type(test_name) is not str
+          or type(runtime) not in {int, long, type(None)}):
+        return
+
+    self.__times = times
+
+  def get_test_time(self, binary, testname):
+    """Return the last duration for the given test as an integer number of
+    milliseconds, or None if the test failed or if there's no record for it."""
+    return self.__times.get((binary, testname), None)
+
+  def record_test_time(self, binary, testname, runtime_ms):
+    """Record that the given test ran in the specified number of
+    milliseconds. If the test failed, runtime_ms should be None."""
+    with self.__lock:
+      self.__times[(binary, testname)] = runtime_ms
+
+  def write_to_file(self, save_file):
+    "Write all the times to file."
+    try:
+      with TestTimes.LockedFile(save_file, 'a+b') as fd:
+        times = TestTimes.__read_test_times_file(fd)
+
+        if times is None:
+          times = self.__times
+        else:
+          times.update(self.__times)
+
+        # We erase data from file while still holding a lock to it. This
+        # way reading old test times and appending new ones are atomic
+        # for external viewer.
+        fd.seek(0)
+        fd.truncate()
+        with gzip.GzipFile(fileobj=fd, mode='wb') as gzf:
+          cPickle.dump(times, gzf, PICKLE_HIGHEST_PROTOCOL)
+    except IOError:
+      pass  # ignore errors---saving the times isn't that important
+
+  @staticmethod
+  def __read_test_times_file(fd):
+    try:
+      with gzip.GzipFile(fileobj=fd, mode='rb') as gzf:
+        times = cPickle.load(gzf)
+    except Exception:
+      # File doesn't exist, isn't readable, is malformed---whatever.
+      # Just ignore it.
+      return None
+    else:
+      return times
+
+
+def find_tests(binaries, additional_args, options, times):
+  test_count = 0
+  tasks = []
+  for test_binary in binaries:
+    command = [test_binary] + additional_args
+    if options.gtest_also_run_disabled_tests:
+      command += ['--gtest_also_run_disabled_tests']
+
+    list_command = command + ['--gtest_list_tests']
+    if options.gtest_filter != '':
+      list_command += ['--gtest_filter=' + options.gtest_filter]
+
+    try:
+      test_list = subprocess.check_output(list_command,
+                                          stderr=subprocess.STDOUT)
+    except subprocess.CalledProcessError as e:
+      sys.exit("%s: %s\n%s" % (test_binary, str(e), e.output))
+
+    try:
+      test_list = test_list.split('\n')
+    except TypeError:
+      # subprocess.check_output() returns bytes in python3
+      test_list = test_list.decode(sys.stdout.encoding).split('\n')
+
+    command += ['--gtest_color=' + options.gtest_color]
+
+    test_group = ''
+    for line in test_list:
+      if not line.strip():
+        continue
+      if line[0] != " ":
+        # Remove comments for typed tests and strip whitespace.
+        test_group = line.split('#')[0].strip()
+        continue
+      # Remove comments for parameterized tests and strip whitespace.
+      line = line.split('#')[0].strip()
+      if not line:
+        continue
+
+      test_name = test_group + line
+      if not options.gtest_also_run_disabled_tests and 'DISABLED_' in test_name:
+        continue
+
+      # Skip PRE_ tests which are used by Chromium.
+      if '.PRE_' in test_name:
+        continue
+
+      last_execution_time = times.get_test_time(test_binary, test_name)
+      if options.failed and last_execution_time is not None:
+        continue
+
+      test_command = command + ['--gtest_filter=' + test_name]
+      if (test_count - options.shard_index) % options.shard_count == 0:
+        for execution_number in range(options.repeat):
+          tasks.append(
+              Task(test_binary, test_name, test_command, execution_number + 1,
+                   last_execution_time, options.output_dir))
+
+      test_count += 1
+
+  # Sort the tasks to run the slowest tests first, so that faster ones can be
+  # finished in parallel.
+  return sorted(tasks, reverse=True)
+
+
+def execute_tasks(tasks, pool_size, task_manager, timeout_seconds,
+                  timeout_per_test, serialize_test_cases):
+  class WorkerFn(object):
+    def __init__(self, tasks, running_groups, timeout_per_test):
+      self.tasks = tasks
+      self.running_groups = running_groups
+      self.timeout_per_test = timeout_per_test
+      self.task_lock = threading.Lock()
+
+    def __call__(self):
+      while True:
+        with self.task_lock:
+          for task_id in range(len(self.tasks)):
+            task = self.tasks[task_id]
+
+            if self.running_groups is not None:
+              test_group = task.test_name.split('.')[0]
+              if test_group in self.running_groups:
+                # Try to find other non-running test group.
+                continue
+              else:
+                self.running_groups.add(test_group)
+
+            del self.tasks[task_id]
+            break
+          else:
+            # Either there is no tasks left or number or remaining test
+            # cases (groups) is less than number or running threads.
+            return
+
+        task_manager.run_task(task, self.timeout_per_test)
+
+        if self.running_groups is not None:
+          with self.task_lock:
+            self.running_groups.remove(test_group)
+
+  def start_daemon(func):
+    t = threading.Thread(target=func)
+    t.daemon = True
+    t.start()
+    return t
+
+  timeout = None
+  try:
+    if timeout_seconds:
+      timeout = threading.Timer(timeout_seconds, sigint_handler.interrupt)
+      timeout.start()
+    running_groups = set() if serialize_test_cases else None
+    worker_fn = WorkerFn(tasks, running_groups, timeout_per_test)
+    workers = [start_daemon(worker_fn) for _ in range(pool_size)]
+    for worker in workers:
+      worker.join()
+  finally:
+    if timeout:
+      timeout.cancel()
+      for task in list(task_manager.started.values()):
+        task.runtime_ms = timeout_seconds * 1000
+        task_manager.register_exit(task)
+
+
+def default_options_parser():
+  parser = optparse.OptionParser(
+      usage='usage: %prog [options] binary [binary ...] -- [additional args]')
+
+  parser.add_option('-d',
+                    '--output_dir',
+                    type='string',
+                    default=None,
+                    help='Output directory for test logs. Logs will be '
+                    'available under gtest-parallel-logs/, so '
+                    '--output_dir=/tmp will results in all logs being '
+                    'available under /tmp/gtest-parallel-logs/.')
+  parser.add_option('-r',
+                    '--repeat',
+                    type='int',
+                    default=1,
+                    help='Number of times to execute all the tests.')
+  parser.add_option('--retry_failed',
+                    type='int',
+                    default=0,
+                    help='Number of times to repeat failed tests.')
+  parser.add_option('--failed',
+                    action='store_true',
+                    default=False,
+                    help='run only failed and new tests')
+  parser.add_option('-w',
+                    '--workers',
+                    type='int',
+                    default=multiprocessing.cpu_count(),
+                    help='number of workers to spawn')
+  parser.add_option('--gtest_color',
+                    type='string',
+                    default='yes',
+                    help='color output')
+  parser.add_option('--gtest_filter',
+                    type='string',
+                    default='',
+                    help='test filter')
+  parser.add_option('--gtest_also_run_disabled_tests',
+                    action='store_true',
+                    default=False,
+                    help='run disabled tests too')
+  parser.add_option(
+      '--print_test_times',
+      action='store_true',
+      default=False,
+      help='list the run time of each test at the end of execution')
+  parser.add_option(
+      '--print_test_command',
+      action='store_true',
+      default=False,
+      help='Print full test command instead of name')
+  parser.add_option('--shard_count',
+                    type='int',
+                    default=1,
+                    help='total number of shards (for sharding test execution '
+                    'between multiple machines)')
+  parser.add_option('--shard_index',
+                    type='int',
+                    default=0,
+                    help='zero-indexed number identifying this shard (for '
+                    'sharding test execution between multiple machines)')
+  parser.add_option(
+      '--dump_json_test_results',
+      type='string',
+      default=None,
+      help='Saves the results of the tests as a JSON machine-'
+      'readable file. The format of the file is specified at '
+      'https://www.chromium.org/developers/the-json-test-results-format')
+  parser.add_option('--timeout',
+                    type='int',
+                    default=None,
+                    help='Interrupt all remaining processes after the given '
+                    'time (in seconds).')
+  parser.add_option('--timeout_per_test',
+                    type='int',
+                    default=None,
+                    help='Interrupt single processes after the given '
+                    'time (in seconds).')
+  parser.add_option('--serialize_test_cases',
+                    action='store_true',
+                    default=False,
+                    help='Do not run tests from the same test '
+                    'case in parallel.')
+  return parser
+
+
+def main():
+  # Remove additional arguments (anything after --).
+  additional_args = []
+
+  for i in range(len(sys.argv)):
+    if sys.argv[i] == '--':
+      additional_args = sys.argv[i + 1:]
+      sys.argv = sys.argv[:i]
+      break
+
+  parser = default_options_parser()
+  (options, binaries) = parser.parse_args()
+
+  if (options.output_dir is not None and not os.path.isdir(options.output_dir)):
+    parser.error('--output_dir value must be an existing directory, '
+                 'current value is "%s"' % options.output_dir)
+
+  # Append gtest-parallel-logs to log output, this is to avoid deleting user
+  # data if an user passes a directory where files are already present. If a
+  # user specifies --output_dir=Docs/, we'll create Docs/gtest-parallel-logs
+  # and clean that directory out on startup, instead of nuking Docs/.
+  if options.output_dir:
+    options.output_dir = os.path.join(options.output_dir, 'gtest-parallel-logs')
+
+  if binaries == []:
+    parser.print_usage()
+    sys.exit(1)
+
+  if options.shard_count < 1:
+    parser.error("Invalid number of shards: %d. Must be at least 1." %
+                 options.shard_count)
+  if not (0 <= options.shard_index < options.shard_count):
+    parser.error("Invalid shard index: %d. Must be between 0 and %d "
+                 "(less than the number of shards)." %
+                 (options.shard_index, options.shard_count - 1))
+
+  # Check that all test binaries have an unique basename. That way we can ensure
+  # the logs are saved to unique files even when two different binaries have
+  # common tests.
+  unique_binaries = set(os.path.basename(binary) for binary in binaries)
+  assert len(unique_binaries) == len(binaries), (
+      "All test binaries must have an unique basename.")
+
+  if options.output_dir:
+    # Remove files from old test runs.
+    if os.path.isdir(options.output_dir):
+      shutil.rmtree(options.output_dir)
+    # Create directory for test log output.
+    try:
+      os.makedirs(options.output_dir)
+    except OSError as e:
+      # Ignore errors if this directory already exists.
+      if e.errno != errno.EEXIST or not os.path.isdir(options.output_dir):
+        raise e
+
+  test_results = None
+  if options.dump_json_test_results is not None:
+    test_results = CollectTestResults(options.dump_json_test_results)
+
+  save_file = get_save_file_path()
+
+  times = TestTimes(save_file)
+  logger = FilterFormat(options.output_dir)
+
+  task_manager = TaskManager(times, logger, test_results, Task,
+                             options.retry_failed, options.repeat + 1)
+
+  tasks = find_tests(binaries, additional_args, options, times)
+  logger.log_tasks(len(tasks))
+  execute_tasks(tasks, options.workers, task_manager, options.timeout,
+                options.timeout_per_test, options.serialize_test_cases)
+
+  print_try_number = options.retry_failed > 0 or options.repeat > 1
+  if task_manager.passed:
+    logger.move_to('passed', task_manager.passed)
+    if options.print_test_times:
+      logger.print_tests('PASSED TESTS', task_manager.passed, print_try_number, options.print_test_command)
+
+  if task_manager.failed:
+    logger.print_tests('FAILED TESTS', task_manager.failed, print_try_number, options.print_test_command)
+    logger.move_to('failed', task_manager.failed)
+
+  if task_manager.timed_out:
+    logger.print_tests('TIMED OUT TESTS', task_manager.timed_out, print_try_number, options.print_test_command)
+    logger.move_to('timed_out', task_manager.timed_out)
+
+  if task_manager.started:
+    logger.print_tests('INTERRUPTED TESTS', task_manager.started.values(),
+                       print_try_number, options.print_test_command)
+    logger.move_to('interrupted', task_manager.started.values())
+
+  if options.repeat > 1 and (task_manager.failed or task_manager.started):
+    logger.summarize(task_manager.passed, task_manager.failed,
+                     task_manager.started.values())
+
+  logger.flush()
+  times.write_to_file(save_file)
+  if test_results:
+    test_results.dump_to_file_and_close()
+
+  if sigint_handler.got_sigint():
+    return -signal.SIGINT
+
+  return task_manager.global_exit_code
+
+
+if __name__ == "__main__":
+  sys.exit(main())

build_tools/rocksptest.sh

@@ -0,0 +1,67 @@
+#!/usr/bin/env bash
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# This source code is licensed under both the GPLv2 (found in the
+# COPYING file in the root directory) and Apache 2.0 License
+# (found in the LICENSE.Apache file in the root directory).
+#
+# Build RocksDB unit test binaries and run them under gtest-parallel,
+# which shards the test cases across CPUs for faster execution.
+#
+# Hardened version of a simple `make <bin> && gtest-parallel ./<bin>` helper:
+#  * AUTO_CLEAN=1 so the Makefile automatically runs a clean when the build
+#    parameters (DEBUG_LEVEL, sanitizers, ASSERT_STATUS_CHECKED, RTTI, etc.)
+#    have changed since the last build, preventing stale/mixed object files.
+#  * Parallel build with -j<NCORES>, computed the same way the Makefile does.
+#  * Uses the gtest-parallel checked in alongside this script (build_tools/),
+#    so it works regardless of PATH.
+#  * `set -euo pipefail` so any failure stops the script.
+#
+# Run from the repository root.
+#
+# Accepts one or more test binaries (gtest-parallel pools all their test cases
+# into one shared worker queue). The leading non-option arguments are treated as
+# binaries; everything from the first option onward is forwarded to
+# gtest-parallel / the test binaries.
+#
+# Example usage:
+#   build_tools/rocksptest.sh db_test
+#   build_tools/rocksptest.sh db_test -r1000 --gtest_filter=*MixedSlowdown*
+#   build_tools/rocksptest.sh db_test env_test db_basic_test
+#   build_tools/rocksptest.sh db_test env_test --gtest_filter=*Foo*
+
+set -euo pipefail
+
+if [ "$#" -lt 1 ]; then
+  echo "usage: $0 <test_binary> [more_test_binaries...] [gtest-parallel/test args...]" >&2
+  echo "example: $0 db_test env_test -r1000 --gtest_filter=*MixedSlowdown*" >&2
+  exit 1
+fi
+
+# First argument must be a test binary, not an option (catches a forgotten name).
+if [ "${1#-}" != "$1" ]; then
+  echo "$0: first argument must be a test binary name, not an option ('$1')" >&2
+  exit 1
+fi
+
+# Collect the leading non-option arguments as test binaries; everything from the
+# first option onward is forwarded to gtest-parallel / the test binaries.
+BINS=()
+while [ "$#" -gt 0 ] && [ "${1#-}" = "$1" ]; do
+  BINS+=("$1")
+  shift
+done
+
+# Paths as gtest-parallel expects them (relative to the current directory).
+BIN_PATHS=()
+for b in "${BINS[@]}"; do
+  BIN_PATHS+=("./$b")
+done
+
+# Directory of this script, so we use the gtest-parallel checked in next to it.
+SCRIPT_DIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+
+# Compute parallelism the same way the Makefile computes NCORES.
+NCORES=$(nproc 2>/dev/null || sysctl -n hw.ncpu 2>/dev/null || echo 4)
+
+make AUTO_CLEAN=1 -j"$NCORES" "${BINS[@]}"
+"$SCRIPT_DIR/gtest-parallel" "${BIN_PATHS[@]}" "$@"

build_tools/rockstest.sh

@@ -0,0 +1,76 @@
+#!/usr/bin/env bash
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# This source code is licensed under both the GPLv2 (found in the
+# COPYING file in the root directory) and Apache 2.0 License
+# (found in the LICENSE.Apache file in the root directory).
+#
+# Build a single RocksDB unit test binary and run it directly (serially).
+# Only recommended with --gtest_filter=... because of speed. See also
+# build_tools/rocksptest.sh
+#
+# Hardened version of a simple `make <bin> && ./<bin>` helper:
+#  * AUTO_CLEAN=1 so the Makefile automatically runs a clean when the build
+#    parameters (DEBUG_LEVEL, sanitizers, ASSERT_STATUS_CHECKED, RTTI, etc.)
+#    have changed since the last build, preventing stale/mixed object files.
+#  * Parallel build with -j<NCORES>, computed the same way the Makefile does.
+#  * `set -euo pipefail` so any failure stops the script.
+#
+# Run from the repository root.
+#
+# Example usage:
+#   build_tools/rockstest.sh db_test --gtest_filter=*MixedSlowdown*
+#   build_tools/rockstest.sh env_test # Slow to run many tests serially
+#
+# Install mode:
+#   build_tools/rockstest.sh install
+#     Creates ~/bin/rockstest and ~/bin/rocksptest shims that defer to
+#     build_tools/rockstest.sh / rocksptest.sh in whatever directory you run
+#     them from, so you can just type `rockstest db_test` from any rocksdb
+#     source root.
+
+set -euo pipefail
+
+# Install mode: write thin ~/bin shims that defer to the build_tools scripts in
+# the current directory.
+if [ "${1:-}" = "install" ]; then
+  mkdir -p "$HOME/bin"
+  for name in rockstest rocksptest; do
+    dest="$HOME/bin/$name"
+    cat > "$dest" <<EOF
+#!/usr/bin/env bash
+# Auto-generated by 'build_tools/rockstest.sh install'. Defers to
+# build_tools/$name.sh in the current rocksdb source root.
+if [ -x build_tools/$name.sh ]; then
+  exec build_tools/$name.sh "\$@"
+else
+  echo "build_tools/$name.sh not found (Not in a rocksdb source root directory?)" >&2
+  exit 1
+fi
+EOF
+    chmod +x "$dest"
+    echo "Installed $dest"
+  done
+  exit 0
+fi
+
+if [ "$#" -lt 1 ]; then
+  echo "usage: $0 <test_binary> [test args...]" >&2
+  echo "       $0 install   # create ~/bin/rockstest and ~/bin/rocksptest shims" >&2
+  echo "example: $0 db_test --gtest_filter=*MixedSlowdown*" >&2
+  exit 1
+fi
+
+# First argument must be a test binary, not an option (catches a forgotten name).
+if [ "${1#-}" != "$1" ]; then
+  echo "$0: first argument must be a test binary name, not an option ('$1')" >&2
+  exit 1
+fi
+
+BIN=$1
+shift
+
+# Compute parallelism the same way the Makefile computes NCORES.
+NCORES=$(nproc 2>/dev/null || sysctl -n hw.ncpu 2>/dev/null || echo 4)
+
+make AUTO_CLEAN=1 -j"$NCORES" "$BIN"
+./"$BIN" "$@"

build_tools/ubuntu22_image/Dockerfile

@@ -5,6 +5,8 @@
 # the docker and docker-registry groups, and logging out and back in to pick
 # those up.)
 #
+# Meta employees: see https://fburl.com/rocksdb-docker to build this on a devvm.
+#
 # Follow https://docs.github.com/en/packages/working-with-a-github-packages-registry/working-with-the-container-registry#authenticating-with-a-personal-access-token-classic
 # to login with your GitHub credentials, as in
 #
@@ -15,8 +17,8 @@
 # Then in the build_tools/ubuntu22_image directory, (bump minor version for
 # random docker file updates, major version tracks Ubuntu release)
 #
-# $ docker build -t ghcr.io/facebook/rocksdb_ubuntu:22.0
-# $ docker push ghcr.io/facebook/rocksdb_ubuntu:22.0
+# $ docker build -t ghcr.io/facebook/rocksdb_ubuntu:22.2
+# $ docker push ghcr.io/facebook/rocksdb_ubuntu:22.2
 #
 # Might need to change visibility to public through
 # https://github.com/orgs/facebook/packages/container/rocksdb_ubuntu/settings
@@ -28,7 +30,7 @@ FROM ubuntu:22.04
 RUN apt-get update
 RUN apt-get upgrade -y
 # install basic tools
-RUN apt-get install -y vim wget curl
+RUN apt-get install -y vim wget curl ccache
 # install tzdata noninteractive
 RUN DEBIAN_FRONTEND=noninteractive TZ=Etc/UTC apt-get -y install tzdata
 # install git and default compilers

build_tools/ubuntu24_image/Dockerfile

@@ -5,6 +5,8 @@
 # the docker and docker-registry groups, and logging out and back in to pick
 # those up.)
 #
+# Meta employees: see https://fburl.com/rocksdb-docker to build this on a devvm.
+#
 # Follow https://docs.github.com/en/packages/working-with-a-github-packages-registry/working-with-the-container-registry#authenticating-with-a-personal-access-token-classic
 # to login with your GitHub credentials, as in
 #
@@ -15,8 +17,8 @@
 # Then in the build_tools/ubuntu24_image directory, (bump minor version for
 # random docker file updates, major version tracks Ubuntu release)
 #
-# $ docker build -t ghcr.io/facebook/rocksdb_ubuntu:24.0
-# $ docker push ghcr.io/facebook/rocksdb_ubuntu:24.0
+# $ docker build -t ghcr.io/facebook/rocksdb_ubuntu:24.1
+# $ docker push ghcr.io/facebook/rocksdb_ubuntu:24.1
 #
 # Might need to change visibility to public through
 # https://github.com/orgs/facebook/packages/container/rocksdb_ubuntu/settings
@@ -28,11 +30,15 @@ FROM ubuntu:24.04
 RUN apt-get update
 RUN apt-get upgrade -y
 # install basic tools
-RUN apt-get install -y vim wget curl
+RUN apt-get install -y vim wget curl ccache
 # install tzdata noninteractive
 RUN DEBIAN_FRONTEND=noninteractive TZ=Etc/UTC apt-get -y install tzdata
 # install git and default compilers
 RUN apt-get install -y git gcc g++ clang clang-tools
+# install clang-21 from LLVM snapshot repo
+RUN curl -fsSL https://apt.llvm.org/llvm-snapshot.gpg.key | gpg --dearmor -o /etc/apt/trusted.gpg.d/llvm.gpg && \
+    echo "deb https://apt.llvm.org/noble/ llvm-toolchain-noble-21 main" > /etc/apt/sources.list.d/llvm-21.list && \
+    apt-get update && apt-get install -y clang-21
 # install basic package
 RUN apt-get install -y lsb-release software-properties-common gnupg
 # install gflags, tbb

build_tools/update_dependencies.sh

@@ -32,7 +32,7 @@ function get_lib_base()
   local lib_platform=$3
 
   local result="$TP2_LATEST/$lib_name/"
-  
+
   # Lib Version
   if [ -z "$lib_version" ] || [ "$lib_version" = "LATEST" ]; then
     # version is not provided, use latest
@@ -40,7 +40,7 @@ function get_lib_base()
   else
     result="$result/$lib_version/"
   fi
-  
+
   # Lib Platform
   if [ -z "$lib_platform" ]; then
     # platform is not provided, use latest gcc
@@ -49,17 +49,17 @@ function get_lib_base()
     echo $lib_platform
     result="$result/$lib_platform/"
   fi
-  
+
   result=`ls -1d $result/*/ | head -n1`
 
   echo Finding link $result
-  
+
   # lib_name => LIB_NAME_BASE
   local __res_var=${lib_name^^}"_BASE"
   __res_var=`echo $__res_var | tr - _`
   # LIB_NAME_BASE=$result
   eval $__res_var=`readlink -f $result`
-  
+
   log_variable $__res_var
 }
 
@@ -75,17 +75,33 @@ touch "$OUTPUT"
 echo "Writing dependencies to $OUTPUT"
 
 # Compilers locations
-GCC_BASE=`readlink -f $TP2_LATEST/gcc/11.x/centos8-native/*/`
-CLANG_BASE=`readlink -f $TP2_LATEST/llvm-fb/15/platform010/*/`
+# GCC is pinned to 11.x because the only newer GCC in third-party2 (13.x) is
+# built for centos9/glibc>=2.35 -- its libgcc_s.so.1 has a hard reference
+# to _dl_find_object@GLIBC_2.35, but platform010 ships glibc 2.34. Bumping
+# GCC requires a platform with glibc >= 2.35.
+GCC_BASE=`readlink -f $TP2_LATEST/gcc/11.x/centos9-native/*/`
+# Clang is pinned to the latest tested major (21). Bump deliberately.
+CLANG_BASE=`readlink -f $TP2_LATEST/llvm-fb/21/platform010/*/`
 
 log_header
 log_variable GCC_BASE
 log_variable CLANG_BASE
 
 # Libraries locations
+# libgcc is pinned to 11.x to match the GCC 11 compiler above (libstdc++/
+# libgcc runtime, ABI, and C++ headers). Bump in lockstep with GCC_BASE.
 get_lib_base libgcc     11.x    platform010
+# glibc 2.34 is the platform010 ABI baseline (ld.so + libc); it is also the
+# only version available in third-party2, and defines the platform -- do
+# not bump independently.
 get_lib_base glibc      2.34    platform010
 get_lib_base snappy     LATEST  platform010
+# zlib is pinned to 1.2.8: it is the latest version with an x86_64
+# platform010 build (1.2.13 / 1.3.1 are centos9 / aarch64 only). At time of
+# writing, LATEST here doesn't work: get_lib_base picks the newest version
+# dir first and only then appends the platform, with no fallback -- so
+# LATEST would resolve to 1.3.1, find no platform010 build, and emit an
+# empty ZLIB_BASE.
 get_lib_base zlib       1.2.8   platform010
 get_lib_base bzip2      LATEST  platform010
 get_lib_base lz4        LATEST  platform010
@@ -94,12 +110,11 @@ get_lib_base gflags     LATEST  platform010
 get_lib_base jemalloc   LATEST  platform010
 get_lib_base numa       LATEST  platform010
 get_lib_base libunwind  LATEST  platform010
-get_lib_base tbb        2018_U5 platform010
 get_lib_base liburing   LATEST  platform010
 get_lib_base benchmark  LATEST  platform010
 
 get_lib_base kernel-headers fb platform010
-get_lib_base binutils   LATEST centos8-native
+get_lib_base binutils   LATEST centos9-native
 get_lib_base valgrind   LATEST platform010
 
 git diff $OUTPUT

c_api_gen/c_generated_block_based_options_subset.cc.inc

@@ -0,0 +1,39 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/generate_c_api.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Source: tools/c_api_gen/spec.json
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/generate_c_api.py
+//     --spec tools/c_api_gen/spec.json --section 'BlockBasedOptions simple'
+//     --header-out
+//     c_api_gen/c_generated_block_based_options_subset.h.inc
+//     --source-out
+//     c_api_gen/c_generated_block_based_options_subset.cc.inc
+
+/* BlockBasedOptions simple */
+
+void rocksdb_block_based_options_set_data_block_hash_ratio(
+    rocksdb_block_based_table_options_t* options, double v) {
+  options->rep.data_block_hash_table_util_ratio = v;
+}
+
+void rocksdb_block_based_options_set_top_level_index_pinning_tier(
+    rocksdb_block_based_table_options_t* options, int v) {
+  options->rep.metadata_cache_options.top_level_index_pinning =
+      static_cast<ROCKSDB_NAMESPACE::PinningTier>(v);
+}
+
+void rocksdb_block_based_options_set_partition_pinning_tier(
+    rocksdb_block_based_table_options_t* options, int v) {
+  options->rep.metadata_cache_options.partition_pinning =
+      static_cast<ROCKSDB_NAMESPACE::PinningTier>(v);
+}
+
+void rocksdb_block_based_options_set_unpartitioned_pinning_tier(
+    rocksdb_block_based_table_options_t* options, int v) {
+  options->rep.metadata_cache_options.unpartitioned_pinning =
+      static_cast<ROCKSDB_NAMESPACE::PinningTier>(v);
+}

c_api_gen/c_generated_block_based_options_subset.h.inc

@@ -0,0 +1,32 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/generate_c_api.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Source: tools/c_api_gen/spec.json
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/generate_c_api.py
+//     --spec tools/c_api_gen/spec.json --section 'BlockBasedOptions simple'
+//     --header-out
+//     c_api_gen/c_generated_block_based_options_subset.h.inc
+//     --source-out
+//     c_api_gen/c_generated_block_based_options_subset.cc.inc
+
+/* BlockBasedOptions simple */
+
+extern ROCKSDB_LIBRARY_API void
+rocksdb_block_based_options_set_data_block_hash_ratio(
+    rocksdb_block_based_table_options_t* options, double v);
+
+extern ROCKSDB_LIBRARY_API void
+rocksdb_block_based_options_set_top_level_index_pinning_tier(
+    rocksdb_block_based_table_options_t* options, int v);
+
+extern ROCKSDB_LIBRARY_API void
+rocksdb_block_based_options_set_partition_pinning_tier(
+    rocksdb_block_based_table_options_t* options, int v);
+
+extern ROCKSDB_LIBRARY_API void
+rocksdb_block_based_options_set_unpartitioned_pinning_tier(
+    rocksdb_block_based_table_options_t* options, int v);

c_api_gen/c_generated_cuckoo_options_subset.cc.inc

@@ -0,0 +1,21 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/generate_c_api.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Source: tools/c_api_gen/spec.json
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/generate_c_api.py
+//     --spec tools/c_api_gen/spec.json --section 'CuckooOptions simple'
+//     --header-out
+//     c_api_gen/c_generated_cuckoo_options_subset.h.inc
+//     --source-out
+//     c_api_gen/c_generated_cuckoo_options_subset.cc.inc
+
+/* CuckooOptions simple */
+
+void rocksdb_cuckoo_options_set_hash_ratio(
+    rocksdb_cuckoo_table_options_t* options, double v) {
+  options->rep.hash_table_ratio = v;
+}

c_api_gen/c_generated_cuckoo_options_subset.h.inc

@@ -0,0 +1,19 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/generate_c_api.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Source: tools/c_api_gen/spec.json
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/generate_c_api.py
+//     --spec tools/c_api_gen/spec.json --section 'CuckooOptions simple'
+//     --header-out
+//     c_api_gen/c_generated_cuckoo_options_subset.h.inc
+//     --source-out
+//     c_api_gen/c_generated_cuckoo_options_subset.cc.inc
+
+/* CuckooOptions simple */
+
+extern ROCKSDB_LIBRARY_API void rocksdb_cuckoo_options_set_hash_ratio(
+    rocksdb_cuckoo_table_options_t* options, double v);

c_api_gen/c_generated_db_simple_subset.cc.inc

@@ -0,0 +1,189 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/generate_c_api.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Source: tools/c_api_gen/spec.json
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/generate_c_api.py
+//     --spec tools/c_api_gen/spec.json --section 'DB data operations'
+//     --header-out
+//     c_api_gen/c_generated_db_simple_subset.h.inc
+//     --source-out
+//     c_api_gen/c_generated_db_simple_subset.cc.inc
+
+/* DB data operations */
+
+void rocksdb_put(rocksdb_t* db, const rocksdb_writeoptions_t* options,
+                 const char* key, size_t keylen, const char* val, size_t vallen,
+                 char** errptr) {
+  SaveError(errptr,
+            db->rep->Put(options->rep, Slice(key, keylen), Slice(val, vallen)));
+}
+
+void rocksdb_put_cf(rocksdb_t* db, const rocksdb_writeoptions_t* options,
+                    rocksdb_column_family_handle_t* column_family,
+                    const char* key, size_t keylen, const char* val,
+                    size_t vallen, char** errptr) {
+  SaveError(errptr, db->rep->Put(options->rep, column_family->rep,
+                                 Slice(key, keylen), Slice(val, vallen)));
+}
+
+void rocksdb_delete(rocksdb_t* db, const rocksdb_writeoptions_t* options,
+                    const char* key, size_t keylen, char** errptr) {
+  SaveError(errptr, db->rep->Delete(options->rep, Slice(key, keylen)));
+}
+
+void rocksdb_delete_cf(rocksdb_t* db, const rocksdb_writeoptions_t* options,
+                       rocksdb_column_family_handle_t* column_family,
+                       const char* key, size_t keylen, char** errptr) {
+  SaveError(errptr, db->rep->Delete(options->rep, column_family->rep,
+                                    Slice(key, keylen)));
+}
+
+void rocksdb_merge(rocksdb_t* db, const rocksdb_writeoptions_t* options,
+                   const char* key, size_t keylen, const char* val,
+                   size_t vallen, char** errptr) {
+  SaveError(errptr, db->rep->Merge(options->rep, Slice(key, keylen),
+                                   Slice(val, vallen)));
+}
+
+void rocksdb_merge_cf(rocksdb_t* db, const rocksdb_writeoptions_t* options,
+                      rocksdb_column_family_handle_t* column_family,
+                      const char* key, size_t keylen, const char* val,
+                      size_t vallen, char** errptr) {
+  SaveError(errptr, db->rep->Merge(options->rep, column_family->rep,
+                                   Slice(key, keylen), Slice(val, vallen)));
+}
+
+void rocksdb_write(rocksdb_t* db, const rocksdb_writeoptions_t* options,
+                   rocksdb_writebatch_t* batch, char** errptr) {
+  SaveError(errptr, db->rep->Write(options->rep, &batch->rep));
+}
+
+void rocksdb_put_with_ts(rocksdb_t* db, const rocksdb_writeoptions_t* options,
+                         const char* key, size_t keylen, const char* ts,
+                         size_t tslen, const char* val, size_t vallen,
+                         char** errptr) {
+  SaveError(errptr, db->rep->Put(options->rep, Slice(key, keylen),
+                                 Slice(ts, tslen), Slice(val, vallen)));
+}
+
+void rocksdb_put_cf_with_ts(rocksdb_t* db,
+                            const rocksdb_writeoptions_t* options,
+                            rocksdb_column_family_handle_t* column_family,
+                            const char* key, size_t keylen, const char* ts,
+                            size_t tslen, const char* val, size_t vallen,
+                            char** errptr) {
+  SaveError(errptr,
+            db->rep->Put(options->rep, column_family->rep, Slice(key, keylen),
+                         Slice(ts, tslen), Slice(val, vallen)));
+}
+
+void rocksdb_delete_with_ts(rocksdb_t* db,
+                            const rocksdb_writeoptions_t* options,
+                            const char* key, size_t keylen, const char* ts,
+                            size_t tslen, char** errptr) {
+  SaveError(errptr, db->rep->Delete(options->rep, Slice(key, keylen),
+                                    Slice(ts, tslen)));
+}
+
+void rocksdb_delete_cf_with_ts(rocksdb_t* db,
+                               const rocksdb_writeoptions_t* options,
+                               rocksdb_column_family_handle_t* column_family,
+                               const char* key, size_t keylen, const char* ts,
+                               size_t tslen, char** errptr) {
+  SaveError(errptr, db->rep->Delete(options->rep, column_family->rep,
+                                    Slice(key, keylen), Slice(ts, tslen)));
+}
+
+void rocksdb_singledelete(rocksdb_t* db, const rocksdb_writeoptions_t* options,
+                          const char* key, size_t keylen, char** errptr) {
+  SaveError(errptr, db->rep->SingleDelete(options->rep, Slice(key, keylen)));
+}
+
+void rocksdb_singledelete_cf(rocksdb_t* db,
+                             const rocksdb_writeoptions_t* options,
+                             rocksdb_column_family_handle_t* column_family,
+                             const char* key, size_t keylen, char** errptr) {
+  SaveError(errptr, db->rep->SingleDelete(options->rep, column_family->rep,
+                                          Slice(key, keylen)));
+}
+
+void rocksdb_singledelete_with_ts(rocksdb_t* db,
+                                  const rocksdb_writeoptions_t* options,
+                                  const char* key, size_t keylen,
+                                  const char* ts, size_t tslen, char** errptr) {
+  SaveError(errptr, db->rep->SingleDelete(options->rep, Slice(key, keylen),
+                                          Slice(ts, tslen)));
+}
+
+void rocksdb_singledelete_cf_with_ts(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, const char* ts, size_t tslen, char** errptr) {
+  SaveError(errptr,
+            db->rep->SingleDelete(options->rep, column_family->rep,
+                                  Slice(key, keylen), Slice(ts, tslen)));
+}
+
+void rocksdb_delete_range_cf(rocksdb_t* db,
+                             const rocksdb_writeoptions_t* options,
+                             rocksdb_column_family_handle_t* column_family,
+                             const char* start_key, size_t start_key_len,
+                             const char* end_key, size_t end_key_len,
+                             char** errptr) {
+  SaveError(errptr, db->rep->DeleteRange(options->rep, column_family->rep,
+                                         Slice(start_key, start_key_len),
+                                         Slice(end_key, end_key_len)));
+}
+
+void rocksdb_flush(rocksdb_t* db, const rocksdb_flushoptions_t* options,
+                   char** errptr) {
+  SaveError(errptr, db->rep->Flush(options->rep));
+}
+
+void rocksdb_flush_cf(rocksdb_t* db, const rocksdb_flushoptions_t* options,
+                      rocksdb_column_family_handle_t* column_family,
+                      char** errptr) {
+  SaveError(errptr, db->rep->Flush(options->rep, column_family->rep));
+}
+
+void rocksdb_flush_wal(rocksdb_t* db, unsigned char sync, char** errptr) {
+  SaveError(errptr, db->rep->FlushWAL(sync));
+}
+
+void rocksdb_pause_background_work(rocksdb_t* db, char** errptr) {
+  SaveError(errptr, db->rep->PauseBackgroundWork());
+}
+
+void rocksdb_continue_background_work(rocksdb_t* db, char** errptr) {
+  SaveError(errptr, db->rep->ContinueBackgroundWork());
+}
+
+void rocksdb_disable_file_deletions(rocksdb_t* db, char** errptr) {
+  SaveError(errptr, db->rep->DisableFileDeletions());
+}
+
+void rocksdb_enable_file_deletions(rocksdb_t* db, char** errptr) {
+  SaveError(errptr, db->rep->EnableFileDeletions());
+}
+
+void rocksdb_verify_checksum(rocksdb_t* db, char** errptr) {
+  SaveError(errptr, db->rep->VerifyChecksum());
+}
+
+void rocksdb_verify_file_checksums(rocksdb_t* db, char** errptr) {
+  SaveError(errptr, db->rep->VerifyFileChecksums(ReadOptions()));
+}
+
+void rocksdb_destroy_db(const rocksdb_options_t* options, const char* name,
+                        char** errptr) {
+  SaveError(errptr, DestroyDB(name, options->rep));
+}
+
+void rocksdb_repair_db(const rocksdb_options_t* options, const char* name,
+                       char** errptr) {
+  SaveError(errptr, RepairDB(name, options->rep));
+}

c_api_gen/c_generated_db_simple_subset.h.inc

@@ -0,0 +1,126 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/generate_c_api.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Source: tools/c_api_gen/spec.json
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/generate_c_api.py
+//     --spec tools/c_api_gen/spec.json --section 'DB data operations'
+//     --header-out
+//     c_api_gen/c_generated_db_simple_subset.h.inc
+//     --source-out
+//     c_api_gen/c_generated_db_simple_subset.cc.inc
+
+/* DB data operations */
+
+extern ROCKSDB_LIBRARY_API void rocksdb_put(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options, const char* key,
+    size_t keylen, const char* val, size_t vallen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_put_cf(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, const char* val, size_t vallen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_delete(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options, const char* key,
+    size_t keylen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_delete_cf(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_merge(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options, const char* key,
+    size_t keylen, const char* val, size_t vallen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_merge_cf(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, const char* val, size_t vallen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_write(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_writebatch_t* batch, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_put_with_ts(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options, const char* key,
+    size_t keylen, const char* ts, size_t tslen, const char* val, size_t vallen,
+    char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_put_cf_with_ts(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, const char* ts, size_t tslen, const char* val, size_t vallen,
+    char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_delete_with_ts(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options, const char* key,
+    size_t keylen, const char* ts, size_t tslen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_delete_cf_with_ts(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, const char* ts, size_t tslen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_singledelete(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options, const char* key,
+    size_t keylen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_singledelete_cf(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_singledelete_with_ts(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options, const char* key,
+    size_t keylen, const char* ts, size_t tslen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_singledelete_cf_with_ts(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, const char* ts, size_t tslen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_delete_range_cf(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* start_key,
+    size_t start_key_len, const char* end_key, size_t end_key_len,
+    char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_flush(
+    rocksdb_t* db, const rocksdb_flushoptions_t* options, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_flush_cf(
+    rocksdb_t* db, const rocksdb_flushoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_flush_wal(rocksdb_t* db,
+                                                  unsigned char sync,
+                                                  char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_pause_background_work(rocksdb_t* db,
+                                                              char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_continue_background_work(rocksdb_t* db,
+                                                                 char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_disable_file_deletions(rocksdb_t* db,
+                                                               char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_enable_file_deletions(rocksdb_t* db,
+                                                              char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_verify_checksum(rocksdb_t* db,
+                                                        char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_verify_file_checksums(rocksdb_t* db,
+                                                              char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_destroy_db(
+    const rocksdb_options_t* options, const char* name, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_repair_db(
+    const rocksdb_options_t* options, const char* name, char** errptr);

c_api_gen/c_generated_jobinfo_auto.cc.inc

@@ -0,0 +1,255 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/auto_simple_bindings.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Sources:
+//   - include/rocksdb/listener.h
+//   - tools/c_api_gen/auto_simple_bindings_blocklist.json
+//   - include/rocksdb/c.h
+//   - tools/c_api_gen/c_base.cc
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/auto_simple_bindings.py
+
+// Output group: Auto-discovered JobInfo metadata simple
+
+/* FlushJobInfo */
+
+uint32_t rocksdb_flushjobinfo_cf_id(const rocksdb_flushjobinfo_t* info) {
+  return info->rep.cf_id;
+}
+
+const char* rocksdb_flushjobinfo_cf_name(const rocksdb_flushjobinfo_t* info,
+                                         size_t* size) {
+  *size = info->rep.cf_name.size();
+  return info->rep.cf_name.data();
+}
+
+const char* rocksdb_flushjobinfo_file_path(const rocksdb_flushjobinfo_t* info,
+                                           size_t* size) {
+  *size = info->rep.file_path.size();
+  return info->rep.file_path.data();
+}
+
+uint64_t rocksdb_flushjobinfo_file_number(const rocksdb_flushjobinfo_t* info) {
+  return info->rep.file_number;
+}
+
+uint64_t rocksdb_flushjobinfo_oldest_blob_file_number(
+    const rocksdb_flushjobinfo_t* info) {
+  return info->rep.oldest_blob_file_number;
+}
+
+uint64_t rocksdb_flushjobinfo_thread_id(const rocksdb_flushjobinfo_t* info) {
+  return info->rep.thread_id;
+}
+
+int rocksdb_flushjobinfo_job_id(const rocksdb_flushjobinfo_t* info) {
+  return info->rep.job_id;
+}
+
+unsigned char rocksdb_flushjobinfo_triggered_writes_slowdown(
+    const rocksdb_flushjobinfo_t* info) {
+  return info->rep.triggered_writes_slowdown;
+}
+
+unsigned char rocksdb_flushjobinfo_triggered_writes_stop(
+    const rocksdb_flushjobinfo_t* info) {
+  return info->rep.triggered_writes_stop;
+}
+
+uint64_t rocksdb_flushjobinfo_smallest_seqno(
+    const rocksdb_flushjobinfo_t* info) {
+  return info->rep.smallest_seqno;
+}
+
+uint64_t rocksdb_flushjobinfo_largest_seqno(
+    const rocksdb_flushjobinfo_t* info) {
+  return info->rep.largest_seqno;
+}
+
+uint32_t rocksdb_flushjobinfo_flush_reason(const rocksdb_flushjobinfo_t* info) {
+  return static_cast<uint32_t>(info->rep.flush_reason);
+}
+
+uint32_t rocksdb_flushjobinfo_blob_compression_type(
+    const rocksdb_flushjobinfo_t* info) {
+  return static_cast<uint32_t>(info->rep.blob_compression_type);
+}
+
+/* CompactionJobInfo */
+
+uint32_t rocksdb_compactionjobinfo_cf_id(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.cf_id;
+}
+
+const char* rocksdb_compactionjobinfo_cf_name(
+    const rocksdb_compactionjobinfo_t* info, size_t* size) {
+  *size = info->rep.cf_name.size();
+  return info->rep.cf_name.data();
+}
+
+void rocksdb_compactionjobinfo_status(const rocksdb_compactionjobinfo_t* info,
+                                      char** errptr) {
+  SaveError(errptr, info->rep.status);
+}
+
+uint64_t rocksdb_compactionjobinfo_thread_id(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.thread_id;
+}
+
+int rocksdb_compactionjobinfo_job_id(const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.job_id;
+}
+
+int rocksdb_compactionjobinfo_num_l0_files(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.num_l0_files;
+}
+
+int rocksdb_compactionjobinfo_base_input_level(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.base_input_level;
+}
+
+int rocksdb_compactionjobinfo_output_level(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.output_level;
+}
+
+uint32_t rocksdb_compactionjobinfo_compaction_reason(
+    const rocksdb_compactionjobinfo_t* info) {
+  return static_cast<uint32_t>(info->rep.compaction_reason);
+}
+
+uint32_t rocksdb_compactionjobinfo_compression(
+    const rocksdb_compactionjobinfo_t* info) {
+  return static_cast<uint32_t>(info->rep.compression);
+}
+
+uint32_t rocksdb_compactionjobinfo_blob_compression_type(
+    const rocksdb_compactionjobinfo_t* info) {
+  return static_cast<uint32_t>(info->rep.blob_compression_type);
+}
+
+unsigned char rocksdb_compactionjobinfo_aborted(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.aborted;
+}
+
+/* SubcompactionJobInfo */
+
+uint32_t rocksdb_subcompactionjobinfo_cf_id(
+    const rocksdb_subcompactionjobinfo_t* info) {
+  return info->rep.cf_id;
+}
+
+const char* rocksdb_subcompactionjobinfo_cf_name(
+    const rocksdb_subcompactionjobinfo_t* info, size_t* size) {
+  *size = info->rep.cf_name.size();
+  return info->rep.cf_name.data();
+}
+
+void rocksdb_subcompactionjobinfo_status(
+    const rocksdb_subcompactionjobinfo_t* info, char** errptr) {
+  SaveError(errptr, info->rep.status);
+}
+
+uint64_t rocksdb_subcompactionjobinfo_thread_id(
+    const rocksdb_subcompactionjobinfo_t* info) {
+  return info->rep.thread_id;
+}
+
+int rocksdb_subcompactionjobinfo_job_id(
+    const rocksdb_subcompactionjobinfo_t* info) {
+  return info->rep.job_id;
+}
+
+int rocksdb_subcompactionjobinfo_subcompaction_job_id(
+    const rocksdb_subcompactionjobinfo_t* info) {
+  return info->rep.subcompaction_job_id;
+}
+
+int rocksdb_subcompactionjobinfo_base_input_level(
+    const rocksdb_subcompactionjobinfo_t* info) {
+  return info->rep.base_input_level;
+}
+
+int rocksdb_subcompactionjobinfo_output_level(
+    const rocksdb_subcompactionjobinfo_t* info) {
+  return info->rep.output_level;
+}
+
+uint32_t rocksdb_subcompactionjobinfo_compaction_reason(
+    const rocksdb_subcompactionjobinfo_t* info) {
+  return static_cast<uint32_t>(info->rep.compaction_reason);
+}
+
+uint32_t rocksdb_subcompactionjobinfo_compression(
+    const rocksdb_subcompactionjobinfo_t* info) {
+  return static_cast<uint32_t>(info->rep.compression);
+}
+
+uint32_t rocksdb_subcompactionjobinfo_blob_compression_type(
+    const rocksdb_subcompactionjobinfo_t* info) {
+  return static_cast<uint32_t>(info->rep.blob_compression_type);
+}
+
+/* ExternalFileIngestionInfo */
+
+const char* rocksdb_externalfileingestioninfo_cf_name(
+    const rocksdb_externalfileingestioninfo_t* info, size_t* size) {
+  *size = info->rep.cf_name.size();
+  return info->rep.cf_name.data();
+}
+
+const char* rocksdb_externalfileingestioninfo_external_file_path(
+    const rocksdb_externalfileingestioninfo_t* info, size_t* size) {
+  *size = info->rep.external_file_path.size();
+  return info->rep.external_file_path.data();
+}
+
+const char* rocksdb_externalfileingestioninfo_internal_file_path(
+    const rocksdb_externalfileingestioninfo_t* info, size_t* size) {
+  *size = info->rep.internal_file_path.size();
+  return info->rep.internal_file_path.data();
+}
+
+uint64_t rocksdb_externalfileingestioninfo_global_seqno(
+    const rocksdb_externalfileingestioninfo_t* info) {
+  return info->rep.global_seqno;
+}
+
+/* MemTableInfo */
+
+const char* rocksdb_memtableinfo_cf_name(const rocksdb_memtableinfo_t* info,
+                                         size_t* size) {
+  *size = info->rep.cf_name.size();
+  return info->rep.cf_name.data();
+}
+
+uint64_t rocksdb_memtableinfo_first_seqno(const rocksdb_memtableinfo_t* info) {
+  return info->rep.first_seqno;
+}
+
+uint64_t rocksdb_memtableinfo_earliest_seqno(
+    const rocksdb_memtableinfo_t* info) {
+  return info->rep.earliest_seqno;
+}
+
+uint64_t rocksdb_memtableinfo_num_entries(const rocksdb_memtableinfo_t* info) {
+  return info->rep.num_entries;
+}
+
+uint64_t rocksdb_memtableinfo_num_deletes(const rocksdb_memtableinfo_t* info) {
+  return info->rep.num_deletes;
+}
+
+const char* rocksdb_memtableinfo_newest_udt(const rocksdb_memtableinfo_t* info,
+                                            size_t* size) {
+  *size = info->rep.newest_udt.size();
+  return info->rep.newest_udt.data();
+}

c_api_gen/c_generated_jobinfo_auto.h.inc

@@ -0,0 +1,173 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/auto_simple_bindings.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Sources:
+//   - include/rocksdb/listener.h
+//   - tools/c_api_gen/auto_simple_bindings_blocklist.json
+//   - include/rocksdb/c.h
+//   - tools/c_api_gen/c_base.cc
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/auto_simple_bindings.py
+
+// Output group: Auto-discovered JobInfo metadata simple
+
+/* FlushJobInfo */
+
+extern ROCKSDB_LIBRARY_API uint32_t
+rocksdb_flushjobinfo_cf_id(const rocksdb_flushjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API const char* rocksdb_flushjobinfo_cf_name(
+    const rocksdb_flushjobinfo_t* info, size_t* size);
+
+extern ROCKSDB_LIBRARY_API const char* rocksdb_flushjobinfo_file_path(
+    const rocksdb_flushjobinfo_t* info, size_t* size);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_flushjobinfo_file_number(const rocksdb_flushjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_flushjobinfo_oldest_blob_file_number(
+    const rocksdb_flushjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_flushjobinfo_thread_id(const rocksdb_flushjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API int rocksdb_flushjobinfo_job_id(
+    const rocksdb_flushjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_flushjobinfo_triggered_writes_slowdown(
+    const rocksdb_flushjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_flushjobinfo_triggered_writes_stop(const rocksdb_flushjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_flushjobinfo_smallest_seqno(const rocksdb_flushjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_flushjobinfo_largest_seqno(const rocksdb_flushjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint32_t
+rocksdb_flushjobinfo_flush_reason(const rocksdb_flushjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint32_t
+rocksdb_flushjobinfo_blob_compression_type(const rocksdb_flushjobinfo_t* info);
+
+/* CompactionJobInfo */
+
+extern ROCKSDB_LIBRARY_API uint32_t
+rocksdb_compactionjobinfo_cf_id(const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API const char* rocksdb_compactionjobinfo_cf_name(
+    const rocksdb_compactionjobinfo_t* info, size_t* size);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_compactionjobinfo_status(
+    const rocksdb_compactionjobinfo_t* info, char** errptr);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compactionjobinfo_thread_id(const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API int rocksdb_compactionjobinfo_job_id(
+    const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API int rocksdb_compactionjobinfo_num_l0_files(
+    const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API int rocksdb_compactionjobinfo_base_input_level(
+    const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API int rocksdb_compactionjobinfo_output_level(
+    const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint32_t rocksdb_compactionjobinfo_compaction_reason(
+    const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint32_t
+rocksdb_compactionjobinfo_compression(const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint32_t
+rocksdb_compactionjobinfo_blob_compression_type(
+    const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API unsigned char rocksdb_compactionjobinfo_aborted(
+    const rocksdb_compactionjobinfo_t* info);
+
+/* SubcompactionJobInfo */
+
+extern ROCKSDB_LIBRARY_API uint32_t
+rocksdb_subcompactionjobinfo_cf_id(const rocksdb_subcompactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API const char* rocksdb_subcompactionjobinfo_cf_name(
+    const rocksdb_subcompactionjobinfo_t* info, size_t* size);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_subcompactionjobinfo_status(
+    const rocksdb_subcompactionjobinfo_t* info, char** errptr);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_subcompactionjobinfo_thread_id(
+    const rocksdb_subcompactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API int rocksdb_subcompactionjobinfo_job_id(
+    const rocksdb_subcompactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API int
+rocksdb_subcompactionjobinfo_subcompaction_job_id(
+    const rocksdb_subcompactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API int rocksdb_subcompactionjobinfo_base_input_level(
+    const rocksdb_subcompactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API int rocksdb_subcompactionjobinfo_output_level(
+    const rocksdb_subcompactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint32_t
+rocksdb_subcompactionjobinfo_compaction_reason(
+    const rocksdb_subcompactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint32_t rocksdb_subcompactionjobinfo_compression(
+    const rocksdb_subcompactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint32_t
+rocksdb_subcompactionjobinfo_blob_compression_type(
+    const rocksdb_subcompactionjobinfo_t* info);
+
+/* ExternalFileIngestionInfo */
+
+extern ROCKSDB_LIBRARY_API const char*
+rocksdb_externalfileingestioninfo_cf_name(
+    const rocksdb_externalfileingestioninfo_t* info, size_t* size);
+
+extern ROCKSDB_LIBRARY_API const char*
+rocksdb_externalfileingestioninfo_external_file_path(
+    const rocksdb_externalfileingestioninfo_t* info, size_t* size);
+
+extern ROCKSDB_LIBRARY_API const char*
+rocksdb_externalfileingestioninfo_internal_file_path(
+    const rocksdb_externalfileingestioninfo_t* info, size_t* size);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_externalfileingestioninfo_global_seqno(
+    const rocksdb_externalfileingestioninfo_t* info);
+
+/* MemTableInfo */
+
+extern ROCKSDB_LIBRARY_API const char* rocksdb_memtableinfo_cf_name(
+    const rocksdb_memtableinfo_t* info, size_t* size);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_memtableinfo_first_seqno(const rocksdb_memtableinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_memtableinfo_earliest_seqno(const rocksdb_memtableinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_memtableinfo_num_entries(const rocksdb_memtableinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_memtableinfo_num_deletes(const rocksdb_memtableinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API const char* rocksdb_memtableinfo_newest_udt(
+    const rocksdb_memtableinfo_t* info, size_t* size);

c_api_gen/c_generated_jobinfo_metadata_subset.cc.inc

@@ -0,0 +1,84 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/generate_c_api.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Source: tools/c_api_gen/spec.json
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/generate_c_api.py
+//     --spec tools/c_api_gen/spec.json --section 'JobInfo metadata simple'
+//     --header-out
+//     c_api_gen/c_generated_jobinfo_metadata_subset.h.inc
+//     --source-out
+//     c_api_gen/c_generated_jobinfo_metadata_subset.cc.inc
+
+/* JobInfo metadata simple */
+
+size_t rocksdb_compactionjobinfo_input_files_count(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.input_files.size();
+}
+
+size_t rocksdb_compactionjobinfo_output_files_count(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.output_files.size();
+}
+
+uint64_t rocksdb_compactionjobinfo_elapsed_micros(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.stats.elapsed_micros;
+}
+
+uint64_t rocksdb_compactionjobinfo_num_corrupt_keys(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.stats.num_corrupt_keys;
+}
+
+uint64_t rocksdb_compactionjobinfo_input_records(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.stats.num_input_records;
+}
+
+uint64_t rocksdb_compactionjobinfo_output_records(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.stats.num_output_records;
+}
+
+uint64_t rocksdb_compactionjobinfo_total_input_bytes(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.stats.total_input_bytes;
+}
+
+uint64_t rocksdb_compactionjobinfo_total_output_bytes(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.stats.total_output_bytes;
+}
+
+size_t rocksdb_compactionjobinfo_num_input_files(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.stats.num_input_files;
+}
+
+size_t rocksdb_compactionjobinfo_num_input_files_at_output_level(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.stats.num_input_files_at_output_level;
+}
+
+const char* rocksdb_writestallinfo_cf_name(const rocksdb_writestallinfo_t* info,
+                                           size_t* size) {
+  *size = info->rep.cf_name.size();
+  return info->rep.cf_name.data();
+}
+
+const rocksdb_writestallcondition_t* rocksdb_writestallinfo_cur(
+    const rocksdb_writestallinfo_t* info) {
+  return reinterpret_cast<const rocksdb_writestallcondition_t*>(
+      &info->rep.condition.cur);
+}
+
+const rocksdb_writestallcondition_t* rocksdb_writestallinfo_prev(
+    const rocksdb_writestallinfo_t* info) {
+  return reinterpret_cast<const rocksdb_writestallcondition_t*>(
+      &info->rep.condition.prev);
+}

c_api_gen/c_generated_jobinfo_metadata_subset.h.inc

@@ -0,0 +1,57 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/generate_c_api.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Source: tools/c_api_gen/spec.json
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/generate_c_api.py
+//     --spec tools/c_api_gen/spec.json --section 'JobInfo metadata simple'
+//     --header-out
+//     c_api_gen/c_generated_jobinfo_metadata_subset.h.inc
+//     --source-out
+//     c_api_gen/c_generated_jobinfo_metadata_subset.cc.inc
+
+/* JobInfo metadata simple */
+
+extern ROCKSDB_LIBRARY_API size_t rocksdb_compactionjobinfo_input_files_count(
+    const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API size_t rocksdb_compactionjobinfo_output_files_count(
+    const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_compactionjobinfo_elapsed_micros(
+    const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_compactionjobinfo_num_corrupt_keys(
+    const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_compactionjobinfo_input_records(
+    const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_compactionjobinfo_output_records(
+    const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_compactionjobinfo_total_input_bytes(
+    const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compactionjobinfo_total_output_bytes(
+    const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API size_t rocksdb_compactionjobinfo_num_input_files(
+    const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API size_t
+rocksdb_compactionjobinfo_num_input_files_at_output_level(
+    const rocksdb_compactionjobinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API const char* rocksdb_writestallinfo_cf_name(
+    const rocksdb_writestallinfo_t* info, size_t* size);
+
+extern ROCKSDB_LIBRARY_API const rocksdb_writestallcondition_t*
+rocksdb_writestallinfo_cur(const rocksdb_writestallinfo_t* info);
+
+extern ROCKSDB_LIBRARY_API const rocksdb_writestallcondition_t*
+rocksdb_writestallinfo_prev(const rocksdb_writestallinfo_t* info);

c_api_gen/c_generated_metadata_structs_auto.cc.inc

@@ -0,0 +1,501 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/auto_simple_bindings.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Sources:
+//   - include/rocksdb/compaction_job_stats.h
+//   - include/rocksdb/listener.h
+//   - include/rocksdb/table_properties.h
+//   - tools/c_api_gen/auto_simple_bindings_blocklist.json
+//   - include/rocksdb/c.h
+//   - tools/c_api_gen/c_base.cc
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/auto_simple_bindings.py
+
+// Output group: Auto-discovered metadata view structs simple
+
+/* TableProperties */
+
+uint64_t rocksdb_table_properties_orig_file_number(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.orig_file_number;
+}
+
+uint64_t rocksdb_table_properties_data_size(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.data_size;
+}
+
+uint64_t rocksdb_table_properties_uncompressed_data_size(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.uncompressed_data_size;
+}
+
+uint64_t rocksdb_table_properties_index_size(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.index_size;
+}
+
+uint64_t rocksdb_table_properties_index_partitions(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.index_partitions;
+}
+
+uint64_t rocksdb_table_properties_top_level_index_size(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.top_level_index_size;
+}
+
+uint64_t rocksdb_table_properties_index_key_is_user_key(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.index_key_is_user_key;
+}
+
+uint64_t rocksdb_table_properties_index_value_is_delta_encoded(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.index_value_is_delta_encoded;
+}
+
+uint64_t rocksdb_table_properties_udi_is_primary_index(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.udi_is_primary_index;
+}
+
+uint64_t rocksdb_table_properties_filter_size(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.filter_size;
+}
+
+uint64_t rocksdb_table_properties_raw_key_size(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.raw_key_size;
+}
+
+uint64_t rocksdb_table_properties_raw_value_size(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.raw_value_size;
+}
+
+uint64_t rocksdb_table_properties_num_data_blocks(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.num_data_blocks;
+}
+
+uint64_t rocksdb_table_properties_num_data_blocks_compression_rejected(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.num_data_blocks_compression_rejected;
+}
+
+uint64_t rocksdb_table_properties_num_data_blocks_compression_bypassed(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.num_data_blocks_compression_bypassed;
+}
+
+uint64_t rocksdb_table_properties_num_uniform_blocks(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.num_uniform_blocks;
+}
+
+uint64_t rocksdb_table_properties_num_entries(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.num_entries;
+}
+
+uint64_t rocksdb_table_properties_num_filter_entries(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.num_filter_entries;
+}
+
+uint64_t rocksdb_table_properties_num_deletions(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.num_deletions;
+}
+
+uint64_t rocksdb_table_properties_num_merge_operands(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.num_merge_operands;
+}
+
+uint64_t rocksdb_table_properties_num_range_deletions(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.num_range_deletions;
+}
+
+uint64_t rocksdb_table_properties_format_version(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.format_version;
+}
+
+uint64_t rocksdb_table_properties_fixed_key_len(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.fixed_key_len;
+}
+
+uint64_t rocksdb_table_properties_column_family_id(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.column_family_id;
+}
+
+uint64_t rocksdb_table_properties_creation_time(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.creation_time;
+}
+
+uint64_t rocksdb_table_properties_oldest_key_time(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.oldest_key_time;
+}
+
+uint64_t rocksdb_table_properties_newest_key_time(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.newest_key_time;
+}
+
+uint64_t rocksdb_table_properties_file_creation_time(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.file_creation_time;
+}
+
+uint64_t rocksdb_table_properties_slow_compression_estimated_data_size(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.slow_compression_estimated_data_size;
+}
+
+uint64_t rocksdb_table_properties_fast_compression_estimated_data_size(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.fast_compression_estimated_data_size;
+}
+
+uint64_t rocksdb_table_properties_external_sst_file_global_seqno_offset(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.external_sst_file_global_seqno_offset;
+}
+
+uint64_t rocksdb_table_properties_tail_start_offset(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.tail_start_offset;
+}
+
+uint64_t rocksdb_table_properties_user_defined_timestamps_persisted(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.user_defined_timestamps_persisted;
+}
+
+uint64_t rocksdb_table_properties_key_largest_seqno(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.key_largest_seqno;
+}
+
+uint64_t rocksdb_table_properties_key_smallest_seqno(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.key_smallest_seqno;
+}
+
+uint64_t rocksdb_table_properties_data_block_restart_interval(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.data_block_restart_interval;
+}
+
+uint64_t rocksdb_table_properties_index_block_restart_interval(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.index_block_restart_interval;
+}
+
+uint64_t rocksdb_table_properties_separate_key_value_in_data_block(
+    const rocksdb_table_properties_t* props) {
+  return props->rep.separate_key_value_in_data_block;
+}
+
+const char* rocksdb_table_properties_db_id(
+    const rocksdb_table_properties_t* props, size_t* size) {
+  *size = props->rep.db_id.size();
+  return props->rep.db_id.data();
+}
+
+const char* rocksdb_table_properties_db_session_id(
+    const rocksdb_table_properties_t* props, size_t* size) {
+  *size = props->rep.db_session_id.size();
+  return props->rep.db_session_id.data();
+}
+
+const char* rocksdb_table_properties_db_host_id(
+    const rocksdb_table_properties_t* props, size_t* size) {
+  *size = props->rep.db_host_id.size();
+  return props->rep.db_host_id.data();
+}
+
+const char* rocksdb_table_properties_column_family_name(
+    const rocksdb_table_properties_t* props, size_t* size) {
+  *size = props->rep.column_family_name.size();
+  return props->rep.column_family_name.data();
+}
+
+const char* rocksdb_table_properties_filter_policy_name(
+    const rocksdb_table_properties_t* props, size_t* size) {
+  *size = props->rep.filter_policy_name.size();
+  return props->rep.filter_policy_name.data();
+}
+
+const char* rocksdb_table_properties_comparator_name(
+    const rocksdb_table_properties_t* props, size_t* size) {
+  *size = props->rep.comparator_name.size();
+  return props->rep.comparator_name.data();
+}
+
+const char* rocksdb_table_properties_merge_operator_name(
+    const rocksdb_table_properties_t* props, size_t* size) {
+  *size = props->rep.merge_operator_name.size();
+  return props->rep.merge_operator_name.data();
+}
+
+const char* rocksdb_table_properties_prefix_extractor_name(
+    const rocksdb_table_properties_t* props, size_t* size) {
+  *size = props->rep.prefix_extractor_name.size();
+  return props->rep.prefix_extractor_name.data();
+}
+
+const char* rocksdb_table_properties_property_collectors_names(
+    const rocksdb_table_properties_t* props, size_t* size) {
+  *size = props->rep.property_collectors_names.size();
+  return props->rep.property_collectors_names.data();
+}
+
+const char* rocksdb_table_properties_compression_name(
+    const rocksdb_table_properties_t* props, size_t* size) {
+  *size = props->rep.compression_name.size();
+  return props->rep.compression_name.data();
+}
+
+const char* rocksdb_table_properties_compression_options(
+    const rocksdb_table_properties_t* props, size_t* size) {
+  *size = props->rep.compression_options.size();
+  return props->rep.compression_options.data();
+}
+
+const char* rocksdb_table_properties_seqno_to_time_mapping(
+    const rocksdb_table_properties_t* props, size_t* size) {
+  *size = props->rep.seqno_to_time_mapping.size();
+  return props->rep.seqno_to_time_mapping.data();
+}
+
+/* CompactionJobStats */
+
+uint64_t rocksdb_compaction_job_stats_elapsed_micros(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.elapsed_micros;
+}
+
+uint64_t rocksdb_compaction_job_stats_cpu_micros(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.cpu_micros;
+}
+
+unsigned char rocksdb_compaction_job_stats_has_accurate_num_input_records(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.has_accurate_num_input_records;
+}
+
+uint64_t rocksdb_compaction_job_stats_num_input_records(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.num_input_records;
+}
+
+uint64_t rocksdb_compaction_job_stats_num_blobs_read(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.num_blobs_read;
+}
+
+size_t rocksdb_compaction_job_stats_num_input_files(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.num_input_files;
+}
+
+size_t rocksdb_compaction_job_stats_num_input_files_trivially_moved(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.num_input_files_trivially_moved;
+}
+
+size_t rocksdb_compaction_job_stats_num_input_files_at_output_level(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.num_input_files_at_output_level;
+}
+
+size_t rocksdb_compaction_job_stats_num_filtered_input_files(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.num_filtered_input_files;
+}
+
+size_t rocksdb_compaction_job_stats_num_filtered_input_files_at_output_level(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.num_filtered_input_files_at_output_level;
+}
+
+uint64_t rocksdb_compaction_job_stats_num_output_records(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.num_output_records;
+}
+
+size_t rocksdb_compaction_job_stats_num_output_files(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.num_output_files;
+}
+
+size_t rocksdb_compaction_job_stats_num_output_files_blob(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.num_output_files_blob;
+}
+
+unsigned char rocksdb_compaction_job_stats_is_full_compaction(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.is_full_compaction;
+}
+
+unsigned char rocksdb_compaction_job_stats_is_manual_compaction(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.is_manual_compaction;
+}
+
+unsigned char rocksdb_compaction_job_stats_is_remote_compaction(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.is_remote_compaction;
+}
+
+uint64_t rocksdb_compaction_job_stats_total_input_bytes(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.total_input_bytes;
+}
+
+uint64_t rocksdb_compaction_job_stats_total_blob_bytes_read(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.total_blob_bytes_read;
+}
+
+uint64_t rocksdb_compaction_job_stats_total_output_bytes(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.total_output_bytes;
+}
+
+uint64_t rocksdb_compaction_job_stats_total_output_bytes_blob(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.total_output_bytes_blob;
+}
+
+uint64_t rocksdb_compaction_job_stats_total_skipped_input_bytes(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.total_skipped_input_bytes;
+}
+
+uint64_t rocksdb_compaction_job_stats_num_records_replaced(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.num_records_replaced;
+}
+
+uint64_t rocksdb_compaction_job_stats_total_input_raw_key_bytes(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.total_input_raw_key_bytes;
+}
+
+uint64_t rocksdb_compaction_job_stats_total_input_raw_value_bytes(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.total_input_raw_value_bytes;
+}
+
+uint64_t rocksdb_compaction_job_stats_num_input_deletion_records(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.num_input_deletion_records;
+}
+
+uint64_t rocksdb_compaction_job_stats_num_expired_deletion_records(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.num_expired_deletion_records;
+}
+
+uint64_t rocksdb_compaction_job_stats_num_corrupt_keys(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.num_corrupt_keys;
+}
+
+uint64_t rocksdb_compaction_job_stats_file_write_nanos(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.file_write_nanos;
+}
+
+uint64_t rocksdb_compaction_job_stats_file_range_sync_nanos(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.file_range_sync_nanos;
+}
+
+uint64_t rocksdb_compaction_job_stats_file_fsync_nanos(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.file_fsync_nanos;
+}
+
+uint64_t rocksdb_compaction_job_stats_file_prepare_write_nanos(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.file_prepare_write_nanos;
+}
+
+const char* rocksdb_compaction_job_stats_smallest_output_key_prefix(
+    const rocksdb_compaction_job_stats_t* stats, size_t* size) {
+  *size = stats->rep.smallest_output_key_prefix.size();
+  return stats->rep.smallest_output_key_prefix.data();
+}
+
+const char* rocksdb_compaction_job_stats_largest_output_key_prefix(
+    const rocksdb_compaction_job_stats_t* stats, size_t* size) {
+  *size = stats->rep.largest_output_key_prefix.size();
+  return stats->rep.largest_output_key_prefix.data();
+}
+
+uint64_t rocksdb_compaction_job_stats_num_single_del_fallthru(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.num_single_del_fallthru;
+}
+
+uint64_t rocksdb_compaction_job_stats_num_single_del_mismatch(
+    const rocksdb_compaction_job_stats_t* stats) {
+  return stats->rep.num_single_del_mismatch;
+}
+
+/* CompactionFileInfo */
+
+int rocksdb_compaction_file_info_level(
+    const rocksdb_compaction_file_info_t* info) {
+  return info->rep.level;
+}
+
+uint64_t rocksdb_compaction_file_info_file_number(
+    const rocksdb_compaction_file_info_t* info) {
+  return info->rep.file_number;
+}
+
+uint64_t rocksdb_compaction_file_info_oldest_blob_file_number(
+    const rocksdb_compaction_file_info_t* info) {
+  return info->rep.oldest_blob_file_number;
+}
+
+/* BlobFileAdditionInfo */
+
+uint64_t rocksdb_blob_file_addition_info_total_blob_count(
+    const rocksdb_blob_file_addition_info_t* info) {
+  return info->rep.total_blob_count;
+}
+
+uint64_t rocksdb_blob_file_addition_info_total_blob_bytes(
+    const rocksdb_blob_file_addition_info_t* info) {
+  return info->rep.total_blob_bytes;
+}
+
+/* BlobFileGarbageInfo */
+
+uint64_t rocksdb_blob_file_garbage_info_garbage_blob_count(
+    const rocksdb_blob_file_garbage_info_t* info) {
+  return info->rep.garbage_blob_count;
+}
+
+uint64_t rocksdb_blob_file_garbage_info_garbage_blob_bytes(
+    const rocksdb_blob_file_garbage_info_t* info) {
+  return info->rep.garbage_blob_bytes;
+}

c_api_gen/c_generated_metadata_structs_auto.h.inc

@@ -0,0 +1,361 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/auto_simple_bindings.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Sources:
+//   - include/rocksdb/compaction_job_stats.h
+//   - include/rocksdb/listener.h
+//   - include/rocksdb/table_properties.h
+//   - tools/c_api_gen/auto_simple_bindings_blocklist.json
+//   - include/rocksdb/c.h
+//   - tools/c_api_gen/c_base.cc
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/auto_simple_bindings.py
+
+// Output group: Auto-discovered metadata view structs simple
+
+/* TableProperties */
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_table_properties_orig_file_number(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_data_size(const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_uncompressed_data_size(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_index_size(const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_table_properties_index_partitions(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_top_level_index_size(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_index_key_is_user_key(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_index_value_is_delta_encoded(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_udi_is_primary_index(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_filter_size(const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_raw_key_size(const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_table_properties_raw_value_size(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_table_properties_num_data_blocks(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_num_data_blocks_compression_rejected(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_num_data_blocks_compression_bypassed(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_table_properties_num_uniform_blocks(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_num_entries(const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_table_properties_num_filter_entries(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_num_deletions(const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_table_properties_num_merge_operands(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_num_range_deletions(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_table_properties_format_version(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_fixed_key_len(const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_table_properties_column_family_id(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_creation_time(const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_table_properties_oldest_key_time(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_table_properties_newest_key_time(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_table_properties_file_creation_time(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_slow_compression_estimated_data_size(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_fast_compression_estimated_data_size(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_external_sst_file_global_seqno_offset(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_table_properties_tail_start_offset(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_user_defined_timestamps_persisted(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_table_properties_key_largest_seqno(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_table_properties_key_smallest_seqno(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_data_block_restart_interval(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_index_block_restart_interval(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_table_properties_separate_key_value_in_data_block(
+    const rocksdb_table_properties_t* props);
+
+extern ROCKSDB_LIBRARY_API const char* rocksdb_table_properties_db_id(
+    const rocksdb_table_properties_t* props, size_t* size);
+
+extern ROCKSDB_LIBRARY_API const char* rocksdb_table_properties_db_session_id(
+    const rocksdb_table_properties_t* props, size_t* size);
+
+extern ROCKSDB_LIBRARY_API const char* rocksdb_table_properties_db_host_id(
+    const rocksdb_table_properties_t* props, size_t* size);
+
+extern ROCKSDB_LIBRARY_API const char*
+rocksdb_table_properties_column_family_name(
+    const rocksdb_table_properties_t* props, size_t* size);
+
+extern ROCKSDB_LIBRARY_API const char*
+rocksdb_table_properties_filter_policy_name(
+    const rocksdb_table_properties_t* props, size_t* size);
+
+extern ROCKSDB_LIBRARY_API const char* rocksdb_table_properties_comparator_name(
+    const rocksdb_table_properties_t* props, size_t* size);
+
+extern ROCKSDB_LIBRARY_API const char*
+rocksdb_table_properties_merge_operator_name(
+    const rocksdb_table_properties_t* props, size_t* size);
+
+extern ROCKSDB_LIBRARY_API const char*
+rocksdb_table_properties_prefix_extractor_name(
+    const rocksdb_table_properties_t* props, size_t* size);
+
+extern ROCKSDB_LIBRARY_API const char*
+rocksdb_table_properties_property_collectors_names(
+    const rocksdb_table_properties_t* props, size_t* size);
+
+extern ROCKSDB_LIBRARY_API const char*
+rocksdb_table_properties_compression_name(
+    const rocksdb_table_properties_t* props, size_t* size);
+
+extern ROCKSDB_LIBRARY_API const char*
+rocksdb_table_properties_compression_options(
+    const rocksdb_table_properties_t* props, size_t* size);
+
+extern ROCKSDB_LIBRARY_API const char*
+rocksdb_table_properties_seqno_to_time_mapping(
+    const rocksdb_table_properties_t* props, size_t* size);
+
+/* CompactionJobStats */
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_compaction_job_stats_elapsed_micros(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_compaction_job_stats_cpu_micros(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_compaction_job_stats_has_accurate_num_input_records(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_num_input_records(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_compaction_job_stats_num_blobs_read(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API size_t rocksdb_compaction_job_stats_num_input_files(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API size_t
+rocksdb_compaction_job_stats_num_input_files_trivially_moved(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API size_t
+rocksdb_compaction_job_stats_num_input_files_at_output_level(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API size_t
+rocksdb_compaction_job_stats_num_filtered_input_files(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API size_t
+rocksdb_compaction_job_stats_num_filtered_input_files_at_output_level(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_num_output_records(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API size_t rocksdb_compaction_job_stats_num_output_files(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API size_t
+rocksdb_compaction_job_stats_num_output_files_blob(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_compaction_job_stats_is_full_compaction(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_compaction_job_stats_is_manual_compaction(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_compaction_job_stats_is_remote_compaction(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_total_input_bytes(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_total_blob_bytes_read(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_total_output_bytes(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_total_output_bytes_blob(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_total_skipped_input_bytes(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_num_records_replaced(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_total_input_raw_key_bytes(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_total_input_raw_value_bytes(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_num_input_deletion_records(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_num_expired_deletion_records(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_num_corrupt_keys(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_file_write_nanos(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_file_range_sync_nanos(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_file_fsync_nanos(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_file_prepare_write_nanos(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API const char*
+rocksdb_compaction_job_stats_smallest_output_key_prefix(
+    const rocksdb_compaction_job_stats_t* stats, size_t* size);
+
+extern ROCKSDB_LIBRARY_API const char*
+rocksdb_compaction_job_stats_largest_output_key_prefix(
+    const rocksdb_compaction_job_stats_t* stats, size_t* size);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_num_single_del_fallthru(
+    const rocksdb_compaction_job_stats_t* stats);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_job_stats_num_single_del_mismatch(
+    const rocksdb_compaction_job_stats_t* stats);
+
+/* CompactionFileInfo */
+
+extern ROCKSDB_LIBRARY_API int rocksdb_compaction_file_info_level(
+    const rocksdb_compaction_file_info_t* info);
+
+extern ROCKSDB_LIBRARY_API uint64_t rocksdb_compaction_file_info_file_number(
+    const rocksdb_compaction_file_info_t* info);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_compaction_file_info_oldest_blob_file_number(
+    const rocksdb_compaction_file_info_t* info);
+
+/* BlobFileAdditionInfo */
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_blob_file_addition_info_total_blob_count(
+    const rocksdb_blob_file_addition_info_t* info);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_blob_file_addition_info_total_blob_bytes(
+    const rocksdb_blob_file_addition_info_t* info);
+
+/* BlobFileGarbageInfo */
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_blob_file_garbage_info_garbage_blob_count(
+    const rocksdb_blob_file_garbage_info_t* info);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_blob_file_garbage_info_garbage_blob_bytes(
+    const rocksdb_blob_file_garbage_info_t* info);

c_api_gen/c_generated_options_subset.cc.inc

@@ -0,0 +1,84 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/generate_c_api.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Source: tools/c_api_gen/spec.json
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/generate_c_api.py
+//     --spec tools/c_api_gen/spec.json --section 'Options simple'
+//     --header-out
+//     c_api_gen/c_generated_options_subset.h.inc
+//     --source-out
+//     c_api_gen/c_generated_options_subset.cc.inc
+
+/* Options simple */
+
+void rocksdb_options_set_compression_options_zstd_max_train_bytes(
+    rocksdb_options_t* opt, int zstd_max_train_bytes) {
+  opt->rep.compression_opts.zstd_max_train_bytes = zstd_max_train_bytes;
+}
+
+int rocksdb_options_get_compression_options_zstd_max_train_bytes(
+    rocksdb_options_t* opt) {
+  return opt->rep.compression_opts.zstd_max_train_bytes;
+}
+
+void rocksdb_options_set_compression_options_use_zstd_dict_trainer(
+    rocksdb_options_t* opt, unsigned char use_zstd_dict_trainer) {
+  opt->rep.compression_opts.use_zstd_dict_trainer = use_zstd_dict_trainer;
+}
+
+unsigned char rocksdb_options_get_compression_options_use_zstd_dict_trainer(
+    rocksdb_options_t* opt) {
+  return opt->rep.compression_opts.use_zstd_dict_trainer;
+}
+
+void rocksdb_options_set_compression_options_parallel_threads(
+    rocksdb_options_t* opt, int value) {
+  opt->rep.compression_opts.parallel_threads = value;
+}
+
+int rocksdb_options_get_compression_options_parallel_threads(
+    rocksdb_options_t* opt) {
+  return opt->rep.compression_opts.parallel_threads;
+}
+
+void rocksdb_options_set_compression_options_max_dict_buffer_bytes(
+    rocksdb_options_t* opt, uint64_t max_dict_buffer_bytes) {
+  opt->rep.compression_opts.max_dict_buffer_bytes = max_dict_buffer_bytes;
+}
+
+uint64_t rocksdb_options_get_compression_options_max_dict_buffer_bytes(
+    rocksdb_options_t* opt) {
+  return opt->rep.compression_opts.max_dict_buffer_bytes;
+}
+
+unsigned char
+rocksdb_options_get_bottommost_compression_options_use_zstd_dict_trainer(
+    rocksdb_options_t* opt) {
+  return opt->rep.bottommost_compression_opts.use_zstd_dict_trainer;
+}
+
+void rocksdb_options_set_use_fsync(rocksdb_options_t* opt, int use_fsync) {
+  opt->rep.use_fsync = use_fsync;
+}
+
+int rocksdb_options_get_use_fsync(rocksdb_options_t* opt) {
+  return opt->rep.use_fsync;
+}
+
+void rocksdb_options_set_disable_auto_compactions(rocksdb_options_t* opt,
+                                                  int disable) {
+  opt->rep.disable_auto_compactions = disable;
+}
+
+void rocksdb_options_set_optimize_filters_for_hits(rocksdb_options_t* opt,
+                                                   int v) {
+  opt->rep.optimize_filters_for_hits = v;
+}
+
+void rocksdb_options_set_report_bg_io_stats(rocksdb_options_t* opt, int v) {
+  opt->rep.report_bg_io_stats = v;
+}

c_api_gen/c_generated_options_subset.h.inc

@@ -0,0 +1,67 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/generate_c_api.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Source: tools/c_api_gen/spec.json
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/generate_c_api.py
+//     --spec tools/c_api_gen/spec.json --section 'Options simple'
+//     --header-out
+//     c_api_gen/c_generated_options_subset.h.inc
+//     --source-out
+//     c_api_gen/c_generated_options_subset.cc.inc
+
+/* Options simple */
+
+extern ROCKSDB_LIBRARY_API void
+rocksdb_options_set_compression_options_zstd_max_train_bytes(
+    rocksdb_options_t* opt, int zstd_max_train_bytes);
+
+extern ROCKSDB_LIBRARY_API int
+rocksdb_options_get_compression_options_zstd_max_train_bytes(
+    rocksdb_options_t* opt);
+
+extern ROCKSDB_LIBRARY_API void
+rocksdb_options_set_compression_options_use_zstd_dict_trainer(
+    rocksdb_options_t* opt, unsigned char use_zstd_dict_trainer);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_options_get_compression_options_use_zstd_dict_trainer(
+    rocksdb_options_t* opt);
+
+extern ROCKSDB_LIBRARY_API void
+rocksdb_options_set_compression_options_parallel_threads(rocksdb_options_t* opt,
+                                                         int value);
+
+extern ROCKSDB_LIBRARY_API int
+rocksdb_options_get_compression_options_parallel_threads(
+    rocksdb_options_t* opt);
+
+extern ROCKSDB_LIBRARY_API void
+rocksdb_options_set_compression_options_max_dict_buffer_bytes(
+    rocksdb_options_t* opt, uint64_t max_dict_buffer_bytes);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_options_get_compression_options_max_dict_buffer_bytes(
+    rocksdb_options_t* opt);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_options_get_bottommost_compression_options_use_zstd_dict_trainer(
+    rocksdb_options_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_options_set_use_fsync(
+    rocksdb_options_t* opt, int use_fsync);
+
+extern ROCKSDB_LIBRARY_API int rocksdb_options_get_use_fsync(
+    rocksdb_options_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_options_set_disable_auto_compactions(
+    rocksdb_options_t* opt, int disable);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_options_set_optimize_filters_for_hits(
+    rocksdb_options_t* opt, int v);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_options_set_report_bg_io_stats(
+    rocksdb_options_t* opt, int v);

c_api_gen/c_generated_readoptions_auto.cc.inc

@@ -0,0 +1,236 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/auto_simple_bindings.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Sources:
+//   - include/rocksdb/options.h
+//   - tools/c_api_gen/auto_simple_bindings_blocklist.json
+//   - include/rocksdb/c.h
+//   - tools/c_api_gen/c_base.cc
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/auto_simple_bindings.py
+
+// Output group: Auto-discovered ReadOptions simple
+
+/* ReadOptions */
+
+void rocksdb_readoptions_set_deadline(rocksdb_readoptions_t* opt,
+                                      uint64_t microseconds) {
+  opt->rep.deadline = std::chrono::microseconds(microseconds);
+}
+
+uint64_t rocksdb_readoptions_get_deadline(rocksdb_readoptions_t* opt) {
+  return opt->rep.deadline.count();
+}
+
+void rocksdb_readoptions_set_io_timeout(rocksdb_readoptions_t* opt,
+                                        uint64_t microseconds) {
+  opt->rep.io_timeout = std::chrono::microseconds(microseconds);
+}
+
+uint64_t rocksdb_readoptions_get_io_timeout(rocksdb_readoptions_t* opt) {
+  return opt->rep.io_timeout.count();
+}
+
+void rocksdb_readoptions_set_read_tier(rocksdb_readoptions_t* opt, int v) {
+  opt->rep.read_tier = static_cast<decltype(opt->rep.read_tier)>(v);
+}
+
+int rocksdb_readoptions_get_read_tier(rocksdb_readoptions_t* opt) {
+  return static_cast<int>(opt->rep.read_tier);
+}
+
+void rocksdb_readoptions_set_rate_limiter_priority(rocksdb_readoptions_t* opt,
+                                                   int v) {
+  opt->rep.rate_limiter_priority =
+      static_cast<decltype(opt->rep.rate_limiter_priority)>(v);
+}
+
+int rocksdb_readoptions_get_rate_limiter_priority(rocksdb_readoptions_t* opt) {
+  return static_cast<int>(opt->rep.rate_limiter_priority);
+}
+
+void rocksdb_readoptions_set_value_size_soft_limit(rocksdb_readoptions_t* opt,
+                                                   uint64_t v) {
+  opt->rep.value_size_soft_limit = v;
+}
+
+uint64_t rocksdb_readoptions_get_value_size_soft_limit(
+    rocksdb_readoptions_t* opt) {
+  return opt->rep.value_size_soft_limit;
+}
+
+void rocksdb_readoptions_set_verify_checksums(rocksdb_readoptions_t* opt,
+                                              unsigned char v) {
+  opt->rep.verify_checksums = v;
+}
+
+unsigned char rocksdb_readoptions_get_verify_checksums(
+    rocksdb_readoptions_t* opt) {
+  return opt->rep.verify_checksums;
+}
+
+void rocksdb_readoptions_set_fill_cache(rocksdb_readoptions_t* opt,
+                                        unsigned char v) {
+  opt->rep.fill_cache = v;
+}
+
+unsigned char rocksdb_readoptions_get_fill_cache(rocksdb_readoptions_t* opt) {
+  return opt->rep.fill_cache;
+}
+
+void rocksdb_readoptions_set_ignore_range_deletions(rocksdb_readoptions_t* opt,
+                                                    unsigned char v) {
+  opt->rep.ignore_range_deletions = v;
+}
+
+unsigned char rocksdb_readoptions_get_ignore_range_deletions(
+    rocksdb_readoptions_t* opt) {
+  return opt->rep.ignore_range_deletions;
+}
+
+void rocksdb_readoptions_set_async_io(rocksdb_readoptions_t* opt,
+                                      unsigned char v) {
+  opt->rep.async_io = v;
+}
+
+unsigned char rocksdb_readoptions_get_async_io(rocksdb_readoptions_t* opt) {
+  return opt->rep.async_io;
+}
+
+void rocksdb_readoptions_set_optimize_multiget_for_io(
+    rocksdb_readoptions_t* opt, unsigned char v) {
+  opt->rep.optimize_multiget_for_io = v;
+}
+
+unsigned char rocksdb_readoptions_get_optimize_multiget_for_io(
+    rocksdb_readoptions_t* opt) {
+  return opt->rep.optimize_multiget_for_io;
+}
+
+void rocksdb_readoptions_set_readahead_size(rocksdb_readoptions_t* opt,
+                                            size_t v) {
+  opt->rep.readahead_size = v;
+}
+
+size_t rocksdb_readoptions_get_readahead_size(rocksdb_readoptions_t* opt) {
+  return opt->rep.readahead_size;
+}
+
+void rocksdb_readoptions_set_max_skippable_internal_keys(
+    rocksdb_readoptions_t* opt, uint64_t v) {
+  opt->rep.max_skippable_internal_keys = v;
+}
+
+uint64_t rocksdb_readoptions_get_max_skippable_internal_keys(
+    rocksdb_readoptions_t* opt) {
+  return opt->rep.max_skippable_internal_keys;
+}
+
+void rocksdb_readoptions_set_tailing(rocksdb_readoptions_t* opt,
+                                     unsigned char v) {
+  opt->rep.tailing = v;
+}
+
+unsigned char rocksdb_readoptions_get_tailing(rocksdb_readoptions_t* opt) {
+  return opt->rep.tailing;
+}
+
+void rocksdb_readoptions_set_total_order_seek(rocksdb_readoptions_t* opt,
+                                              unsigned char v) {
+  opt->rep.total_order_seek = v;
+}
+
+unsigned char rocksdb_readoptions_get_total_order_seek(
+    rocksdb_readoptions_t* opt) {
+  return opt->rep.total_order_seek;
+}
+
+void rocksdb_readoptions_set_auto_prefix_mode(rocksdb_readoptions_t* opt,
+                                              unsigned char v) {
+  opt->rep.auto_prefix_mode = v;
+}
+
+unsigned char rocksdb_readoptions_get_auto_prefix_mode(
+    rocksdb_readoptions_t* opt) {
+  return opt->rep.auto_prefix_mode;
+}
+
+void rocksdb_readoptions_set_prefix_same_as_start(rocksdb_readoptions_t* opt,
+                                                  unsigned char v) {
+  opt->rep.prefix_same_as_start = v;
+}
+
+unsigned char rocksdb_readoptions_get_prefix_same_as_start(
+    rocksdb_readoptions_t* opt) {
+  return opt->rep.prefix_same_as_start;
+}
+
+void rocksdb_readoptions_set_pin_data(rocksdb_readoptions_t* opt,
+                                      unsigned char v) {
+  opt->rep.pin_data = v;
+}
+
+unsigned char rocksdb_readoptions_get_pin_data(rocksdb_readoptions_t* opt) {
+  return opt->rep.pin_data;
+}
+
+void rocksdb_readoptions_set_adaptive_readahead(rocksdb_readoptions_t* opt,
+                                                unsigned char v) {
+  opt->rep.adaptive_readahead = v;
+}
+
+unsigned char rocksdb_readoptions_get_adaptive_readahead(
+    rocksdb_readoptions_t* opt) {
+  return opt->rep.adaptive_readahead;
+}
+
+void rocksdb_readoptions_set_background_purge_on_iterator_cleanup(
+    rocksdb_readoptions_t* opt, unsigned char v) {
+  opt->rep.background_purge_on_iterator_cleanup = v;
+}
+
+unsigned char rocksdb_readoptions_get_background_purge_on_iterator_cleanup(
+    rocksdb_readoptions_t* opt) {
+  return opt->rep.background_purge_on_iterator_cleanup;
+}
+
+void rocksdb_readoptions_set_auto_readahead_size(rocksdb_readoptions_t* opt,
+                                                 unsigned char v) {
+  opt->rep.auto_readahead_size = v;
+}
+
+unsigned char rocksdb_readoptions_get_auto_readahead_size(
+    rocksdb_readoptions_t* opt) {
+  return opt->rep.auto_readahead_size;
+}
+
+void rocksdb_readoptions_set_allow_unprepared_value(rocksdb_readoptions_t* opt,
+                                                    unsigned char v) {
+  opt->rep.allow_unprepared_value = v;
+}
+
+unsigned char rocksdb_readoptions_get_allow_unprepared_value(
+    rocksdb_readoptions_t* opt) {
+  return opt->rep.allow_unprepared_value;
+}
+
+void rocksdb_readoptions_set_auto_refresh_iterator_with_snapshot(
+    rocksdb_readoptions_t* opt, unsigned char v) {
+  opt->rep.auto_refresh_iterator_with_snapshot = v;
+}
+
+unsigned char rocksdb_readoptions_get_auto_refresh_iterator_with_snapshot(
+    rocksdb_readoptions_t* opt) {
+  return opt->rep.auto_refresh_iterator_with_snapshot;
+}
+
+void rocksdb_readoptions_set_io_activity(rocksdb_readoptions_t* opt, int v) {
+  opt->rep.io_activity = static_cast<decltype(opt->rep.io_activity)>(v);
+}
+
+int rocksdb_readoptions_get_io_activity(rocksdb_readoptions_t* opt) {
+  return static_cast<int>(opt->rep.io_activity);
+}

c_api_gen/c_generated_readoptions_auto.h.inc

@@ -0,0 +1,161 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/auto_simple_bindings.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Sources:
+//   - include/rocksdb/options.h
+//   - tools/c_api_gen/auto_simple_bindings_blocklist.json
+//   - include/rocksdb/c.h
+//   - tools/c_api_gen/c_base.cc
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/auto_simple_bindings.py
+
+// Output group: Auto-discovered ReadOptions simple
+
+/* ReadOptions */
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_deadline(
+    rocksdb_readoptions_t* opt, uint64_t microseconds);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_readoptions_get_deadline(rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_io_timeout(
+    rocksdb_readoptions_t* opt, uint64_t microseconds);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_readoptions_get_io_timeout(rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_read_tier(
+    rocksdb_readoptions_t* opt, int v);
+
+extern ROCKSDB_LIBRARY_API int rocksdb_readoptions_get_read_tier(
+    rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_rate_limiter_priority(
+    rocksdb_readoptions_t* opt, int v);
+
+extern ROCKSDB_LIBRARY_API int rocksdb_readoptions_get_rate_limiter_priority(
+    rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_value_size_soft_limit(
+    rocksdb_readoptions_t* opt, uint64_t v);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_readoptions_get_value_size_soft_limit(rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_verify_checksums(
+    rocksdb_readoptions_t* opt, unsigned char v);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_readoptions_get_verify_checksums(rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_fill_cache(
+    rocksdb_readoptions_t* opt, unsigned char v);
+
+extern ROCKSDB_LIBRARY_API unsigned char rocksdb_readoptions_get_fill_cache(
+    rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_ignore_range_deletions(
+    rocksdb_readoptions_t* opt, unsigned char v);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_readoptions_get_ignore_range_deletions(rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_async_io(
+    rocksdb_readoptions_t* opt, unsigned char v);
+
+extern ROCKSDB_LIBRARY_API unsigned char rocksdb_readoptions_get_async_io(
+    rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void
+rocksdb_readoptions_set_optimize_multiget_for_io(rocksdb_readoptions_t* opt,
+                                                 unsigned char v);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_readoptions_get_optimize_multiget_for_io(rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_readahead_size(
+    rocksdb_readoptions_t* opt, size_t v);
+
+extern ROCKSDB_LIBRARY_API size_t
+rocksdb_readoptions_get_readahead_size(rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void
+rocksdb_readoptions_set_max_skippable_internal_keys(rocksdb_readoptions_t* opt,
+                                                    uint64_t v);
+
+extern ROCKSDB_LIBRARY_API uint64_t
+rocksdb_readoptions_get_max_skippable_internal_keys(rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_tailing(
+    rocksdb_readoptions_t* opt, unsigned char v);
+
+extern ROCKSDB_LIBRARY_API unsigned char rocksdb_readoptions_get_tailing(
+    rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_total_order_seek(
+    rocksdb_readoptions_t* opt, unsigned char v);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_readoptions_get_total_order_seek(rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_auto_prefix_mode(
+    rocksdb_readoptions_t* opt, unsigned char v);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_readoptions_get_auto_prefix_mode(rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_prefix_same_as_start(
+    rocksdb_readoptions_t* opt, unsigned char v);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_readoptions_get_prefix_same_as_start(rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_pin_data(
+    rocksdb_readoptions_t* opt, unsigned char v);
+
+extern ROCKSDB_LIBRARY_API unsigned char rocksdb_readoptions_get_pin_data(
+    rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_adaptive_readahead(
+    rocksdb_readoptions_t* opt, unsigned char v);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_readoptions_get_adaptive_readahead(rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void
+rocksdb_readoptions_set_background_purge_on_iterator_cleanup(
+    rocksdb_readoptions_t* opt, unsigned char v);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_readoptions_get_background_purge_on_iterator_cleanup(
+    rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_auto_readahead_size(
+    rocksdb_readoptions_t* opt, unsigned char v);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_readoptions_get_auto_readahead_size(rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_allow_unprepared_value(
+    rocksdb_readoptions_t* opt, unsigned char v);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_readoptions_get_allow_unprepared_value(rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void
+rocksdb_readoptions_set_auto_refresh_iterator_with_snapshot(
+    rocksdb_readoptions_t* opt, unsigned char v);
+
+extern ROCKSDB_LIBRARY_API unsigned char
+rocksdb_readoptions_get_auto_refresh_iterator_with_snapshot(
+    rocksdb_readoptions_t* opt);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_readoptions_set_io_activity(
+    rocksdb_readoptions_t* opt, int v);
+
+extern ROCKSDB_LIBRARY_API int rocksdb_readoptions_get_io_activity(
+    rocksdb_readoptions_t* opt);

c_api_gen/c_generated_subset.cc.inc

@@ -0,0 +1,371 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/generate_c_api.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Source: tools/c_api_gen/spec.json
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/generate_c_api.py
+//     --spec tools/c_api_gen/spec.json --section 'DB data operations'
+//     --section 'WriteBatch' --section 'Transaction'
+//     --section 'TransactionDB simple'
+//     --header-out
+//     c_api_gen/c_generated_subset.h.inc
+//     --source-out
+//     c_api_gen/c_generated_subset.cc.inc
+
+/* DB data operations */
+
+void rocksdb_put(rocksdb_t* db, const rocksdb_writeoptions_t* options,
+                 const char* key, size_t keylen, const char* val, size_t vallen,
+                 char** errptr) {
+  SaveError(errptr,
+            db->rep->Put(options->rep, Slice(key, keylen), Slice(val, vallen)));
+}
+
+void rocksdb_put_cf(rocksdb_t* db, const rocksdb_writeoptions_t* options,
+                    rocksdb_column_family_handle_t* column_family,
+                    const char* key, size_t keylen, const char* val,
+                    size_t vallen, char** errptr) {
+  SaveError(errptr, db->rep->Put(options->rep, column_family->rep,
+                                 Slice(key, keylen), Slice(val, vallen)));
+}
+
+void rocksdb_delete(rocksdb_t* db, const rocksdb_writeoptions_t* options,
+                    const char* key, size_t keylen, char** errptr) {
+  SaveError(errptr, db->rep->Delete(options->rep, Slice(key, keylen)));
+}
+
+void rocksdb_delete_cf(rocksdb_t* db, const rocksdb_writeoptions_t* options,
+                       rocksdb_column_family_handle_t* column_family,
+                       const char* key, size_t keylen, char** errptr) {
+  SaveError(errptr, db->rep->Delete(options->rep, column_family->rep,
+                                    Slice(key, keylen)));
+}
+
+void rocksdb_merge(rocksdb_t* db, const rocksdb_writeoptions_t* options,
+                   const char* key, size_t keylen, const char* val,
+                   size_t vallen, char** errptr) {
+  SaveError(errptr, db->rep->Merge(options->rep, Slice(key, keylen),
+                                   Slice(val, vallen)));
+}
+
+void rocksdb_merge_cf(rocksdb_t* db, const rocksdb_writeoptions_t* options,
+                      rocksdb_column_family_handle_t* column_family,
+                      const char* key, size_t keylen, const char* val,
+                      size_t vallen, char** errptr) {
+  SaveError(errptr, db->rep->Merge(options->rep, column_family->rep,
+                                   Slice(key, keylen), Slice(val, vallen)));
+}
+
+void rocksdb_write(rocksdb_t* db, const rocksdb_writeoptions_t* options,
+                   rocksdb_writebatch_t* batch, char** errptr) {
+  SaveError(errptr, db->rep->Write(options->rep, &batch->rep));
+}
+
+void rocksdb_put_with_ts(rocksdb_t* db, const rocksdb_writeoptions_t* options,
+                         const char* key, size_t keylen, const char* ts,
+                         size_t tslen, const char* val, size_t vallen,
+                         char** errptr) {
+  SaveError(errptr, db->rep->Put(options->rep, Slice(key, keylen),
+                                 Slice(ts, tslen), Slice(val, vallen)));
+}
+
+void rocksdb_put_cf_with_ts(rocksdb_t* db,
+                            const rocksdb_writeoptions_t* options,
+                            rocksdb_column_family_handle_t* column_family,
+                            const char* key, size_t keylen, const char* ts,
+                            size_t tslen, const char* val, size_t vallen,
+                            char** errptr) {
+  SaveError(errptr,
+            db->rep->Put(options->rep, column_family->rep, Slice(key, keylen),
+                         Slice(ts, tslen), Slice(val, vallen)));
+}
+
+void rocksdb_delete_with_ts(rocksdb_t* db,
+                            const rocksdb_writeoptions_t* options,
+                            const char* key, size_t keylen, const char* ts,
+                            size_t tslen, char** errptr) {
+  SaveError(errptr, db->rep->Delete(options->rep, Slice(key, keylen),
+                                    Slice(ts, tslen)));
+}
+
+void rocksdb_delete_cf_with_ts(rocksdb_t* db,
+                               const rocksdb_writeoptions_t* options,
+                               rocksdb_column_family_handle_t* column_family,
+                               const char* key, size_t keylen, const char* ts,
+                               size_t tslen, char** errptr) {
+  SaveError(errptr, db->rep->Delete(options->rep, column_family->rep,
+                                    Slice(key, keylen), Slice(ts, tslen)));
+}
+
+void rocksdb_singledelete(rocksdb_t* db, const rocksdb_writeoptions_t* options,
+                          const char* key, size_t keylen, char** errptr) {
+  SaveError(errptr, db->rep->SingleDelete(options->rep, Slice(key, keylen)));
+}
+
+void rocksdb_singledelete_cf(rocksdb_t* db,
+                             const rocksdb_writeoptions_t* options,
+                             rocksdb_column_family_handle_t* column_family,
+                             const char* key, size_t keylen, char** errptr) {
+  SaveError(errptr, db->rep->SingleDelete(options->rep, column_family->rep,
+                                          Slice(key, keylen)));
+}
+
+void rocksdb_singledelete_with_ts(rocksdb_t* db,
+                                  const rocksdb_writeoptions_t* options,
+                                  const char* key, size_t keylen,
+                                  const char* ts, size_t tslen, char** errptr) {
+  SaveError(errptr, db->rep->SingleDelete(options->rep, Slice(key, keylen),
+                                          Slice(ts, tslen)));
+}
+
+void rocksdb_singledelete_cf_with_ts(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, const char* ts, size_t tslen, char** errptr) {
+  SaveError(errptr,
+            db->rep->SingleDelete(options->rep, column_family->rep,
+                                  Slice(key, keylen), Slice(ts, tslen)));
+}
+
+void rocksdb_delete_range_cf(rocksdb_t* db,
+                             const rocksdb_writeoptions_t* options,
+                             rocksdb_column_family_handle_t* column_family,
+                             const char* start_key, size_t start_key_len,
+                             const char* end_key, size_t end_key_len,
+                             char** errptr) {
+  SaveError(errptr, db->rep->DeleteRange(options->rep, column_family->rep,
+                                         Slice(start_key, start_key_len),
+                                         Slice(end_key, end_key_len)));
+}
+
+void rocksdb_flush(rocksdb_t* db, const rocksdb_flushoptions_t* options,
+                   char** errptr) {
+  SaveError(errptr, db->rep->Flush(options->rep));
+}
+
+void rocksdb_flush_cf(rocksdb_t* db, const rocksdb_flushoptions_t* options,
+                      rocksdb_column_family_handle_t* column_family,
+                      char** errptr) {
+  SaveError(errptr, db->rep->Flush(options->rep, column_family->rep));
+}
+
+void rocksdb_flush_wal(rocksdb_t* db, unsigned char sync, char** errptr) {
+  SaveError(errptr, db->rep->FlushWAL(sync));
+}
+
+void rocksdb_pause_background_work(rocksdb_t* db, char** errptr) {
+  SaveError(errptr, db->rep->PauseBackgroundWork());
+}
+
+void rocksdb_continue_background_work(rocksdb_t* db, char** errptr) {
+  SaveError(errptr, db->rep->ContinueBackgroundWork());
+}
+
+void rocksdb_disable_file_deletions(rocksdb_t* db, char** errptr) {
+  SaveError(errptr, db->rep->DisableFileDeletions());
+}
+
+void rocksdb_enable_file_deletions(rocksdb_t* db, char** errptr) {
+  SaveError(errptr, db->rep->EnableFileDeletions());
+}
+
+void rocksdb_verify_checksum(rocksdb_t* db, char** errptr) {
+  SaveError(errptr, db->rep->VerifyChecksum());
+}
+
+void rocksdb_verify_file_checksums(rocksdb_t* db, char** errptr) {
+  SaveError(errptr, db->rep->VerifyFileChecksums(ReadOptions()));
+}
+
+void rocksdb_destroy_db(const rocksdb_options_t* options, const char* name,
+                        char** errptr) {
+  SaveError(errptr, DestroyDB(name, options->rep));
+}
+
+void rocksdb_repair_db(const rocksdb_options_t* options, const char* name,
+                       char** errptr) {
+  SaveError(errptr, RepairDB(name, options->rep));
+}
+
+/* WriteBatch */
+
+void rocksdb_writebatch_clear(rocksdb_writebatch_t* b) { b->rep.Clear(); }
+
+void rocksdb_writebatch_put(rocksdb_writebatch_t* b, const char* key,
+                            size_t klen, const char* val, size_t vlen) {
+  b->rep.Put(Slice(key, klen), Slice(val, vlen));
+}
+
+void rocksdb_writebatch_put_cf(rocksdb_writebatch_t* b,
+                               rocksdb_column_family_handle_t* column_family,
+                               const char* key, size_t klen, const char* val,
+                               size_t vlen) {
+  b->rep.Put(column_family->rep, Slice(key, klen), Slice(val, vlen));
+}
+
+void rocksdb_writebatch_delete(rocksdb_writebatch_t* b, const char* key,
+                               size_t klen) {
+  b->rep.Delete(Slice(key, klen));
+}
+
+void rocksdb_writebatch_put_log_data(rocksdb_writebatch_t* b, const char* blob,
+                                     size_t len) {
+  b->rep.PutLogData(Slice(blob, len));
+}
+
+void rocksdb_writebatch_set_save_point(rocksdb_writebatch_t* b) {
+  b->rep.SetSavePoint();
+}
+
+void rocksdb_writebatch_rollback_to_save_point(rocksdb_writebatch_t* b,
+                                               char** errptr) {
+  SaveError(errptr, b->rep.RollbackToSavePoint());
+}
+
+void rocksdb_writebatch_pop_save_point(rocksdb_writebatch_t* b, char** errptr) {
+  SaveError(errptr, b->rep.PopSavePoint());
+}
+
+void rocksdb_writebatch_verify_checksum(rocksdb_writebatch_t* b,
+                                        char** errptr) {
+  SaveError(errptr, b->rep.VerifyChecksum());
+}
+
+/* Transaction */
+
+void rocksdb_transaction_put(rocksdb_transaction_t* txn, const char* key,
+                             size_t klen, const char* val, size_t vlen,
+                             char** errptr) {
+  SaveError(errptr, txn->rep->Put(Slice(key, klen), Slice(val, vlen)));
+}
+
+void rocksdb_transaction_put_cf(rocksdb_transaction_t* txn,
+                                rocksdb_column_family_handle_t* column_family,
+                                const char* key, size_t klen, const char* val,
+                                size_t vlen, char** errptr) {
+  SaveError(errptr, txn->rep->Put(column_family->rep, Slice(key, klen),
+                                  Slice(val, vlen)));
+}
+
+void rocksdb_transaction_merge(rocksdb_transaction_t* txn, const char* key,
+                               size_t klen, const char* val, size_t vlen,
+                               char** errptr) {
+  SaveError(errptr, txn->rep->Merge(Slice(key, klen), Slice(val, vlen)));
+}
+
+void rocksdb_transaction_merge_cf(rocksdb_transaction_t* txn,
+                                  rocksdb_column_family_handle_t* column_family,
+                                  const char* key, size_t klen, const char* val,
+                                  size_t vlen, char** errptr) {
+  SaveError(errptr, txn->rep->Merge(column_family->rep, Slice(key, klen),
+                                    Slice(val, vlen)));
+}
+
+void rocksdb_transaction_delete(rocksdb_transaction_t* txn, const char* key,
+                                size_t klen, char** errptr) {
+  SaveError(errptr, txn->rep->Delete(Slice(key, klen)));
+}
+
+void rocksdb_transaction_delete_cf(
+    rocksdb_transaction_t* txn, rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t klen, char** errptr) {
+  SaveError(errptr, txn->rep->Delete(column_family->rep, Slice(key, klen)));
+}
+
+void rocksdb_transaction_commit(rocksdb_transaction_t* txn, char** errptr) {
+  SaveError(errptr, txn->rep->Commit());
+}
+
+void rocksdb_transaction_rollback(rocksdb_transaction_t* txn, char** errptr) {
+  SaveError(errptr, txn->rep->Rollback());
+}
+
+void rocksdb_transaction_put_log_data(rocksdb_transaction_t* txn,
+                                      const char* blob, size_t len) {
+  txn->rep->PutLogData(Slice(blob, len));
+}
+
+void rocksdb_transaction_set_savepoint(rocksdb_transaction_t* txn) {
+  txn->rep->SetSavePoint();
+}
+
+void rocksdb_transaction_rollback_to_savepoint(rocksdb_transaction_t* txn,
+                                               char** errptr) {
+  SaveError(errptr, txn->rep->RollbackToSavePoint());
+}
+
+/* TransactionDB simple */
+
+void rocksdb_transactiondb_put(rocksdb_transactiondb_t* txn_db,
+                               const rocksdb_writeoptions_t* options,
+                               const char* key, size_t klen, const char* val,
+                               size_t vlen, char** errptr) {
+  SaveError(errptr,
+            txn_db->rep->Put(options->rep, Slice(key, klen), Slice(val, vlen)));
+}
+
+void rocksdb_transactiondb_put_cf(rocksdb_transactiondb_t* txn_db,
+                                  const rocksdb_writeoptions_t* options,
+                                  rocksdb_column_family_handle_t* column_family,
+                                  const char* key, size_t keylen,
+                                  const char* val, size_t vallen,
+                                  char** errptr) {
+  SaveError(errptr, txn_db->rep->Put(options->rep, column_family->rep,
+                                     Slice(key, keylen), Slice(val, vallen)));
+}
+
+void rocksdb_transactiondb_write(rocksdb_transactiondb_t* db,
+                                 const rocksdb_writeoptions_t* options,
+                                 rocksdb_writebatch_t* batch, char** errptr) {
+  SaveError(errptr, db->rep->Write(options->rep, &batch->rep));
+}
+
+void rocksdb_transactiondb_merge(rocksdb_transactiondb_t* txn_db,
+                                 const rocksdb_writeoptions_t* options,
+                                 const char* key, size_t klen, const char* val,
+                                 size_t vlen, char** errptr) {
+  SaveError(errptr, txn_db->rep->Merge(options->rep, Slice(key, klen),
+                                       Slice(val, vlen)));
+}
+
+void rocksdb_transactiondb_merge_cf(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key, size_t klen,
+    const char* val, size_t vlen, char** errptr) {
+  SaveError(errptr, txn_db->rep->Merge(options->rep, column_family->rep,
+                                       Slice(key, klen), Slice(val, vlen)));
+}
+
+void rocksdb_transactiondb_delete(rocksdb_transactiondb_t* txn_db,
+                                  const rocksdb_writeoptions_t* options,
+                                  const char* key, size_t klen, char** errptr) {
+  SaveError(errptr, txn_db->rep->Delete(options->rep, Slice(key, klen)));
+}
+
+void rocksdb_transactiondb_delete_cf(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, char** errptr) {
+  SaveError(errptr, txn_db->rep->Delete(options->rep, column_family->rep,
+                                        Slice(key, keylen)));
+}
+
+void rocksdb_transactiondb_flush_wal(rocksdb_transactiondb_t* txn_db,
+                                     unsigned char sync, char** errptr) {
+  SaveError(errptr, txn_db->rep->FlushWAL(sync));
+}
+
+void rocksdb_transactiondb_flush(rocksdb_transactiondb_t* txn_db,
+                                 const rocksdb_flushoptions_t* options,
+                                 char** errptr) {
+  SaveError(errptr, txn_db->rep->Flush(options->rep));
+}
+
+void rocksdb_transactiondb_flush_cf(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_flushoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, char** errptr) {
+  SaveError(errptr, txn_db->rep->Flush(options->rep, column_family->rep));
+}

c_api_gen/c_generated_subset.h.inc

@@ -0,0 +1,245 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/generate_c_api.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Source: tools/c_api_gen/spec.json
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/generate_c_api.py
+//     --spec tools/c_api_gen/spec.json --section 'DB data operations'
+//     --section 'WriteBatch' --section 'Transaction'
+//     --section 'TransactionDB simple'
+//     --header-out
+//     c_api_gen/c_generated_subset.h.inc
+//     --source-out
+//     c_api_gen/c_generated_subset.cc.inc
+
+/* DB data operations */
+
+extern ROCKSDB_LIBRARY_API void rocksdb_put(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options, const char* key,
+    size_t keylen, const char* val, size_t vallen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_put_cf(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, const char* val, size_t vallen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_delete(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options, const char* key,
+    size_t keylen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_delete_cf(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_merge(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options, const char* key,
+    size_t keylen, const char* val, size_t vallen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_merge_cf(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, const char* val, size_t vallen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_write(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_writebatch_t* batch, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_put_with_ts(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options, const char* key,
+    size_t keylen, const char* ts, size_t tslen, const char* val, size_t vallen,
+    char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_put_cf_with_ts(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, const char* ts, size_t tslen, const char* val, size_t vallen,
+    char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_delete_with_ts(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options, const char* key,
+    size_t keylen, const char* ts, size_t tslen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_delete_cf_with_ts(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, const char* ts, size_t tslen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_singledelete(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options, const char* key,
+    size_t keylen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_singledelete_cf(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_singledelete_with_ts(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options, const char* key,
+    size_t keylen, const char* ts, size_t tslen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_singledelete_cf_with_ts(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, const char* ts, size_t tslen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_delete_range_cf(
+    rocksdb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* start_key,
+    size_t start_key_len, const char* end_key, size_t end_key_len,
+    char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_flush(
+    rocksdb_t* db, const rocksdb_flushoptions_t* options, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_flush_cf(
+    rocksdb_t* db, const rocksdb_flushoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_flush_wal(rocksdb_t* db,
+                                                  unsigned char sync,
+                                                  char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_pause_background_work(rocksdb_t* db,
+                                                              char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_continue_background_work(rocksdb_t* db,
+                                                                 char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_disable_file_deletions(rocksdb_t* db,
+                                                               char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_enable_file_deletions(rocksdb_t* db,
+                                                              char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_verify_checksum(rocksdb_t* db,
+                                                        char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_verify_file_checksums(rocksdb_t* db,
+                                                              char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_destroy_db(
+    const rocksdb_options_t* options, const char* name, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_repair_db(
+    const rocksdb_options_t* options, const char* name, char** errptr);
+
+/* WriteBatch */
+
+extern ROCKSDB_LIBRARY_API void rocksdb_writebatch_clear(
+    rocksdb_writebatch_t* b);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_writebatch_put(rocksdb_writebatch_t* b,
+                                                       const char* key,
+                                                       size_t klen,
+                                                       const char* val,
+                                                       size_t vlen);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_writebatch_put_cf(
+    rocksdb_writebatch_t* b, rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t klen, const char* val, size_t vlen);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_writebatch_delete(
+    rocksdb_writebatch_t* b, const char* key, size_t klen);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_writebatch_put_log_data(
+    rocksdb_writebatch_t* b, const char* blob, size_t len);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_writebatch_set_save_point(
+    rocksdb_writebatch_t* b);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_writebatch_rollback_to_save_point(
+    rocksdb_writebatch_t* b, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_writebatch_pop_save_point(
+    rocksdb_writebatch_t* b, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_writebatch_verify_checksum(
+    rocksdb_writebatch_t* b, char** errptr);
+
+/* Transaction */
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_put(
+    rocksdb_transaction_t* txn, const char* key, size_t klen, const char* val,
+    size_t vlen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_put_cf(
+    rocksdb_transaction_t* txn, rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t klen, const char* val, size_t vlen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_merge(
+    rocksdb_transaction_t* txn, const char* key, size_t klen, const char* val,
+    size_t vlen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_merge_cf(
+    rocksdb_transaction_t* txn, rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t klen, const char* val, size_t vlen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_delete(
+    rocksdb_transaction_t* txn, const char* key, size_t klen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_delete_cf(
+    rocksdb_transaction_t* txn, rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t klen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_commit(
+    rocksdb_transaction_t* txn, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_rollback(
+    rocksdb_transaction_t* txn, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_put_log_data(
+    rocksdb_transaction_t* txn, const char* blob, size_t len);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_set_savepoint(
+    rocksdb_transaction_t* txn);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_rollback_to_savepoint(
+    rocksdb_transaction_t* txn, char** errptr);
+
+/* TransactionDB simple */
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_put(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_writeoptions_t* options,
+    const char* key, size_t klen, const char* val, size_t vlen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_put_cf(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, const char* val, size_t vallen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_write(
+    rocksdb_transactiondb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_writebatch_t* batch, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_merge(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_writeoptions_t* options,
+    const char* key, size_t klen, const char* val, size_t vlen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_merge_cf(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key, size_t klen,
+    const char* val, size_t vlen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_delete(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_writeoptions_t* options,
+    const char* key, size_t klen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_delete_cf(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_flush_wal(
+    rocksdb_transactiondb_t* txn_db, unsigned char sync, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_flush(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_flushoptions_t* options,
+    char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_flush_cf(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_flushoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, char** errptr);

c_api_gen/c_generated_transaction_subset.cc.inc

@@ -0,0 +1,77 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/generate_c_api.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Source: tools/c_api_gen/spec.json
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/generate_c_api.py
+//     --spec tools/c_api_gen/spec.json --section 'Transaction'
+//     --header-out
+//     c_api_gen/c_generated_transaction_subset.h.inc
+//     --source-out
+//     c_api_gen/c_generated_transaction_subset.cc.inc
+
+/* Transaction */
+
+void rocksdb_transaction_put(rocksdb_transaction_t* txn, const char* key,
+                             size_t klen, const char* val, size_t vlen,
+                             char** errptr) {
+  SaveError(errptr, txn->rep->Put(Slice(key, klen), Slice(val, vlen)));
+}
+
+void rocksdb_transaction_put_cf(rocksdb_transaction_t* txn,
+                                rocksdb_column_family_handle_t* column_family,
+                                const char* key, size_t klen, const char* val,
+                                size_t vlen, char** errptr) {
+  SaveError(errptr, txn->rep->Put(column_family->rep, Slice(key, klen),
+                                  Slice(val, vlen)));
+}
+
+void rocksdb_transaction_merge(rocksdb_transaction_t* txn, const char* key,
+                               size_t klen, const char* val, size_t vlen,
+                               char** errptr) {
+  SaveError(errptr, txn->rep->Merge(Slice(key, klen), Slice(val, vlen)));
+}
+
+void rocksdb_transaction_merge_cf(rocksdb_transaction_t* txn,
+                                  rocksdb_column_family_handle_t* column_family,
+                                  const char* key, size_t klen, const char* val,
+                                  size_t vlen, char** errptr) {
+  SaveError(errptr, txn->rep->Merge(column_family->rep, Slice(key, klen),
+                                    Slice(val, vlen)));
+}
+
+void rocksdb_transaction_delete(rocksdb_transaction_t* txn, const char* key,
+                                size_t klen, char** errptr) {
+  SaveError(errptr, txn->rep->Delete(Slice(key, klen)));
+}
+
+void rocksdb_transaction_delete_cf(
+    rocksdb_transaction_t* txn, rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t klen, char** errptr) {
+  SaveError(errptr, txn->rep->Delete(column_family->rep, Slice(key, klen)));
+}
+
+void rocksdb_transaction_commit(rocksdb_transaction_t* txn, char** errptr) {
+  SaveError(errptr, txn->rep->Commit());
+}
+
+void rocksdb_transaction_rollback(rocksdb_transaction_t* txn, char** errptr) {
+  SaveError(errptr, txn->rep->Rollback());
+}
+
+void rocksdb_transaction_put_log_data(rocksdb_transaction_t* txn,
+                                      const char* blob, size_t len) {
+  txn->rep->PutLogData(Slice(blob, len));
+}
+
+void rocksdb_transaction_set_savepoint(rocksdb_transaction_t* txn) {
+  txn->rep->SetSavePoint();
+}
+
+void rocksdb_transaction_rollback_to_savepoint(rocksdb_transaction_t* txn,
+                                               char** errptr) {
+  SaveError(errptr, txn->rep->RollbackToSavePoint());
+}

c_api_gen/c_generated_transaction_subset.h.inc

@@ -0,0 +1,54 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/generate_c_api.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Source: tools/c_api_gen/spec.json
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/generate_c_api.py
+//     --spec tools/c_api_gen/spec.json --section 'Transaction'
+//     --header-out
+//     c_api_gen/c_generated_transaction_subset.h.inc
+//     --source-out
+//     c_api_gen/c_generated_transaction_subset.cc.inc
+
+/* Transaction */
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_put(
+    rocksdb_transaction_t* txn, const char* key, size_t klen, const char* val,
+    size_t vlen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_put_cf(
+    rocksdb_transaction_t* txn, rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t klen, const char* val, size_t vlen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_merge(
+    rocksdb_transaction_t* txn, const char* key, size_t klen, const char* val,
+    size_t vlen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_merge_cf(
+    rocksdb_transaction_t* txn, rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t klen, const char* val, size_t vlen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_delete(
+    rocksdb_transaction_t* txn, const char* key, size_t klen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_delete_cf(
+    rocksdb_transaction_t* txn, rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t klen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_commit(
+    rocksdb_transaction_t* txn, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_rollback(
+    rocksdb_transaction_t* txn, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_put_log_data(
+    rocksdb_transaction_t* txn, const char* blob, size_t len);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_set_savepoint(
+    rocksdb_transaction_t* txn);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transaction_rollback_to_savepoint(
+    rocksdb_transaction_t* txn, char** errptr);

c_api_gen/c_generated_transactiondb_subset.cc.inc

@@ -0,0 +1,87 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/generate_c_api.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Source: tools/c_api_gen/spec.json
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/generate_c_api.py
+//     --spec tools/c_api_gen/spec.json --section 'TransactionDB simple'
+//     --header-out
+//     c_api_gen/c_generated_transactiondb_subset.h.inc
+//     --source-out
+//     c_api_gen/c_generated_transactiondb_subset.cc.inc
+
+/* TransactionDB simple */
+
+void rocksdb_transactiondb_put(rocksdb_transactiondb_t* txn_db,
+                               const rocksdb_writeoptions_t* options,
+                               const char* key, size_t klen, const char* val,
+                               size_t vlen, char** errptr) {
+  SaveError(errptr,
+            txn_db->rep->Put(options->rep, Slice(key, klen), Slice(val, vlen)));
+}
+
+void rocksdb_transactiondb_put_cf(rocksdb_transactiondb_t* txn_db,
+                                  const rocksdb_writeoptions_t* options,
+                                  rocksdb_column_family_handle_t* column_family,
+                                  const char* key, size_t keylen,
+                                  const char* val, size_t vallen,
+                                  char** errptr) {
+  SaveError(errptr, txn_db->rep->Put(options->rep, column_family->rep,
+                                     Slice(key, keylen), Slice(val, vallen)));
+}
+
+void rocksdb_transactiondb_write(rocksdb_transactiondb_t* db,
+                                 const rocksdb_writeoptions_t* options,
+                                 rocksdb_writebatch_t* batch, char** errptr) {
+  SaveError(errptr, db->rep->Write(options->rep, &batch->rep));
+}
+
+void rocksdb_transactiondb_merge(rocksdb_transactiondb_t* txn_db,
+                                 const rocksdb_writeoptions_t* options,
+                                 const char* key, size_t klen, const char* val,
+                                 size_t vlen, char** errptr) {
+  SaveError(errptr, txn_db->rep->Merge(options->rep, Slice(key, klen),
+                                       Slice(val, vlen)));
+}
+
+void rocksdb_transactiondb_merge_cf(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key, size_t klen,
+    const char* val, size_t vlen, char** errptr) {
+  SaveError(errptr, txn_db->rep->Merge(options->rep, column_family->rep,
+                                       Slice(key, klen), Slice(val, vlen)));
+}
+
+void rocksdb_transactiondb_delete(rocksdb_transactiondb_t* txn_db,
+                                  const rocksdb_writeoptions_t* options,
+                                  const char* key, size_t klen, char** errptr) {
+  SaveError(errptr, txn_db->rep->Delete(options->rep, Slice(key, klen)));
+}
+
+void rocksdb_transactiondb_delete_cf(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, char** errptr) {
+  SaveError(errptr, txn_db->rep->Delete(options->rep, column_family->rep,
+                                        Slice(key, keylen)));
+}
+
+void rocksdb_transactiondb_flush_wal(rocksdb_transactiondb_t* txn_db,
+                                     unsigned char sync, char** errptr) {
+  SaveError(errptr, txn_db->rep->FlushWAL(sync));
+}
+
+void rocksdb_transactiondb_flush(rocksdb_transactiondb_t* txn_db,
+                                 const rocksdb_flushoptions_t* options,
+                                 char** errptr) {
+  SaveError(errptr, txn_db->rep->Flush(options->rep));
+}
+
+void rocksdb_transactiondb_flush_cf(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_flushoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, char** errptr) {
+  SaveError(errptr, txn_db->rep->Flush(options->rep, column_family->rep));
+}

c_api_gen/c_generated_transactiondb_subset.h.inc

@@ -0,0 +1,58 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/generate_c_api.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Source: tools/c_api_gen/spec.json
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/generate_c_api.py
+//     --spec tools/c_api_gen/spec.json --section 'TransactionDB simple'
+//     --header-out
+//     c_api_gen/c_generated_transactiondb_subset.h.inc
+//     --source-out
+//     c_api_gen/c_generated_transactiondb_subset.cc.inc
+
+/* TransactionDB simple */
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_put(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_writeoptions_t* options,
+    const char* key, size_t klen, const char* val, size_t vlen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_put_cf(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, const char* val, size_t vallen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_write(
+    rocksdb_transactiondb_t* db, const rocksdb_writeoptions_t* options,
+    rocksdb_writebatch_t* batch, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_merge(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_writeoptions_t* options,
+    const char* key, size_t klen, const char* val, size_t vlen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_merge_cf(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key, size_t klen,
+    const char* val, size_t vlen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_delete(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_writeoptions_t* options,
+    const char* key, size_t klen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_delete_cf(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_flush_wal(
+    rocksdb_transactiondb_t* txn_db, unsigned char sync, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_flush(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_flushoptions_t* options,
+    char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_transactiondb_flush_cf(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_flushoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, char** errptr);

c_api_gen/c_generated_writebatch_subset.cc.inc

@@ -0,0 +1,58 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/generate_c_api.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Source: tools/c_api_gen/spec.json
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/generate_c_api.py
+//     --spec tools/c_api_gen/spec.json --section 'WriteBatch'
+//     --header-out
+//     c_api_gen/c_generated_writebatch_subset.h.inc
+//     --source-out
+//     c_api_gen/c_generated_writebatch_subset.cc.inc
+
+/* WriteBatch */
+
+void rocksdb_writebatch_clear(rocksdb_writebatch_t* b) { b->rep.Clear(); }
+
+void rocksdb_writebatch_put(rocksdb_writebatch_t* b, const char* key,
+                            size_t klen, const char* val, size_t vlen) {
+  b->rep.Put(Slice(key, klen), Slice(val, vlen));
+}
+
+void rocksdb_writebatch_put_cf(rocksdb_writebatch_t* b,
+                               rocksdb_column_family_handle_t* column_family,
+                               const char* key, size_t klen, const char* val,
+                               size_t vlen) {
+  b->rep.Put(column_family->rep, Slice(key, klen), Slice(val, vlen));
+}
+
+void rocksdb_writebatch_delete(rocksdb_writebatch_t* b, const char* key,
+                               size_t klen) {
+  b->rep.Delete(Slice(key, klen));
+}
+
+void rocksdb_writebatch_put_log_data(rocksdb_writebatch_t* b, const char* blob,
+                                     size_t len) {
+  b->rep.PutLogData(Slice(blob, len));
+}
+
+void rocksdb_writebatch_set_save_point(rocksdb_writebatch_t* b) {
+  b->rep.SetSavePoint();
+}
+
+void rocksdb_writebatch_rollback_to_save_point(rocksdb_writebatch_t* b,
+                                               char** errptr) {
+  SaveError(errptr, b->rep.RollbackToSavePoint());
+}
+
+void rocksdb_writebatch_pop_save_point(rocksdb_writebatch_t* b, char** errptr) {
+  SaveError(errptr, b->rep.PopSavePoint());
+}
+
+void rocksdb_writebatch_verify_checksum(rocksdb_writebatch_t* b,
+                                        char** errptr) {
+  SaveError(errptr, b->rep.VerifyChecksum());
+}

c_api_gen/c_generated_writebatch_subset.h.inc

@@ -0,0 +1,47 @@
+// @generated
+// -----------------------------------------------------------------------------
+// Auto-generated by tools/c_api_gen/generate_c_api.py.
+// DO NOT EDIT THIS FILE DIRECTLY.
+// Source: tools/c_api_gen/spec.json
+// -----------------------------------------------------------------------------
+
+// To regenerate this file:
+//   python3 tools/c_api_gen/generate_c_api.py
+//     --spec tools/c_api_gen/spec.json --section 'WriteBatch'
+//     --header-out
+//     c_api_gen/c_generated_writebatch_subset.h.inc
+//     --source-out
+//     c_api_gen/c_generated_writebatch_subset.cc.inc
+
+/* WriteBatch */
+
+extern ROCKSDB_LIBRARY_API void rocksdb_writebatch_clear(
+    rocksdb_writebatch_t* b);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_writebatch_put(rocksdb_writebatch_t* b,
+                                                       const char* key,
+                                                       size_t klen,
+                                                       const char* val,
+                                                       size_t vlen);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_writebatch_put_cf(
+    rocksdb_writebatch_t* b, rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t klen, const char* val, size_t vlen);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_writebatch_delete(
+    rocksdb_writebatch_t* b, const char* key, size_t klen);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_writebatch_put_log_data(
+    rocksdb_writebatch_t* b, const char* blob, size_t len);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_writebatch_set_save_point(
+    rocksdb_writebatch_t* b);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_writebatch_rollback_to_save_point(
+    rocksdb_writebatch_t* b, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_writebatch_pop_save_point(
+    rocksdb_writebatch_t* b, char** errptr);
+
+extern ROCKSDB_LIBRARY_API void rocksdb_writebatch_verify_checksum(
+    rocksdb_writebatch_t* b, char** errptr);

cache/cache_key.h

@@ -127,6 +127,24 @@ class OffsetableCacheKey : private CacheKey {
     return CacheKey(file_num_etc64_, offset_etc64_ ^ offset);
   }
 
+  // Construct a CacheKey for a record located at byte `offset` within a file in
+  // which every distinct cached region is at least 5 bytes long. The low two
+  // offset bits are dropped, which is collision-free under that >= 5-byte (>= 4
+  // would suffice) guarantee and lets different kinds of records in the same
+  // file share a single keyspace without colliding.
+  //
+  // This is the canonical cache-key scheme for byte-offset records that carry
+  // the block-based "min 5 byte" guarantee: block-based SST blocks (see
+  // BlockBasedTable::GetCacheKey) and SimpleGen2Blob records (which always
+  // include a >= 5-byte trailer; see db/blob/blob_gen2_format.h). Because both
+  // use this same function, an SST's data blocks and its embedded blob records
+  // never collide even when the block cache and blob cache are the same cache.
+  // Keeping a single implementation here avoids a divergence bug that would
+  // silently alias the two keyspaces.
+  inline CacheKey WithOffsetForMinSizeRecord(uint64_t offset) const {
+    return WithOffset(offset >> 2);
+  }
+
   // The "common prefix" is a shared prefix for all the returned CacheKeys.
   // It is specific to the file but the same for all offsets within the file.
   static constexpr size_t kCommonPrefixSize = 8;

cache/clock_cache.cc

@@ -3108,7 +3108,7 @@ AutoHyperClockTable::HandleImpl* AutoHyperClockTable::Lookup(
   HandleImpl* const arr = array_.Get();
   NextWithShift next_with_shift = arr[home].head_next_with_shift.LoadRelaxed();
   for (size_t i = 0; !next_with_shift.IsEnd() && i < 10; ++i) {
-    HandleImpl* h = &arr[next_with_shift.IsEnd()];
+    HandleImpl* h = &arr[next_with_shift.GetNext()];
     // Attempt cheap key match without acquiring a read ref. This could give a
     // false positive, which is re-checked after acquiring read ref, or false
     // negative, which is re-checked in the full Lookup. Also, this is a

claude_md/add_option.md

@@ -46,7 +46,7 @@ This document provides guidance on how to add new options to RocksDB's public AP
 
 ## Pattern 1: Adding a Standard Column Family Option
 
-Example reference: commit `94e65a2e0b4f817aa4bfa4c96cdf867e7980d7bc` (memtable_veirfy_per_key_checksum_on_seek)
+Example reference: commit `94e65a2e0b4f817aa4bfa4c96cdf867e7980d7bc` (memtable_verify_per_key_checksum_on_seek)
 
 ### Step 1: Define the Option in Public Header
 
@@ -63,7 +63,7 @@ Add the option with documentation in `AdvancedColumnFamilyOptions` struct:
 // operation.
 // This option depends on memtable_protection_bytes_per_key to be non zero.
 // If memtable_protection_bytes_per_key is zero, no validation is performed.
-bool memtable_veirfy_per_key_checksum_on_seek = false;
+bool memtable_verify_per_key_checksum_on_seek = false;
 ```
 
 ### Step 2: Add to Internal Options Structs
@@ -74,14 +74,14 @@ Add to `MutableCFOptions` struct (or `ImmutableCFOptions` for immutable options)
 
 ```cpp
 // In MutableCFOptions constructor from Options:
-memtable_veirfy_per_key_checksum_on_seek(
-    options.memtable_veirfy_per_key_checksum_on_seek),
+memtable_verify_per_key_checksum_on_seek(
+    options.memtable_verify_per_key_checksum_on_seek),
 
 // In MutableCFOptions default constructor:
-memtable_veirfy_per_key_checksum_on_seek(false),
+memtable_verify_per_key_checksum_on_seek(false),
 
 // In MutableCFOptions struct member declarations:
-bool memtable_veirfy_per_key_checksum_on_seek;
+bool memtable_verify_per_key_checksum_on_seek;
 ```
 
 ### Step 3: Register for Serialization/Deserialization
@@ -91,9 +91,9 @@ bool memtable_veirfy_per_key_checksum_on_seek;
 Add to the options type info map for serialization:
 
 ```cpp
-{"memtable_veirfy_per_key_checksum_on_seek",
+{"memtable_verify_per_key_checksum_on_seek",
  {offsetof(struct MutableCFOptions,
-           memtable_veirfy_per_key_checksum_on_seek),
+           memtable_verify_per_key_checksum_on_seek),
   OptionType::kBoolean, OptionVerificationType::kNormal,
   OptionTypeFlags::kMutable}},
 ```
@@ -101,8 +101,8 @@ Add to the options type info map for serialization:
 Add logging in `MutableCFOptions::Dump()`:
 
 ```cpp
-ROCKS_LOG_INFO(log, "memtable_veirfy_per_key_checksum_on_seek: %d",
-               memtable_veirfy_per_key_checksum_on_seek);
+ROCKS_LOG_INFO(log, "memtable_verify_per_key_checksum_on_seek: %d",
+               memtable_verify_per_key_checksum_on_seek);
 ```
 
 ### Step 4: Update Options Helper
@@ -112,8 +112,8 @@ ROCKS_LOG_INFO(log, "memtable_veirfy_per_key_checksum_on_seek: %d",
 Add to `UpdateColumnFamilyOptions()`:
 
 ```cpp
-cf_opts->memtable_veirfy_per_key_checksum_on_seek =
-    moptions.memtable_veirfy_per_key_checksum_on_seek;
+cf_opts->memtable_verify_per_key_checksum_on_seek =
+    moptions.memtable_verify_per_key_checksum_on_seek;
 ```
 
 ### Step 5: Add to Options Settable Test
@@ -123,7 +123,7 @@ cf_opts->memtable_veirfy_per_key_checksum_on_seek =
 Add to the test string in `ColumnFamilyOptionsAllFieldsSettable`:
 
 ```cpp
-"memtable_veirfy_per_key_checksum_on_seek=1;"
+"memtable_verify_per_key_checksum_on_seek=1;"
 ```
 
 ### Step 6: Add db_stress Support
@@ -131,23 +131,23 @@ Add to the test string in `ColumnFamilyOptionsAllFieldsSettable`:
 **File: `db_stress_tool/db_stress_common.h`**
 
 ```cpp
-DECLARE_bool(memtable_veirfy_per_key_checksum_on_seek);
+DECLARE_bool(memtable_verify_per_key_checksum_on_seek);
 ```
 
 **File: `db_stress_tool/db_stress_gflags.cc`**
 
 ```cpp
 DEFINE_bool(
-    memtable_veirfy_per_key_checksum_on_seek,
-    ROCKSDB_NAMESPACE::Options().memtable_veirfy_per_key_checksum_on_seek,
-    "Sets CF option memtable_veirfy_per_key_checksum_on_seek.");
+    memtable_verify_per_key_checksum_on_seek,
+    ROCKSDB_NAMESPACE::Options().memtable_verify_per_key_checksum_on_seek,
+    "Sets CF option memtable_verify_per_key_checksum_on_seek.");
 ```
 
 **File: `db_stress_tool/db_stress_test_base.cc`**
 
 ```cpp
-options.memtable_veirfy_per_key_checksum_on_seek =
-    FLAGS_memtable_veirfy_per_key_checksum_on_seek;
+options.memtable_verify_per_key_checksum_on_seek =
+    FLAGS_memtable_verify_per_key_checksum_on_seek;
 ```
 
 ### Step 7: Add db_bench Support
@@ -156,12 +156,12 @@ options.memtable_veirfy_per_key_checksum_on_seek =
 
 ```cpp
 // Flag definition (near related flags):
-DEFINE_bool(memtable_veirfy_per_key_checksum_on_seek, false,
-            "Sets CF option memtable_veirfy_per_key_checksum_on_seek");
+DEFINE_bool(memtable_verify_per_key_checksum_on_seek, false,
+            "Sets CF option memtable_verify_per_key_checksum_on_seek");
 
 // Apply flag to options (in InitializeOptionsFromFlags or similar):
-options.memtable_veirfy_per_key_checksum_on_seek =
-    FLAGS_memtable_veirfy_per_key_checksum_on_seek;
+options.memtable_verify_per_key_checksum_on_seek =
+    FLAGS_memtable_verify_per_key_checksum_on_seek;
 ```
 
 ### Step 8: Add Crash Test Support
@@ -169,7 +169,7 @@ options.memtable_veirfy_per_key_checksum_on_seek =
 **File: `tools/db_crashtest.py`**
 
 ```python
-"memtable_veirfy_per_key_checksum_on_seek": lambda: random.choice([0] * 7 + [1]),
+"memtable_verify_per_key_checksum_on_seek": lambda: random.choice([0] * 7 + [1]),
 ```
 
 Also add constraint handling in `finalize_and_sanitize()` if needed:
@@ -177,7 +177,7 @@ Also add constraint handling in `finalize_and_sanitize()` if needed:
 ```python
 # only skip list memtable representation supports paranoid memory checks
 if dest_params.get("memtablerep") != "skip_list":
-    dest_params["memtable_veirfy_per_key_checksum_on_seek"] = 0
+    dest_params["memtable_verify_per_key_checksum_on_seek"] = 0
 ```
 
 ### Step 9: Add Release Note
@@ -185,7 +185,7 @@ if dest_params.get("memtablerep") != "skip_list":
 **File: `unreleased_history/new_features/<descriptive_name>.md`**
 
 ```markdown
-A new flag memtable_veirfy_per_key_checksum_on_seek is added to AdvancedColumnFamilyOptions. When it is enabled, it will validate key checksum along the binary search path on skiplist based memtable during seek operation.
+A new flag memtable_verify_per_key_checksum_on_seek is added to AdvancedColumnFamilyOptions. When it is enabled, it will validate key checksum along the binary search path on skiplist based memtable during seek operation.
 ```
 
 ---
@@ -421,6 +421,30 @@ Example reference: commit `429b36c22d76403d275dd0e6877b08d4cea2bc90` (block_alig
 
 If an option already exists but needs C API support:
 
+**First decide whether the option is auto-managed.**
+
+Many simple option fields are no longer maintained by hand in
+`include/rocksdb/c.h` and `db/c.cc`. Before adding a manual wrapper:
+
+- Check whether the option belongs to a family handled by
+  `tools/c_api_gen/auto_simple_bindings.py`.
+- If it is a simple scalar/enum/string field in an auto-managed family, run
+  `python3 tools/c_api_gen/regen_all.py` and let the generator add the C API.
+- After regenerating, run
+  `python3 tools/c_api_gen/verify_generated_up_to_date.py` to confirm the
+  checked-in generated fragments stay stable.
+- Do not edit generated `.inc` files under `c_api_gen/` by hand.
+- If the new field is not ready for C API support yet, add an entry to
+  `tools/c_api_gen/auto_simple_bindings_blocklist.json` with
+  `"policy": "deferred"` and a concrete reason so the build stays intentional
+  rather than silently drifting.
+- If the field needs custom ownership, callback, vector, or nested-struct
+  marshalling, keep it manual or move it into `tools/c_api_gen/spec.json`
+  depending on the API shape.
+
+If the option is not auto-managed, or the C API really is a custom/manual case,
+the traditional hand-written pattern still applies:
+
 **File: `db/c.cc`**
 
 ```cpp
@@ -509,4 +533,3 @@ Common option types used in serialization:
 - [ ] unreleased_history markdown file
 - [ ] Java API (for BlockBasedTableOptions)
 - [ ] C API (if needed)
-

claude_md/add_public_api.md

@@ -168,9 +168,18 @@ Status YourNewAPI(const YourAPIOptions& /*options*/,
 
 ### Step 5: Add C API Bindings
 
-**Header:** `include/rocksdb/c.h`
+> **Important:** `include/rocksdb/c.h` and `db/c.cc` are `@generated` and must
+> NOT be edited by hand — your changes would be overwritten on the next
+> regeneration. Choose the right path first (see "Before writing C API code by
+> hand" below). For hand-written (manual) wrappers, edit the source templates
+> `tools/c_api_gen/c_base.h` (declarations) and `tools/c_api_gen/c_base.cc`
+> (implementations), then run `python3 tools/c_api_gen/regen_all.py` to produce
+> `c.h` / `c.cc`. The snippets below show the *declaration* and *implementation*
+> you would add to those templates.
 
-\`\`\`c
+**Header (declaration → `tools/c_api_gen/c_base.h`):**
+
+```c
 // Basic version
 extern ROCKSDB_LIBRARY_API void rocksdb_your_new_api(
     rocksdb_t* db,
@@ -191,7 +200,7 @@ extern ROCKSDB_LIBRARY_API void rocksdb_your_new_api_opt(
     char** errptr);
 \`\`\`
 
-**Implementation:** `db/c.cc`
+**Implementation (→ `tools/c_api_gen/c_base.cc`):**
 
 \`\`\`cpp
 void rocksdb_your_new_api(rocksdb_t* db, const char* start_key,
@@ -217,6 +226,38 @@ void rocksdb_your_new_api_cf(rocksdb_t* db,
 }
 \`\`\`
 
+**Before writing C API code by hand, choose the right path:**
+
+- **Auto-managed simple struct fields:** If you added a new field to a managed
+  public struct such as `ReadOptions`, selected option structs, or managed
+  metadata structs, first check `tools/c_api_gen/auto_simple_bindings.py`.
+  Simple scalar/enum/string fields should be regenerated with
+  `python3 tools/c_api_gen/regen_all.py` rather than hand-editing
+  `include/rocksdb/c.h` / `db/c.cc`.
+  After regenerating, run
+  `python3 tools/c_api_gen/verify_generated_up_to_date.py` to confirm the
+  checked-in generated fragments are stable.
+- **Spec-driven wrappers:** If the API is still mechanically generated but needs
+  an explicit C shape, naming, or `Status`/`char** errptr` policy, add it to
+  `tools/c_api_gen/spec.json`.
+- **Fully manual wrappers:** If the API uses callbacks, ownership transfer,
+  vectors/maps, open flows, or otherwise irregular marshalling, keep the C API
+  hand-written.
+
+**Temporary deferral for auto-managed families:**
+
+If a new field lands in an auto-managed family but the C API shape is not ready
+yet, add a checked-in entry to
+`tools/c_api_gen/auto_simple_bindings_blocklist.json` with:
+
+- `policy: "manual"` when the field is intentionally outside simple auto-gen
+- `policy: "deferred"` when the field should be revisited later
+- a concrete `reason`
+- `tracking_issue` when available
+
+If a field in an auto-managed family is not supported by the generator and is
+not covered by the blocklist, regeneration should fail. That is intentional.
+
 **If you have options, also add:**
 
 \`\`\`cpp
@@ -482,8 +523,8 @@ public class YourAPIOptionsTest {
 | StackableDB | `include/rocksdb/utilities/stackable_db.h` | ✓ |
 | Secondary DB | `db/db_impl/db_impl_secondary.h` | If not supported |
 | Compacted DB | `db/db_impl/compacted_db_impl.h` | If not supported |
-| C API Header | `include/rocksdb/c.h` | ✓ |
-| C API Implementation | `db/c.cc` | ✓ |
+| C API Header (manual wrappers) | `tools/c_api_gen/c_base.h` → regen → `include/rocksdb/c.h` (`@generated`) | ✓ |
+| C API Implementation (manual wrappers) | `tools/c_api_gen/c_base.cc` → regen → `db/c.cc` (`@generated`) | ✓ |
 | Java API | `java/src/main/java/org/rocksdb/RocksDB.java` | ✓ |
 | Java Options | `java/src/main/java/org/rocksdb/YourAPIOptions.java` | If needed |
 | JNI Implementation | `java/rocksjni/rocksjni.cc` | ✓ |

claude_md/ci_complexity_eval_prompt.md

@@ -0,0 +1,61 @@
+# PR Complexity Classifier — CI Triage Prompt
+
+## Purpose
+Quickly classify a RocksDB pull request as **simple** or **complex** so the
+downstream reviewer can pick an appropriate thinking budget.
+
+## Output Contract (STRICT)
+Your final response MUST be exactly one of these two lowercase tokens, on a
+line by itself, with no other text:
+
+```
+simple
+```
+
+or
+
+```
+complex
+```
+
+If you are unsure, output `complex`. The downstream review will use this token
+to set its thinking budget — when in doubt, prefer the more thorough budget.
+
+## Classification Heuristics
+
+Treat a PR as **simple** when ALL of the following hold:
+- Diff touches < ~200 lines of non-test code
+- No changes to public headers under `include/rocksdb/`
+- No changes to on-disk format (SST block layout, WAL record layout,
+  manifest, options file format, version edit encoding)
+- No changes to compaction picker, write-path locking, recovery, or
+  WAL/memtable lifecycle
+- No new public option, no new statistics counter, no new RPC/serialized
+  message
+- Tests-only or docs-only changes
+- Trivial refactors (rename, comment, dead-code removal, lint fixes)
+- One-line bug fix with obvious root cause and a focused regression test
+
+Treat a PR as **complex** when ANY of the following hold:
+- Touches files in `db/`, `db/compaction/`, `db/blob/`, `cache/`,
+  `table/block_based/`, `file/`, `env/io_*` with > ~50 lines of logic change
+- Modifies thread-safety/synchronization (mutexes, atomics, memory ordering,
+  refcounting, condvars)
+- Changes serialization or on-disk format
+- Changes public API surface in `include/rocksdb/`
+- Adds a new feature flag, option, or compaction style
+- Touches transactions (`utilities/transactions/`), backup/checkpoint,
+  user-defined timestamps, or remote compaction
+- Cross-cutting refactor spanning > 5 directories
+
+## How To Decide
+1. Read the diff summary and changed-file list provided below.
+2. Skim the diff hunks — focus on which subsystems are touched and whether
+   the change is mechanical vs. semantic.
+3. Apply the heuristics above. Default to `complex` if the rules conflict.
+4. Output the single token. No preamble. No explanation.
+
+## Tool Use
+You may use `View`, `GlobTool`, and `GrepTool` to peek at one or two files
+if the diff snippet alone is ambiguous, but keep this fast — this is a
+triage step, not a review. Do NOT spawn sub-agents.

claude_md/ci_query_prompt.md

@@ -0,0 +1,3 @@
+You are an expert C++ engineer for RocksDB.
+Read CLAUDE.md in the repo root and claude_md/ files for project context.
+Answer thoroughly with exact file paths and line numbers.

claude_md/ci_recovery_prompt.md

@@ -0,0 +1,54 @@
+Read review-findings.md. This file contains partial code review findings
+from a session that ran out of turns before finishing. Reformat them into
+a clean final review comment.
+
+## Required Output Structure
+
+Use this exact structure so the PR page stays scrollable:
+
+```markdown
+## Summary
+
+> **Partial review** — the analysis was interrupted before all perspectives
+> were completed. Findings below cover only the perspectives that finished.
+> You can request a fresh full review with `/claude-review`.
+
+<!-- One or two sentences of overall assessment. -->
+
+**High-severity findings (N):**
+- **[file.cc:123]** One-line description. <!-- repeat per HIGH finding -->
+
+<!-- If no HIGH findings were salvaged, write: -->
+<!-- _No high-severity findings recovered._ -->
+
+<details>
+<summary>Recovered findings (click to expand)</summary>
+
+### Findings
+
+#### :red_circle: HIGH
+... (H1, H2, ...)
+
+#### :yellow_circle: MEDIUM
+... (M1, M2, ...)
+
+#### :green_circle: LOW / NIT
+... (L1, L2, ...)
+
+### Cross-Component Analysis
+<!-- Whatever cross-component results were captured. -->
+
+### Positive Observations
+<!-- Optional. -->
+
+</details>
+```
+
+## Rules
+- The `## Summary`, the partial-review notice, and the HIGH bullet list MUST
+  stay outside the `<details>` block.
+- All per-finding detail and analysis MUST live inside the `<details>` block.
+- Do NOT nest `<details>` blocks.
+- Do NOT add any text after the closing `</details>` — the comment-builder
+  appends its own footer.
+- Output ONLY the formatted review. No commentary, no explanation.

claude_md/ci_review_prompt.md

@@ -0,0 +1,661 @@
+# Code Review Workflow — CI Prompt
+
+## Purpose
+Thorough multi-agent code review for RocksDB commits following CLAUDE.md guidelines,
+with structured codebase exploration, inter-agent debate, and verification.
+
+IMPORTANT — Incremental output: After completing EACH major phase (Setup,
+Codebase Context, Initial Review, Debate, Synthesis), append your findings
+to review-findings.md using the Write tool. This ensures partial results are
+saved even if the review is interrupted by the turn limit. After the final
+synthesis, write the complete review to review-findings.md, replacing
+previous content. Also output the final review as your response text.
+
+## Workflow Steps
+
+### 1. Setup Phase
+- Read CLAUDE.md to understand review guidelines
+- Identify the commit/diff/PR to review
+- The PR diff is provided below in this prompt. The repository is already
+  checked out — use local file search tools (Grep, Glob, Read) to explore.
+- Parse the changed files and categorize them (API, core logic, tests, etc.)
+- Read all changed files to build full context
+
+### 2. Codebase Context Phase (CRITICAL — Do Not Skip)
+
+**Why this matters**: Review agents that only see the diff miss systemic bugs.
+For example, a change to an iterator's return value might look correct in
+isolation but break multi-SST iteration 5 layers up the call stack. A change
+to error handling might silently drop errors that a caller 3 levels up relies
+on for recovery. This phase builds the context that prevents those blind spots.
+
+Go deep. Surface-level reading leads to reviews that miss caching layers,
+existing helper functions, or concurrency requirements. The most expensive
+failure mode is findings that look correct in isolation but miss how the
+change breaks the surrounding system.
+
+The team lead spawns one or more research agents to perform the following
+analyses using local file search tools (Grep, Glob, Read). The research
+must be written to `context.md` BEFORE any review agent is spawned.
+
+#### 2a. Subsystem Deep-Read
+Read the changed files AND their surrounding subsystem **in depth** — not just
+signatures, but actual logic, edge cases, data flows, locking protocols,
+concurrency patterns, and interactions between components.
+
+For each changed file, also read:
+- The header that declares its interface (if the change is in a .cc file)
+- The sibling implementation that does the same thing the "standard" way
+  (e.g., for a UDI iterator change, read how `IndexBlockIter` handles the
+  same scenario — it's the reference implementation)
+- The wrapper/adapter layer that sits between this code and its callers
+- Any existing tests to understand what behaviors are currently guaranteed
+
+#### 2b. Caller-Chain Analysis (upstream impact)
+For each changed function/method, trace the call chain UPWARD 3-5 levels to
+understand who consumes the changed behavior and what invariants they rely on.
+**Focus on control flow decisions** — if/else branches, while loop conditions,
+early returns — that depend on the changed return values or state.
+
+**How to do this**:
+- Search for callers of each changed function (search for the function name, the
+  class method, virtual overrides)
+- For each caller, search for ITS callers, repeating 3-5 levels up
+- At each level, read the actual code to understand the control flow decision
+- Stop when you reach a "terminal" caller (e.g., user-facing API, top-level
+  iterator) or when the propagation is clearly safe
+
+#### 2c. Callee-Chain Analysis (downstream dependencies AND side effects)
+For each changed function, trace what it calls and whether those calls have
+new preconditions or changed semantics. **Critically, trace SIDE EFFECTS —
+not just return values.** Many bugs hide in side effects on shared state
+(counters, sequence numbers, flags) that are invisible if you only check
+return values.
+
+**For each callee, ask:**
+1. What does it RETURN? (the obvious check)
+2. What shared state does it MUTATE? (the non-obvious check)
+3. Under what PRECONDITIONS does it operate correctly?
+4. Are those preconditions met in the new calling context?
+
+#### 2d. Sibling Implementation Comparison
+Identify the "reference" or "standard" implementation that does the same thing
+the changed code does, and compare their behavior. This is critical for plugin/
+extension APIs where the implementation must match an implicit contract.
+
+#### 2e. Invariant Documentation
+Document the key invariants that the changed code must maintain. Read existing
+comments, assertions, and test expectations to discover these invariants.
+
+#### 2f. Related Functionality and Existing Conventions
+Identify related functionality that already exists in the codebase:
+- Are there existing helper functions the change should use (or is duplicating)?
+- Are there existing patterns for the same kind of change (e.g., how other
+  Customizable subclasses handle deprecation)?
+- Are there existing tests that test similar scenarios (e.g., multi-SST
+  iteration tests that could be extended)?
+
+#### 2g. Cross-Component Data Consumer Analysis (CRITICAL — common blind spot)
+
+**Why this is separate from caller-chain analysis:** Sections 2b-2c trace
+who CALLS the changed functions. This section traces who CONSUMES the data
+that the changed code PRODUCES. These are often completely different
+subsystems.
+
+When the change writes data to a shared structure (memtable, SST block, cache
+entry, statistics counter), ask: **"Who reads this data, under what rules,
+and do those rules match the writer's assumptions?"**
+
+**How to do this:**
+1. Identify every piece of data the change WRITES (e.g., a range tombstone
+   entry, a counter update, a metadata field)
+2. For each, search the codebase for ALL readers of that data structure
+3. For each reader, check:
+   - Does the reader apply visibility rules (seqno, read_callback, snapshot)?
+   - Are those rules compatible with the writer's assumptions?
+   - Does the reader assume invariants about the data (e.g., seqno ordering,
+     monotonicity) that the writer might violate?
+
+#### 2h. Alternative Execution Contexts
+
+The same code path may run in different contexts with different assumptions.
+**Enumerate all contexts and check each one.** This is where "works in the
+common case but breaks in edge configurations" bugs hide.
+
+For RocksDB, always check whether the changed code interacts differently with:
+
+| Context | Key difference | Common failure mode |
+|---------|---------------|-------------------|
+| WritePreparedTxnDB / WriteUnpreparedTxnDB | `read_callback_` controls visibility, not just seqno | Visibility bypass |
+| ReadOnly DB / SecondaryInstance | No mutable memtable, writes not allowed | Null pointer or no-op needed |
+| CompactionService / Remote compaction | Different process, serialized state | Serialization mismatch |
+| User-defined timestamps | Extra dimension in key comparison and visibility | Wrong ordering |
+| MemPurge | Memtable-to-memtable, not memtable-to-SST | Missing data |
+| Column family with BlobDB | Values may be in blob files, not inline | Missing dereference |
+| Snapshots (old, held long-term) | Snapshot seqno may be far behind current state | Metadata corruption |
+| Concurrent writers (allow_concurrent_memtable_write) | Lock-free paths vs locked paths | Lost updates |
+| FIFO / Universal compaction | Different compaction invariants than Level | Wrong assumptions |
+| Prefix seek / total order seek | Different iterator behavior | Wrong iteration results |
+
+**For each context, ask:**
+1. Does the code execute in this context? (Check constructor, feature flags)
+2. If yes, do the assumptions still hold?
+3. If not, should the feature be disabled in this context?
+
+#### 2i. Assumption Stress-Testing (the "logically X" audit)
+
+When the change claims a property (e.g., "logically redundant," "no-op in
+this case," "safe because X"), **systematically break it:**
+
+1. **State the claim precisely.** E.g., "The inserted range tombstone is
+   logically redundant — it covers only keys already deleted by point
+   tombstones."
+2. **List the preconditions** that must be true for the claim to hold.
+3. **For each precondition, construct a counterexample** — a concrete scenario
+   where the precondition is violated.
+4. **Verify each counterexample against the code.** Does the code guard
+   against it? If not, it's a bug.
+
+**Anti-pattern: treating asserts as proofs.** When you see an assert, do NOT
+conclude "the invariant holds." Instead ask: "Can I construct an input where
+this assert fires?" If yes, it's a bug (the assert catches it in debug but
+release builds silently corrupt). If you cannot construct a counterexample
+after trying, document WHY it's impossible.
+
+#### 2j. Write Context Document
+Write the complete analysis to `context.md` in the review folder. This document
+is provided to ALL review agents as part of their prompt.
+
+**Context document template**:
+```markdown
+# Codebase Context for Review
+
+## How the Relevant Subsystem Works Today
+[Detailed description of the subsystem architecture, data flows,
+and component interactions. Not just "what" but "how" and "why."]
+
+## Changed Functions and Their Call Chains
+
+### function_name() (file:line)
+**What changed**: [brief description of the behavioral change]
+
+**Upstream callers** (who depends on this):
+1. CallerA::method() (file:line) — uses return value to decide X
+2. CallerB::method() (file:line) — passes result to CallerC
+3. CallerC::method() (file:line) — THE CRITICAL DECISION POINT: ...
+
+**Downstream callees** (what this depends on):
+1. calleeA() — behavior unchanged
+2. calleeB() — NEW dependency, requires X
+
+**Sibling implementation** (how the "standard" version handles this):
+- StandardImpl::method() does Y at file:line, which ensures invariant Z
+- The changed code must match or document any deviation
+
+**Key invariants**:
+- Must never return X when Y is true because CallerC will...
+- Must always set Z before returning because CallerB assumes...
+
+## System Architecture Context
+[How the changed components fit into the overall system]
+
+## Known Invariants That Must Be Preserved
+1. ...
+2. ...
+
+## Existing Conventions and Related Code
+- [helper functions, patterns, existing tests that are relevant]
+
+## Cross-Component Data Consumers
+For each piece of data the change WRITES to a shared structure:
+### [data item] written to [structure]
+**Writer assumptions**: [what the writer assumes about how this data is consumed]
+**All readers**:
+1. ReaderA::method() (file:line) — reads under [visibility rules]
+   - Compatible with writer? YES/NO. If NO, explain the mismatch.
+2. ReaderB::method() (file:line) — reads under [different rules]
+   - Compatible with writer? YES/NO.
+
+## Alternative Execution Contexts
+| Context | Does code execute? | Assumptions hold? | Action needed? |
+|---------|-------------------|-------------------|----------------|
+| WritePreparedTxnDB | YES/NO | YES/NO | [disable/guard/safe] |
+| Old snapshots | YES/NO | YES/NO | ... |
+| User-defined timestamps | YES/NO | YES/NO | ... |
+| ReadOnly DB | YES/NO | YES/NO | ... |
+| ... | ... | ... | ... |
+
+## Assumption Stress Test
+### Claim: "[the design claim, e.g., logically redundant]"
+**Preconditions for claim to hold:**
+1. [precondition] — counterexample: [scenario]. Guarded? YES/NO.
+2. [precondition] — counterexample: [scenario]. Guarded? YES/NO.
+
+## Potential Pitfalls
+- [specific scenarios where the change could interact badly with
+  the surrounding system, identified from the caller-chain analysis]
+```
+
+### 3. Initial Review Phase (Parallel)
+Create a team and spawn 5 review agents in parallel. **Include the context
+document from Phase 2 in each agent's prompt.** Each agent writes findings
+to its own file and sends a summary message to the team lead.
+
+Each agent's prompt should include:
+```
+## Codebase Context (READ THIS FIRST)
+[paste or reference the context.md file]
+
+You MUST consider how your findings interact with the upstream callers
+and system invariants documented above. A finding that looks correct
+in isolation may be a critical bug when you consider the full call chain.
+```
+
+#### Agent: Design & Approach Reviewer
+- Read the commit message deeply to understand the problem being solved
+- Analyze whether this is the right approach to the problem
+- Consider alternative designs and trade-offs:
+  - Are there simpler solutions that achieve the same goal?
+  - Does the approach introduce unnecessary complexity?
+  - Would a different data structure or algorithm be more appropriate?
+- Evaluate whether the change follows existing architectural patterns or
+  introduces new patterns that may be inconsistent
+- Assess the scope: is the change too large? Should it be split into
+  smaller, independently reviewable pieces?
+- Check if the approach has precedent in the codebase or in the academic
+  literature it references (e.g., SuRF paper)
+
+#### Agent: Correctness Reviewer
+- Thread safety and concurrency analysis
+- Error handling and propagation via Status type
+- Edge cases (empty inputs, overflow, boundary conditions)
+- Data corruption scenarios
+- Logic correctness of new algorithms
+- **Behavioral contract changes**: Do return value semantics match what
+  upstream callers expect? (Use the caller-chain analysis from context.md)
+- **Callee side effects**: Does the change call functions that mutate shared
+  state (counters, seqnos, metadata)? Are the mutations correct for the new
+  calling context? (Use the callee-chain analysis from context.md)
+
+#### Agent: Cross-Component & Adversarial Reviewer
+
+This agent exists because the most critical bugs hide at component boundaries,
+not within a single component. It uses a fundamentally different methodology
+than the correctness reviewer: instead of verifying "does the code do what it
+intends?", it asks "what breaks when we change the assumptions?"
+
+**Data-flow analysis:** Perform the cross-component data consumer analysis
+described in section 2g. For every piece of data the change WRITES to a shared
+structure, trace ALL READERS and verify their visibility rules are compatible
+with the writer's assumptions. This is a DATA-FLOW question, not a
+CONTROL-FLOW question — readers may be in completely different subsystems.
+
+**Alternative execution contexts:** Use the canonical table from section 2h.
+Enumerate all contexts where the changed code executes and verify assumptions
+hold in each.
+
+**Assumption stress-testing and assert-breaking:** Follow the methodology from
+section 2i. Identify every design claim, enumerate preconditions, construct
+counterexamples. Treat asserts as hypotheses to break, not proofs.
+**Red flag words**: "logically redundant", "safe because", "no-op",
+"always true", "cannot happen", "invariant holds"
+
+**Callee side-effect audit:** Perform the callee side-effect audit described
+in section 2c. For every callee, ask what it RETURNS and what it MUTATES.
+
+Write findings to `findings-cross-component.md`.
+
+#### Agent: Invariant Adversary
+
+This agent does NOT read the diff in detail. It takes the feature summary and
+systematically tries to break it. While other agents verify "does the code do
+what it says?", this agent asks "what existing system invariants does this
+feature violate?" and "under what inputs do the design claims fail?"
+
+**Steps 1-4: Assumption stress-testing.** Follow the methodology from section
+2i: extract design claims, enumerate preconditions, construct counterexamples,
+and attempt to break every assert. Be exhaustive about input parameter ranges,
+shared state configurations, and concurrent interleaving.
+
+**Step 5: Callee side-effect audit.** Perform the callee side-effect audit
+described in section 2c. For every function the changed code calls, list ALL
+mutations to shared state (atomic CAS loops, counter increments, flag/metadata
+updates, cache invalidation). A function returning `Status::OK()` does NOT
+mean its side effects are correct.
+
+Write findings to `findings-invariant-adversary.md`.
+
+#### Agent: Caller-Context Auditor
+
+This agent traces BACKWARD from every entry point the changed code hooks into.
+While other agents read the changed code forward ("what does it do?"), this
+agent enumerates WHO invokes it and WITH WHAT parameter ranges.
+
+The key insight: a function that is correct for all *typical* inputs may be
+catastrophically wrong for *atypical-but-reachable* inputs. This agent's job
+is to find the atypical inputs.
+
+**Step 1: Identify entry points.** List every function/constructor/method in
+the changed code that receives external input (parameters, config values,
+pointers to shared state). Include:
+- Constructor parameters (e.g., `active_mem`, `read_callback`, `sequence`)
+- Virtual method overrides that callers invoke polymorphically
+- Functions called from multiple subsystems
+
+**Step 2: Enumerate all callers (3-5 levels up).** For each entry point,
+search the codebase for ALL callers. Do NOT stop at the first caller — trace
+the full call chain. Use `Grep` and `Glob` to find:
+- Direct callers of the function
+- Callers of wrapper/adapter functions that forward to it
+- Factory functions that construct the object
+- Virtual dispatch sites (search for the base class method name)
+
+**Step 3: Parameter range analysis.** For each caller, determine:
+- What values does it pass for each parameter?
+- Under what conditions does it call this function?
+- Are there callers that pass unusual values? (e.g., `kMaxSequenceNumber`,
+  `nullptr`, old snapshot seqnos, non-null `read_callback_`)
+
+Build a parameter range table for each entry point, listing all callers
+and the values they pass for each parameter.
+
+**Step 4: Configuration matrix.** Enumerate option/config combinations that
+affect the changed code path:
+- Which options enable/disable the feature?
+- Which options change the execution context? (e.g.,
+  `allow_concurrent_memtable_write`, `use_trie_index`)
+- Are there option combinations that create contradictions?
+
+**Step 5: Cross-reference with context.md.** Compare your caller analysis
+with the invariants and execution contexts documented in context.md. Flag
+any caller that violates an assumed precondition.
+
+Write findings to `findings-caller-audit.md`.
+
+#### Agent: Performance Reviewer
+- Memory allocation on hot paths
+- Unnecessary copies (string, buffer)
+- Cache efficiency and data locality
+- Loop optimization opportunities
+- Branch prediction (LIKELY/UNLIKELY)
+- Zero-overhead design verification
+
+#### Agent: API & Compatibility Reviewer
+- Public API backwards compatibility
+- API consistency with existing patterns
+- Documentation completeness
+- Deprecation policy compliance
+- Forward compatibility (struct extensibility)
+- **Behavioral compatibility**: Do changed semantics (e.g., kOutOfBound vs
+  kUnknown) break implicit contracts with callers?
+
+#### Agent: Serialization & Deserialization Reviewer
+- Format correctness and versioning
+- Alignment requirements
+- Integer overflow in size computations
+- Bounds checking on untrusted data
+- Crafted input attack vectors
+
+#### Agent: Test Coverage Reviewer
+- Edge case coverage
+- Failure mode testing (corruption, truncation)
+- Round-trip testing (build → serialize → deserialize → verify)
+- Integration test coverage
+- **System-level test coverage**: Are the caller-chain interactions tested?
+  (e.g., multi-SST iteration with UDI, level compaction with new behavior)
+- Test quality (no flaky patterns, good assertions)
+
+### 4. Round-Robin Debate Phase
+After all 5 agents complete their initial review, the team lead orchestrates
+a structured debate:
+
+#### Step 4a: Share findings
+- Team lead sends each agent a summary of ALL other agents' findings
+  (or instructs each agent to read the other agents' findings files)
+
+#### Step 4b: Critique round (2 rounds)
+Each agent reviews the findings from the other agents and sends messages to
+challenge, support, or refine them:
+
+- **Challenge**: "I disagree with [agent] [finding] because [evidence]."
+- **Support**: "I independently found the same issue — this confirms it."
+- **Refine**: "[Finding A] is actually a consequence of [Finding B].
+  The root cause is the same."
+
+The debate assignment follows a round-robin pattern:
+```
+correctness-reviewer      →  critiques  →  invariant-adversary, serialization
+cross-component-reviewer  →  critiques  →  correctness, caller-audit
+invariant-adversary       →  critiques  →  correctness, cross-component
+caller-audit              →  critiques  →  invariant-adversary, cross-component
+performance-reviewer      →  critiques  →  api, cross-component
+api-reviewer              →  critiques  →  correctness, performance
+serialization-reviewer    →  critiques  →  correctness, invariant-adversary
+test-reviewer             →  critiques  →  serialization, caller-audit
+design-reviewer           →  critiques  →  cross-component, invariant-adversary
+```
+
+**Note:** The invariant-adversary and caller-audit agents are deliberately
+cross-linked with correctness and cross-component because their findings
+often reveal the same underlying bug from different angles (invariant
+violation vs reachable bad input vs data-flow mismatch).
+
+Each agent should:
+1. Read the assigned agents' findings files
+2. Send a critique message to each assigned agent
+3. Respond to critiques received from other agents
+4. Update their own findings file with any revisions (upgraded/downgraded severity,
+   withdrawn findings, new findings inspired by others)
+
+#### Step 4c: Team lead collects debate results
+- Read all updated findings files
+- Review the debate messages
+- Build consensus: findings supported by 2+ agents → HIGH confidence
+- Write consensus document with final severity classifications
+
+### 5. Consensus Phase
+Team lead synthesizes the debate into a consensus document:
+- Cross-reference overlapping findings across agents
+- Count agreement (majority = 2+ agents independently flagging or supporting)
+- Note disagreements and how they were resolved
+- Classify findings as HIGH/MEDIUM/LOW severity
+- Write consensus document
+
+### 6. Summary Phase
+- Compile all validated findings ordered by severity
+- Include detailed bug analysis (root cause, vulnerable code paths)
+- Include suggested fixes
+- Note which findings were debated and the outcome
+- Write the complete review to review-findings.md and output as response
+
+**Final report quality rules:**
+- The final report must be CLEAN and POLISHED. No stream-of-consciousness,
+  no "Wait, actually...", no "on closer re-examination". Those belong in
+  the working notes, not the output.
+- If a finding was raised during review but later disproven during debate
+  or deeper analysis, REMOVE IT entirely from the final report. Do not
+  include it with a retraction — just drop it.
+- If a finding was downgraded (e.g., Critical → Suggestion), present it
+  at the final severity only. Do not narrate the severity change.
+- Each finding in the final report should read as a confident, verified
+  conclusion — not a record of the analysis process.
+- The reader should never see the reviewer arguing with itself.
+
+**REQUIRED output structure (so the PR page stays scrollable):**
+
+The final response (and contents of `review-findings.md`) MUST follow this
+exact structure. The summary appears first so reviewers can see HIGH findings
+at a glance; everything else is hidden behind a `<details>` block.
+
+```markdown
+## Summary
+
+<!-- One or two sentences of overall assessment. -->
+
+**High-severity findings (N):**
+- **[file.cc:123]** One-line description of the issue. <!-- repeat per HIGH finding -->
+
+<!-- If there are NO high-severity findings, write exactly: -->
+<!-- _No high-severity findings._ -->
+
+<details>
+<summary>Full review (click to expand)</summary>
+
+### Findings
+
+#### :red_circle: HIGH
+
+##### H1. <Title> — `file.cc:123`
+- **Issue:** ...
+- **Root cause:** ...
+- **Suggested fix:** ...
+
+#### :yellow_circle: MEDIUM
+... (same structure: M1, M2, ...)
+
+#### :green_circle: LOW / NIT
+... (same structure: L1, L2, ...)
+
+### Cross-Component Analysis
+<!-- Execution-context table and assumption stress-test results. -->
+
+### Positive Observations
+<!-- Optional: good patterns, clever optimizations. -->
+
+</details>
+```
+
+Rules for this structure:
+- The top-level `## Summary` and the bullet list of HIGH findings MUST stay
+  outside the `<details>` block — they are always visible.
+- Every detail (per-finding root cause, fix, debate outcomes, cross-component
+  analysis, positive observations) MUST live inside the `<details>` block.
+- Do NOT nest a `<details>` inside another `<details>`.
+- Do NOT add any text after the closing `</details>` — the comment-builder
+  appends its own footer.
+- If the review is partial (recovery path), still produce the summary block
+  first, and put whatever was salvaged inside the `<details>` block.
+
+## Review Checklist
+
+### Context Phase (must be completed before agents spawn)
+- [ ] Subsystem deep-read — read changed files AND surrounding subsystem
+      in depth (logic, edge cases, data flows, locking, concurrency)
+- [ ] Caller chain for every changed public/virtual method (3-5 levels up)
+- [ ] Critical decision points where callers branch on changed behavior
+- [ ] Callee chain with SIDE EFFECTS — downstream dependencies, new
+      preconditions, AND mutations to shared state (section 2c)
+- [ ] Sibling implementations — how does the standard version handle same scenarios?
+- [ ] Invariants that callers rely on
+- [ ] Related functionality — existing helpers, patterns, conventions
+- [ ] Cross-component data consumers — for every data WRITTEN, find ALL readers
+      and verify visibility rules match (section 2g)
+- [ ] Alternative execution contexts verified (section 2h table)
+- [ ] Assumption stress-test — for every design claim, list preconditions
+      and construct counterexamples (section 2i)
+- [ ] Multi-component interactions (compaction, recovery, snapshots, iterators)
+- [ ] Configuration dependencies and unexpected option combinations
+- [ ] Potential pitfalls from caller-chain analysis
+
+### Review Phase (verified by agents)
+- [ ] Database semantics preserved (snapshot isolation, key ordering)
+- [ ] All error cases handled
+- [ ] Thread-safe with correct synchronization
+- [ ] No data races or deadlocks
+- [ ] Appropriate test coverage (edge cases, failure modes, system-level)
+- [ ] No unnecessary allocations or copies in hot paths
+- [ ] Backwards compatible
+- [ ] New APIs consistent with existing patterns and documented
+- [ ] Code follows RocksDB style conventions
+- [ ] Behavioral contracts with upstream callers preserved
+- [ ] All callers enumerated with parameter range table
+
+## Output Structure
+Write all review artifacts to the working directory root:
+- `review-findings.md` — Incremental findings (appended after each phase),
+  then replaced with the final synthesized review at the end
+- `context.md` — Codebase context (call chains, invariants)
+- `findings-*.md` — Per-agent findings (design, correctness, cross-component,
+  invariant-adversary, caller-audit, performance, api, serialization, tests)
+- `consensus.md` — Cross-review consensus (post-debate)
+
+
+
+## Team Structure
+```
+Team Lead (you)
+│
+├── Phase 2: Codebase Context (team lead or dedicated research agent)
+│   └── context-researcher    (general-purpose agent)
+│       ├── Trace caller chains (3-5 levels up)
+│       ├── Trace callee chains (dependencies AND side effects)
+│       ├── Trace data consumers (who reads what the change writes?)
+│       └── Document invariants
+│
+├── Phase 3: Initial Review (parallel, run_in_background)
+│   │  (all agents receive context.md in their prompt)
+│   ├── design-reviewer            (general-purpose agent)
+│   ├── correctness-reviewer       (general-purpose agent)
+│   ├── cross-component-reviewer   (general-purpose agent)
+│   ├── invariant-adversary        (general-purpose agent)  ← NEW
+│   ├── caller-audit               (general-purpose agent)  ← NEW
+│   ├── performance-reviewer       (general-purpose agent)
+│   ├── api-reviewer               (general-purpose agent)
+│   ├── serialization-reviewer     (general-purpose agent)
+│   └── test-reviewer              (general-purpose agent)
+│
+├── Phase 4: Debate (agents message each other)
+│   ├── correctness ↔ invariant-adversary, serialization
+│   ├── cross-component ↔ correctness, caller-audit
+│   ├── invariant-adversary ↔ correctness, cross-component
+│   ├── caller-audit ↔ invariant-adversary, cross-component
+│   ├── performance ↔ api, cross-component
+│   ├── api ↔ correctness, performance
+│   ├── serialization ↔ correctness, invariant-adversary
+│   ├── test-coverage ↔ serialization, caller-audit
+│   └── design ↔ cross-component, invariant-adversary
+│
+```
+
+
+## Agent Communication Protocol
+
+### During Initial Review (Phase 3)
+- Each agent writes findings to its own file
+- Each agent sends a summary message to team lead when done
+- Agents do NOT communicate with each other yet
+
+### During Debate (Phase 4)
+- Team lead sends each agent a message with instructions to:
+  1. Read the assigned agents' findings files
+  2. Send critique messages to those agents
+  3. Respond to incoming critiques
+  4. Update their own findings file with revisions
+- Each critique message should include:
+  - Which finding they're addressing (e.g., "Correctness F1")
+  - Whether they AGREE, DISAGREE, or want to REFINE
+  - Their reasoning with code evidence
+  - Suggested severity adjustment (if any)
+
+### Message format example
+```
+To: correctness-reviewer
+Re: Your Finding F1
+
+AGREE/DISAGREE/REFINE - [reasoning with code evidence].
+[Suggested severity adjustment if any.]
+```
+
+## Review Anti-Patterns
+
+These recurring failure modes lead to missed bugs. Each is detailed in the
+referenced section; this table is a quick-reference checklist.
+
+| Anti-Pattern | Fix | Reference |
+|---|---|---|
+| Return-Value Tunnel Vision | Trace callee MUTATIONS, not just returns | Section 2c |
+| Default-Configuration Bias | Enumerate all execution contexts | Section 2h |
+| Assert-as-Proof | Try to BREAK every assert | Section 2i |
+| Write-Path-Only Analysis | Trace data readers, not just writers | Section 2g |
+| Confirmation-Seeking Research | Use adversarial prompts ("find where X fails") | Invariant Adversary agent |
+| Data-Flow vs Control-Flow Confusion | Separate who CALLS from who READS the data | Section 2g |

cmake/RocksDBConfig.cmake.in

@@ -44,10 +44,6 @@ if(@WITH_NUMA@)
   find_dependency(NUMA)
 endif()
 
-if(@WITH_TBB@)
-  find_dependency(TBB)
-endif()
-
 find_dependency(Threads)
 
 include("${CMAKE_CURRENT_LIST_DIR}/RocksDBTargets.cmake")

cmake/modules/FindTBB.cmake

@@ -1,33 +0,0 @@
-# - Find TBB
-# Find the Thread Building Blocks library and includes
-#
-# TBB_INCLUDE_DIRS - where to find tbb.h, etc.
-# TBB_LIBRARIES - List of libraries when using TBB.
-# TBB_FOUND - True if TBB found.
-
-if(NOT DEFINED TBB_ROOT_DIR)
-  set(TBB_ROOT_DIR "$ENV{TBBROOT}")
-endif()
-
-find_path(TBB_INCLUDE_DIRS
-  NAMES tbb/tbb.h
-  HINTS ${TBB_ROOT_DIR}/include)
-
-find_library(TBB_LIBRARIES
-  NAMES tbb
-  HINTS ${TBB_ROOT_DIR}/lib ENV LIBRARY_PATH)
-
-include(FindPackageHandleStandardArgs)
-find_package_handle_standard_args(TBB DEFAULT_MSG TBB_LIBRARIES TBB_INCLUDE_DIRS)
-
-mark_as_advanced(
-  TBB_LIBRARIES
-  TBB_INCLUDE_DIRS)
-
-if(TBB_FOUND AND NOT (TARGET TBB::TBB))
-  add_library (TBB::TBB UNKNOWN IMPORTED)
-  set_target_properties(TBB::TBB
-    PROPERTIES
-      IMPORTED_LOCATION ${TBB_LIBRARIES}
-      INTERFACE_INCLUDE_DIRECTORIES ${TBB_INCLUDE_DIRS})
-endif()

crash_test.mk

@@ -7,6 +7,13 @@ DB_STRESS_CMD?=./db_stress
 
 include common.mk
 
+ifdef COMPILE_WITH_TSAN
+	# Keep direct `make -f crash_test.mk COMPILE_WITH_TSAN=1 ...` runs
+	# aligned with the main Makefile's TSAN runtime options.
+	TSAN_OPTIONS?=suppressions=$(CURDIR)/tools/tsan_suppressions.txt
+	export TSAN_OPTIONS
+endif
+
 CRASHTEST_MAKE=$(MAKE) -f crash_test.mk
 CRASHTEST_PY=$(PYTHON) -u tools/db_crashtest.py --stress_cmd=$(DB_STRESS_CMD) --cleanup_cmd='$(DB_CLEANUP_CMD)' --destroy_db_initially=1
 

db/arena_wrapped_db_iter.cc

@@ -30,13 +30,78 @@ inline static SequenceNumber GetSeqNum(const DBImpl* db, const Snapshot* s) {
 Status ArenaWrappedDBIter::GetProperty(std::string prop_name,
                                        std::string* prop) {
   if (prop_name == "rocksdb.iterator.super-version-number") {
+    if (!internal_iter_initialized_) {
+      *prop = std::to_string(sv_number_);
+      return Status::OK();
+    }
     // First try to pass the value returned from inner iterator.
     if (!db_iter_->GetProperty(prop_name, prop).ok()) {
       *prop = std::to_string(sv_number_);
     }
     return Status::OK();
   }
-  return db_iter_->GetProperty(prop_name, prop);
+  if (!EnsureInternalIteratorInitialized(nullptr).ok()) {
+    return db_iter_->status();
+  }
+  return db_iter_->GetProperty(std::move(prop_name), prop);
+}
+
+void ArenaWrappedDBIter::CleanupDeferredSuperVersion() {
+  if (deferred_sv_ != nullptr) {
+    assert(db_impl_ != nullptr);
+    db_impl_->CleanupIteratorSuperVersion(
+        deferred_sv_, read_options_.background_purge_on_iterator_cleanup);
+    deferred_sv_ = nullptr;
+  }
+}
+
+void ArenaWrappedDBIter::DestroyDBIter() {
+  db_iter_->~DBIter();
+  CleanupDeferredSuperVersion();
+}
+
+void ArenaWrappedDBIter::ColumnFamilyDataUnrefDeleter::operator()(
+    ColumnFamilyData* cfd) const {
+  if (cfd == nullptr) {
+    return;
+  }
+  assert(db_impl != nullptr);
+
+  InstrumentedMutexLock lock(db_impl->mutex());
+  cfd->UnrefAndTryDelete();
+}
+
+void ArenaWrappedDBIter::DestroyDBIterAndArena() {
+  DestroyDBIter();
+  arena_.~Arena();
+}
+
+Status ArenaWrappedDBIter::EnsureInternalIteratorInitialized(
+    const MultiScanArgs* scan_opts) {
+  if (internal_iter_initialized_) {
+    return Status::OK();
+  }
+  if (db_impl_ == nullptr || deferred_cfd_ == nullptr ||
+      deferred_sv_ == nullptr) {
+    Status s = Status::InvalidArgument(
+        "Internal iterator cannot be initialized without deferred DB state");
+    db_iter_->set_status(s);
+    db_iter_->set_valid(false);
+    return s;
+  }
+
+  const MultiScanArgs* pruning_scan_opts =
+      scan_opts != nullptr && scan_opts->HasBoundedScanRanges() ? scan_opts
+                                                                : nullptr;
+  child_read_options_ = read_options_;
+  child_read_options_.snapshot = nullptr;
+  InternalIterator* internal_iter = db_impl_->NewInternalIterator(
+      child_read_options_, deferred_cfd_, deferred_sv_, &arena_, sequence_,
+      /*allow_unprepared_value=*/true, this, pruning_scan_opts);
+  deferred_cfd_ = nullptr;
+  deferred_sv_ = nullptr;
+  SetIterUnderDBIterImpl(internal_iter);
+  return Status::OK();
 }
 
 void ArenaWrappedDBIter::Init(
@@ -44,18 +109,26 @@ void ArenaWrappedDBIter::Init(
     const MutableCFOptions& mutable_cf_options, const Version* version,
     const SequenceNumber& sequence, uint64_t version_number,
     ReadCallback* read_callback, ColumnFamilyHandleImpl* cfh,
-    bool expose_blob_index, bool allow_refresh, ReadOnlyMemTable* active_mem) {
+    bool expose_blob_index, bool allow_refresh, ReadOnlyMemTable* active_mem,
+    DBImpl* db_impl, ColumnFamilyData* cfd) {
   read_options_ = read_options;
+  child_read_options_ = read_options;
   if (!CheckFSFeatureSupport(env->GetFileSystem().get(),
                              FSSupportedOps::kAsyncIO)) {
     read_options_.async_io = false;
   }
   read_options_.total_order_seek |= ioptions.prefix_seek_opt_in_only;
 
-  db_iter_ = DBIter::NewIter(
-      env, read_options_, ioptions, mutable_cf_options,
-      ioptions.user_comparator, /*internal_iter=*/nullptr, version, sequence,
-      read_callback, active_mem, cfh, expose_blob_index, &arena_);
+  if (cfh != nullptr) {
+    db_impl = cfh->db();
+    cfd = cfh->cfd();
+  }
+
+  db_iter_ = DBIter::NewIter(env, read_options_, ioptions, mutable_cf_options,
+                             ioptions.user_comparator,
+                             /*internal_iter=*/nullptr, version, sequence,
+                             read_callback, active_mem, /*cfh=*/nullptr,
+                             expose_blob_index, &arena_, db_impl, cfd);
 
   sv_number_ = version_number;
   allow_refresh_ = allow_refresh;
@@ -65,13 +138,13 @@ void ArenaWrappedDBIter::Init(
 
 void ArenaWrappedDBIter::MaybeAutoRefresh(bool is_seek,
                                           DBIter::Direction direction) {
-  if (cfh_ != nullptr && read_options_.snapshot != nullptr && allow_refresh_ &&
-      read_options_.auto_refresh_iterator_with_snapshot) {
+  if (cfd_ref_ != nullptr && read_options_.snapshot != nullptr &&
+      allow_refresh_ && read_options_.auto_refresh_iterator_with_snapshot) {
     // The intent here is to capture the superversion number change
     // reasonably soon from the time it actually happened. As such,
     // we're fine with weaker synchronization / ordering guarantees
     // provided by relaxed atomic (in favor of less CPU / mem overhead).
-    uint64_t cur_sv_number = cfh_->cfd()->GetSuperVersionNumberRelaxed();
+    uint64_t cur_sv_number = cfd_ref_->GetSuperVersionNumberRelaxed();
     if ((sv_number_ != cur_sv_number) && status().ok()) {
       // Changing iterators' direction is pretty heavy-weight operation and
       // could have unintended consequences when it comes to prefix seek.
@@ -150,13 +223,13 @@ void ArenaWrappedDBIter::DoRefresh(const Snapshot* snapshot,
   // present in the error log, but won't be reflected in the iterator status.
   // This is by design as we expect compaction to clean up those obsolete files
   // eventually.
-  db_iter_->~DBIter();
-
-  arena_.~Arena();
+  DestroyDBIterAndArena();
   new (&arena_) Arena();
 
-  auto cfd = cfh_->cfd();
-  auto db_impl = cfh_->db();
+  auto cfd = cfd_ref_.get();
+  auto db_impl = db_impl_;
+  assert(cfd != nullptr);
+  assert(db_impl != nullptr);
 
   SuperVersion* sv = cfd->GetReferencedSuperVersion(db_impl);
   assert(sv->version_number >= sv_number);
@@ -164,23 +237,27 @@ void ArenaWrappedDBIter::DoRefresh(const Snapshot* snapshot,
   if (read_callback_) {
     read_callback_->Refresh(read_seq);
   }
+  // TODO: Preserve Prepare() scan options across Refresh() so a refreshed
+  // MultiScan iterator can rebuild the same pruned tree.
   Init(env, read_options_, cfd->ioptions(), sv->mutable_cf_options, sv->current,
-       read_seq, sv->version_number, read_callback_, cfh_, expose_blob_index_,
-       allow_refresh_, allow_mark_memtable_for_flush_ ? sv->mem : nullptr);
+       read_seq, sv->version_number, read_callback_, nullptr,
+       expose_blob_index_, allow_refresh_,
+       allow_mark_memtable_for_flush_ ? sv->mem : nullptr, db_impl, cfd);
 
   InternalIterator* internal_iter = db_impl->NewInternalIterator(
       read_options_, cfd, sv, &arena_, read_seq,
       /* allow_unprepared_value */ true, /* db_iter */ this);
   SetIterUnderDBIter(internal_iter);
+  internal_iter_initialized_ = true;
 }
 
 Status ArenaWrappedDBIter::Refresh(const Snapshot* snapshot) {
-  if (cfh_ == nullptr || !allow_refresh_) {
+  if (cfd_ref_ == nullptr || db_impl_ == nullptr || !allow_refresh_) {
     return Status::NotSupported("Creating renew iterator is not allowed.");
   }
   assert(db_iter_ != nullptr);
-  auto cfd = cfh_->cfd();
-  auto db_impl = cfh_->db();
+  auto cfd = cfd_ref_.get();
+  auto db_impl = db_impl_;
 
   // TODO(yiwu): For last_seq_same_as_publish_seq_==false, this is not the
   // correct behavior. Will be corrected automatically when we take a snapshot
@@ -193,6 +270,13 @@ Status ArenaWrappedDBIter::Refresh(const Snapshot* snapshot) {
   TEST_SYNC_POINT("ArenaWrappedDBIter::Refresh:1");
   TEST_SYNC_POINT("ArenaWrappedDBIter::Refresh:2");
 
+  if (!internal_iter_initialized_) {
+    Status s = EnsureInternalIteratorInitialized(nullptr);
+    if (!s.ok()) {
+      return s;
+    }
+  }
+
   while (true) {
     if (sv_number_ != cur_sv_number) {
       DoRefresh(snapshot, cur_sv_number);
@@ -250,25 +334,47 @@ Status ArenaWrappedDBIter::Refresh(const Snapshot* snapshot) {
   return Status::OK();
 }
 
+void ArenaWrappedDBIter::Prepare(const MultiScanArgs& scan_opts) {
+  if (prepare_called_) {
+    db_iter_->set_status(Status::InvalidArgument(
+        "Prepare called more than once on the same iterator"));
+    db_iter_->set_valid(false);
+    return;
+  }
+  prepare_called_ = true;
+
+  Status s = db_iter_->SetScanOptionsForPrepare(scan_opts);
+  if (!s.ok()) {
+    return;
+  }
+
+  const MultiScanArgs* pruning_scan_opts =
+      scan_opts.HasBoundedScanRanges() ? &scan_opts : nullptr;
+  s = EnsureInternalIteratorInitialized(pruning_scan_opts);
+  if (!s.ok()) {
+    return;
+  }
+
+  db_iter_->PrepareInternalChildren();
+}
+
 ArenaWrappedDBIter* NewArenaWrappedDbIterator(
     Env* env, const ReadOptions& read_options, ColumnFamilyHandleImpl* cfh,
     SuperVersion* sv, const SequenceNumber& sequence,
     ReadCallback* read_callback, DBImpl* db_impl, bool expose_blob_index,
     bool allow_refresh, bool allow_mark_memtable_for_flush) {
   ArenaWrappedDBIter* db_iter = new ArenaWrappedDBIter();
+  ColumnFamilyData* cfd = cfh->cfd();
   db_iter->Init(env, read_options, cfh->cfd()->ioptions(),
                 sv->mutable_cf_options, sv->current, sequence,
                 sv->version_number, read_callback, cfh, expose_blob_index,
                 allow_refresh,
                 allow_mark_memtable_for_flush ? sv->mem : nullptr);
-  if (cfh != nullptr && allow_refresh) {
-    db_iter->StoreRefreshInfo(cfh, read_callback, expose_blob_index);
+  if (allow_refresh && cfd != nullptr) {
+    db_iter->StoreRefreshInfo(read_callback, expose_blob_index);
   }
-
-  InternalIterator* internal_iter = db_impl->NewInternalIterator(
-      db_iter->GetReadOptions(), cfh->cfd(), sv, db_iter->GetArena(), sequence,
-      /*allow_unprepared_value=*/true, db_iter);
-  db_iter->SetIterUnderDBIter(internal_iter);
+  db_iter->StoreDeferredInitInfo(db_impl, cfd, sv, sequence,
+                                 allow_mark_memtable_for_flush);
 
   return db_iter;
 }

db/arena_wrapped_db_iter.h

@@ -10,6 +10,7 @@
 #pragma once
 #include <stdint.h>
 
+#include <memory>
 #include <string>
 
 #include "db/db_impl/db_impl.h"
@@ -33,10 +34,17 @@ class Version;
 // When using the class's Iterator interface, the behavior is exactly
 // the same as the inner DBIter.
 class ArenaWrappedDBIter : public Iterator {
+  struct ColumnFamilyDataUnrefDeleter {
+    DBImpl* db_impl = nullptr;
+    void operator()(ColumnFamilyData* cfd) const;
+  };
+  using ColumnFamilyDataRef =
+      std::unique_ptr<ColumnFamilyData, ColumnFamilyDataUnrefDeleter>;
+
  public:
   ~ArenaWrappedDBIter() override {
     if (db_iter_ != nullptr) {
-      db_iter_->~DBIter();
+      DestroyDBIter();
     } else {
       assert(false);
     }
@@ -51,7 +59,7 @@ class ArenaWrappedDBIter : public Iterator {
   // Set the internal iterator wrapped inside the DB Iterator. Usually it is
   // a merging iterator.
   virtual void SetIterUnderDBIter(InternalIterator* iter) {
-    db_iter_->SetIter(iter);
+    SetIterUnderDBIterImpl(iter);
   }
 
   void SetMemtableRangetombstoneIter(
@@ -60,26 +68,48 @@ class ArenaWrappedDBIter : public Iterator {
   }
 
   bool Valid() const override { return db_iter_->Valid(); }
-  void SeekToFirst() override { db_iter_->SeekToFirst(); }
-  void SeekToLast() override { db_iter_->SeekToLast(); }
+  void SeekToFirst() override {
+    if (!EnsureInternalIteratorInitialized(nullptr).ok()) {
+      return;
+    }
+    db_iter_->SeekToFirst();
+  }
+  void SeekToLast() override {
+    if (!EnsureInternalIteratorInitialized(nullptr).ok()) {
+      return;
+    }
+    db_iter_->SeekToLast();
+  }
   // 'target' does not contain timestamp, even if user timestamp feature is
   // enabled.
   void Seek(const Slice& target) override {
+    if (!EnsureInternalIteratorInitialized(nullptr).ok()) {
+      return;
+    }
     MaybeAutoRefresh(true /* is_seek */, DBIter::kForward);
     db_iter_->Seek(target);
   }
 
   void SeekForPrev(const Slice& target) override {
+    if (!EnsureInternalIteratorInitialized(nullptr).ok()) {
+      return;
+    }
     MaybeAutoRefresh(true /* is_seek */, DBIter::kReverse);
     db_iter_->SeekForPrev(target);
   }
 
   void Next() override {
+    if (!EnsureInternalIteratorInitialized(nullptr).ok()) {
+      return;
+    }
     db_iter_->Next();
     MaybeAutoRefresh(false /* is_seek */, DBIter::kForward);
   }
 
   void Prev() override {
+    if (!EnsureInternalIteratorInitialized(nullptr).ok()) {
+      return;
+    }
     db_iter_->Prev();
     MaybeAutoRefresh(false /* is_seek */, DBIter::kReverse);
   }
@@ -96,12 +126,13 @@ class ArenaWrappedDBIter : public Iterator {
   Status Refresh() override;
   Status Refresh(const Snapshot*) override;
 
-  bool PrepareValue() override { return db_iter_->PrepareValue(); }
-
-  void Prepare(const MultiScanArgs& scan_opts) override {
-    db_iter_->Prepare(scan_opts);
+  bool PrepareValue() override {
+    return EnsureInternalIteratorInitialized(nullptr).ok() &&
+           db_iter_->PrepareValue();
   }
 
+  void Prepare(const MultiScanArgs& scan_opts) override;
+
   // FIXME: we could just pass SV in for mutable cf option, version and version
   // number, but this is used by SstFileReader which does not have a SV.
   void Init(Env* env, const ReadOptions& read_options,
@@ -110,27 +141,53 @@ class ArenaWrappedDBIter : public Iterator {
             const SequenceNumber& sequence, uint64_t version_number,
             ReadCallback* read_callback, ColumnFamilyHandleImpl* cfh,
             bool expose_blob_index, bool allow_refresh,
-            ReadOnlyMemTable* active_mem);
+            ReadOnlyMemTable* active_mem, DBImpl* db_impl = nullptr,
+            ColumnFamilyData* cfd = nullptr);
 
-  // Store some parameters so we can refresh the iterator at a later point
-  // with these same params
-  void StoreRefreshInfo(ColumnFamilyHandleImpl* cfh,
-                        ReadCallback* read_callback, bool expose_blob_index) {
-    cfh_ = cfh;
+  // Store parameters used only by explicit/auto-refresh.
+  void StoreRefreshInfo(ReadCallback* read_callback, bool expose_blob_index) {
     read_callback_ = read_callback;
     expose_blob_index_ = expose_blob_index;
   }
 
+  void StoreDeferredInitInfo(DBImpl* db_impl, ColumnFamilyData* cfd,
+                             SuperVersion* sv, const SequenceNumber& sequence,
+                             bool allow_mark_memtable_for_flush) {
+    assert(cfd != nullptr);
+    db_impl_ = db_impl;
+    cfd->Ref();
+    cfd_ref_ = ColumnFamilyDataRef(cfd, ColumnFamilyDataUnrefDeleter{db_impl});
+    deferred_cfd_ = cfd;
+    deferred_sv_ = sv;
+    sequence_ = sequence;
+    allow_mark_memtable_for_flush_ = allow_mark_memtable_for_flush;
+  }
+
  private:
+  Status EnsureInternalIteratorInitialized(const MultiScanArgs* scan_opts);
+  void SetIterUnderDBIterImpl(InternalIterator* iter) {
+    db_iter_->SetIter(iter);
+    internal_iter_initialized_ = true;
+  }
+  void CleanupDeferredSuperVersion();
+  void DestroyDBIter();
+  void DestroyDBIterAndArena();
   void DoRefresh(const Snapshot* snapshot, uint64_t sv_number);
   void MaybeAutoRefresh(bool is_seek, DBIter::Direction direction);
 
   DBIter* db_iter_ = nullptr;
   Arena arena_;
-  uint64_t sv_number_;
-  ColumnFamilyHandleImpl* cfh_ = nullptr;
+  uint64_t sv_number_ = 0;
+  DBImpl* db_impl_ = nullptr;
+  ColumnFamilyDataRef cfd_ref_{nullptr, ColumnFamilyDataUnrefDeleter{}};
+  ColumnFamilyData* deferred_cfd_ = nullptr;
+  SuperVersion* deferred_sv_ = nullptr;
+  SequenceNumber sequence_ = kMaxSequenceNumber;
+  bool internal_iter_initialized_ = false;
+  bool prepare_called_ = false;
   ReadOptions read_options_;
-  ReadCallback* read_callback_;
+  ReadOptions child_read_options_;
+  ReadCallback* read_callback_ = nullptr;
   bool expose_blob_index_ = false;
   bool allow_refresh_ = true;
   bool allow_mark_memtable_for_flush_ = true;

db/blob/blob_constants.h

@@ -11,6 +11,24 @@
 
 namespace ROCKSDB_NAMESPACE {
 
+// WARNING: This value is == kCurrentFileBlobIndexFileNumber.
+// Use this name only where file number zero means "no valid blob file" or
+// "current file" is not understood/supported.
 constexpr uint64_t kInvalidBlobFileNumber = 0;
 
+// WARNING: This value is == kInvalidBlobFileNumber.
+// Use this name only for BlobIndex references to the same physical file as what
+// is currently being read; generic blob-file metadata must treat zero as
+// invalid. (Using a distinct value like 1 was found to be more problematic,
+// e.g. because of legacy "stackable" blob implementation.)
+//
+// This "zero is invalid unless you are the embedded reader/writer" contract is
+// enforced by integrity checks that reject file number zero on generic paths;
+// see FileMetaData::UpdateBoundaries (write/output path) and Version::GetBlob /
+// Version::MultiGetBlob (read path). Same-file references must be resolved (by
+// EmbeddedBlobResolvingIterator) before they reach those paths. Do not weaken
+// those checks -- they catch leaks/corruption closer to the root cause.
+constexpr uint64_t kCurrentFileBlobIndexFileNumber = kInvalidBlobFileNumber;
+static_assert(kCurrentFileBlobIndexFileNumber == kInvalidBlobFileNumber);
+
 }  // namespace ROCKSDB_NAMESPACE

db/blob/blob_fetcher.cc

@@ -5,6 +5,8 @@
 
 #include "db/blob/blob_fetcher.h"
 
+#include "db/blob/blob_file_partition_manager.h"
+#include "db/blob/blob_index.h"
 #include "db/version_set.h"
 
 namespace ROCKSDB_NAMESPACE {
@@ -14,10 +16,13 @@ Status BlobFetcher::FetchBlob(const Slice& user_key,
                               FilePrefetchBuffer* prefetch_buffer,
                               PinnableSlice* blob_value,
                               uint64_t* bytes_read) const {
-  assert(version_);
-
-  return version_->GetBlob(read_options_, user_key, blob_index_slice,
-                           prefetch_buffer, blob_value, bytes_read);
+  BlobIndex blob_index;
+  Status status = blob_index.DecodeFrom(blob_index_slice);
+  if (status.ok()) {
+    status = FetchBlob(user_key, blob_index, prefetch_buffer, blob_value,
+                       bytes_read);
+  }
+  return status;
 }
 
 Status BlobFetcher::FetchBlob(const Slice& user_key,
@@ -25,10 +30,16 @@ Status BlobFetcher::FetchBlob(const Slice& user_key,
                               FilePrefetchBuffer* prefetch_buffer,
                               PinnableSlice* blob_value,
                               uint64_t* bytes_read) const {
-  assert(version_);
+  if (!allow_write_path_fallback_) {
+    assert(version_);
 
-  return version_->GetBlob(read_options_, user_key, blob_index, prefetch_buffer,
-                           blob_value, bytes_read);
+    return version_->GetBlob(read_options_, user_key, blob_index,
+                             prefetch_buffer, blob_value, bytes_read);
+  }
+
+  return BlobFilePartitionManager::ResolveBlobDirectWriteIndex(
+      read_options_, user_key, blob_index, version_, blob_file_cache_,
+      prefetch_buffer, blob_value, bytes_read);
 }
 
 }  // namespace ROCKSDB_NAMESPACE

db/blob/blob_fetcher.h

@@ -10,17 +10,25 @@
 
 namespace ROCKSDB_NAMESPACE {
 
+class BlobFileCache;
 class Version;
 class Slice;
 class FilePrefetchBuffer;
 class PinnableSlice;
 class BlobIndex;
 
-// A thin wrapper around the blob retrieval functionality of Version.
+// A thin wrapper around blob retrieval. By default it reads through Version,
+// and it can optionally fall back to direct-write blob files that are not yet
+// manifest-visible.
 class BlobFetcher {
  public:
-  BlobFetcher(const Version* version, const ReadOptions& read_options)
-      : version_(version), read_options_(read_options) {}
+  BlobFetcher(const Version* version, const ReadOptions& read_options,
+              BlobFileCache* blob_file_cache = nullptr,
+              bool allow_write_path_fallback = false)
+      : version_(version),
+        read_options_(read_options),
+        blob_file_cache_(blob_file_cache),
+        allow_write_path_fallback_(allow_write_path_fallback) {}
 
   Status FetchBlob(const Slice& user_key, const Slice& blob_index_slice,
                    FilePrefetchBuffer* prefetch_buffer,
@@ -33,5 +41,7 @@ class BlobFetcher {
  private:
   const Version* version_;
   ReadOptions read_options_;
+  BlobFileCache* blob_file_cache_;
+  bool allow_write_path_fallback_;
 };
 }  // namespace ROCKSDB_NAMESPACE

db/blob/blob_file_builder.cc

@@ -22,6 +22,7 @@
 #include "logging/logging.h"
 #include "options/cf_options.h"
 #include "options/options_helper.h"
+#include "rocksdb/file_system.h"
 #include "rocksdb/slice.h"
 #include "rocksdb/status.h"
 #include "test_util/sync_point.h"
@@ -67,13 +68,11 @@ BlobFileBuilder::BlobFileBuilder(
       min_blob_size_(mutable_cf_options->min_blob_size),
       blob_file_size_(mutable_cf_options->blob_file_size),
       blob_compression_type_(mutable_cf_options->blob_compression_type),
-      // TODO: support most CompressionOptions with a new CF option
-      // blob_compression_opts
       // TODO with schema change: support custom compression manager and options
       // such as max_compressed_bytes_per_kb
       // NOTE: returns nullptr for kNoCompression
       blob_compressor_(GetBuiltinV2CompressionManager()->GetCompressor(
-          CompressionOptions{}, blob_compression_type_)),
+          mutable_cf_options->blob_compression_opts, blob_compression_type_)),
       blob_compressor_wa_(blob_compressor_
                               ? blob_compressor_->ObtainWorkingArea()
                               : Compressor::ManagedWorkingArea{}),
@@ -111,7 +110,7 @@ Status BlobFileBuilder::Add(const Slice& key, const Slice& value,
   assert(blob_index);
   assert(blob_index->empty());
 
-  if (value.size() < min_blob_size_) {
+  if (value.empty() || value.size() < min_blob_size_) {
     return Status::OK();
   }
 
@@ -202,6 +201,8 @@ Status BlobFileBuilder::OpenBlobFileIfNeeded() {
   {
     assert(file_options_);
     fo_copy = *file_options_;
+    fo_copy.open_contract = FileOpenContract::kNoReopenForWrite |
+                            FileOpenContract::kNoReadersWhileOpenForWrite;
     fo_copy.write_hint = write_hint_;
     Status s = NewWritableFile(fs_, blob_file_path, &file, fo_copy);
 

db/blob/blob_file_cache.cc

@@ -9,6 +9,7 @@
 #include <memory>
 
 #include "db/blob/blob_file_reader.h"
+#include "logging/logging.h"
 #include "options/cf_options.h"
 #include "rocksdb/cache.h"
 #include "rocksdb/slice.h"
@@ -38,7 +39,8 @@ BlobFileCache::BlobFileCache(Cache* cache,
 
 Status BlobFileCache::GetBlobFileReader(
     const ReadOptions& read_options, uint64_t blob_file_number,
-    CacheHandleGuard<BlobFileReader>* blob_file_reader) {
+    CacheHandleGuard<BlobFileReader>* blob_file_reader,
+    bool allow_footer_skip_retry) {
   assert(blob_file_reader);
   assert(blob_file_reader->IsEmpty());
 
@@ -73,10 +75,22 @@ Status BlobFileCache::GetBlobFileReader(
 
   {
     assert(file_options_);
-    const Status s = BlobFileReader::Create(
+    Status s = BlobFileReader::Create(
         *immutable_options_, read_options, *file_options_, column_family_id_,
-        blob_file_read_hist_, blob_file_number, io_tracer_, &reader);
+        blob_file_read_hist_, blob_file_number, io_tracer_,
+        /*skip_footer_validation=*/false, &reader);
+    if (!s.ok() && s.IsCorruption() && allow_footer_skip_retry) {
+      reader.reset();
+      s = BlobFileReader::Create(
+          *immutable_options_, read_options, *file_options_, column_family_id_,
+          blob_file_read_hist_, blob_file_number, io_tracer_,
+          /*skip_footer_validation=*/true, &reader);
+    }
     if (!s.ok()) {
+      ROCKS_LOG_WARN(immutable_options_->logger,
+                     "BlobFileCache open failed for blob file %" PRIu64
+                     " in CF %u: %s",
+                     blob_file_number, column_family_id_, s.ToString().c_str());
       RecordTick(statistics, NO_FILE_ERRORS);
       return s;
     }
@@ -99,12 +113,121 @@ Status BlobFileCache::GetBlobFileReader(
   return Status::OK();
 }
 
+Status BlobFileCache::OpenBlobFileReaderUncached(
+    const ReadOptions& read_options, uint64_t blob_file_number,
+    std::unique_ptr<BlobFileReader>* blob_file_reader,
+    bool allow_footer_skip_retry) {
+  assert(blob_file_reader);
+  assert(!*blob_file_reader);
+
+  Statistics* const statistics = immutable_options_->stats;
+  RecordTick(statistics, NO_FILE_OPENS);
+
+  assert(file_options_);
+  Status s = BlobFileReader::Create(
+      *immutable_options_, read_options, *file_options_, column_family_id_,
+      blob_file_read_hist_, blob_file_number, io_tracer_,
+      /*skip_footer_validation=*/false, blob_file_reader);
+  if (!s.ok() && s.IsCorruption() && allow_footer_skip_retry) {
+    blob_file_reader->reset();
+    s = BlobFileReader::Create(
+        *immutable_options_, read_options, *file_options_, column_family_id_,
+        blob_file_read_hist_, blob_file_number, io_tracer_,
+        /*skip_footer_validation=*/true, blob_file_reader);
+  }
+  if (!s.ok()) {
+    RecordTick(statistics, NO_FILE_ERRORS);
+  }
+  return s;
+}
+
+Status BlobFileCache::InsertBlobFileReader(
+    uint64_t blob_file_number,
+    std::unique_ptr<BlobFileReader>* blob_file_reader,
+    CacheHandleGuard<BlobFileReader>* cached_blob_file_reader) {
+  assert(blob_file_reader);
+  assert(*blob_file_reader);
+  assert(cached_blob_file_reader);
+  assert(cached_blob_file_reader->IsEmpty());
+
+  const Slice key = GetSliceForKey(&blob_file_number);
+  // Serialize refreshes for the same blob file. The cache API does not expose
+  // a conditional replace, so refresh is intentionally modeled as
+  // Lookup/optional Erase/Insert under this per-key mutex.
+  MutexLock lock(&mutex_.Get(key));
+
+  TypedHandle* handle = cache_.Lookup(key);
+  if (handle) {
+    *cached_blob_file_reader = cache_.Guard(handle);
+    blob_file_reader->reset();
+    return Status::OK();
+  }
+
+  constexpr size_t charge = 1;
+  Status s = cache_.Insert(key, blob_file_reader->get(), charge, &handle);
+  if (!s.ok()) {
+    RecordTick(immutable_options_->stats, NO_FILE_ERRORS);
+    return s;
+  }
+
+  // Ownership transferred to the cache.
+  [[maybe_unused]] BlobFileReader* released_reader =
+      blob_file_reader->release();
+  *cached_blob_file_reader = cache_.Guard(handle);
+  return Status::OK();
+}
+
+Status BlobFileCache::RefreshBlobFileReader(
+    uint64_t blob_file_number,
+    std::unique_ptr<BlobFileReader>* blob_file_reader,
+    CacheHandleGuard<BlobFileReader>* cached_blob_file_reader) {
+  assert(blob_file_reader);
+  assert(*blob_file_reader);
+  assert(cached_blob_file_reader);
+  assert(cached_blob_file_reader->IsEmpty());
+
+  const Slice key = GetSliceForKey(&blob_file_number);
+  MutexLock lock(&mutex_.Get(key));
+
+  TypedHandle* handle = cache_.Lookup(key);
+  if (handle) {
+    BlobFileReader* const cached_reader = cache_.Value(handle);
+    assert(cached_reader != nullptr);
+
+    // Active direct-write blob files can grow between refresh attempts. Keep
+    // whichever reader observed the larger on-disk size so an older refresh
+    // cannot overwrite a newer one that another thread already installed.
+    if (cached_reader->GetFileSize() >= (*blob_file_reader)->GetFileSize()) {
+      *cached_blob_file_reader = cache_.Guard(handle);
+      blob_file_reader->reset();
+      return Status::OK();
+    }
+
+    cache_.Release(handle);
+    cache_.get()->Erase(key);
+  }
+
+  constexpr size_t charge = 1;
+  Status s = cache_.Insert(key, blob_file_reader->get(), charge, &handle);
+  if (!s.ok()) {
+    RecordTick(immutable_options_->stats, NO_FILE_ERRORS);
+    return s;
+  }
+
+  // Ownership transferred to the cache.
+  [[maybe_unused]] BlobFileReader* released_reader =
+      blob_file_reader->release();
+  *cached_blob_file_reader = cache_.Guard(handle);
+  return Status::OK();
+}
+
 void BlobFileCache::Evict(uint64_t blob_file_number) {
   // NOTE: sharing same Cache with table_cache
   const Slice key = GetSliceForKey(&blob_file_number);
 
   assert(cache_);
 
+  MutexLock lock(&mutex_.Get(key));
   cache_.get()->Erase(key);
 }
 

db/blob/blob_file_cache.h

@@ -32,9 +32,35 @@ class BlobFileCache {
   BlobFileCache(const BlobFileCache&) = delete;
   BlobFileCache& operator=(const BlobFileCache&) = delete;
 
+  // Returns a cached reader for `blob_file_number`, opening and caching it on
+  // miss. If `allow_footer_skip_retry` is true, a footer-validation corruption
+  // retries once without requiring a footer.
   Status GetBlobFileReader(const ReadOptions& read_options,
                            uint64_t blob_file_number,
-                           CacheHandleGuard<BlobFileReader>* blob_file_reader);
+                           CacheHandleGuard<BlobFileReader>* blob_file_reader,
+                           bool allow_footer_skip_retry = false);
+
+  // Opens a blob file reader without inserting it into the cache.
+  Status OpenBlobFileReaderUncached(
+      const ReadOptions& read_options, uint64_t blob_file_number,
+      std::unique_ptr<BlobFileReader>* blob_file_reader,
+      bool allow_footer_skip_retry = false);
+
+  // Inserts a freshly opened uncached reader unless another thread already
+  // cached the same blob file.
+  Status InsertBlobFileReader(
+      uint64_t blob_file_number,
+      std::unique_ptr<BlobFileReader>* blob_file_reader,
+      CacheHandleGuard<BlobFileReader>* cached_blob_file_reader);
+
+  // Installs a refresh-opened reader into the cache. If another thread has
+  // already cached a reader opened on at least as large a file size, keep that
+  // reader instead so a racing refresh cannot reintroduce an older active-file
+  // view after the blob file grows.
+  Status RefreshBlobFileReader(
+      uint64_t blob_file_number,
+      std::unique_ptr<BlobFileReader>* blob_file_reader,
+      CacheHandleGuard<BlobFileReader>* cached_blob_file_reader);
 
   // Called when a blob file is obsolete to ensure it is removed from the cache
   // to avoid effectively leaking the open file and assicated memory

db/blob/blob_file_cache_test.cc

@@ -192,6 +192,102 @@ TEST_F(BlobFileCacheTest, GetBlobFileReader_Race) {
   SyncPoint::GetInstance()->ClearAllCallBacks();
 }
 
+TEST_F(BlobFileCacheTest, RefreshBlobFileReaderPrefersLargestObservedFileSize) {
+  Options options;
+  options.env = mock_env_.get();
+  options.statistics = CreateDBStatistics();
+  options.cf_paths.emplace_back(
+      test::PerThreadDBPath(
+          mock_env_.get(),
+          "BlobFileCacheTest_"
+          "RefreshBlobFileReaderPrefersLargestObservedFileSize"),
+      0);
+  options.enable_blob_files = true;
+
+  constexpr uint32_t column_family_id = 1;
+  ImmutableOptions immutable_options(options);
+  constexpr uint64_t blob_file_number = 123;
+
+  const std::string blob_file_path =
+      BlobFileName(immutable_options.cf_paths.front().path, blob_file_number);
+
+  std::unique_ptr<FSWritableFile> file;
+  ASSERT_OK(NewWritableFile(immutable_options.fs.get(), blob_file_path, &file,
+                            FileOptions()));
+
+  std::unique_ptr<WritableFileWriter> file_writer(new WritableFileWriter(
+      std::move(file), blob_file_path, FileOptions(), immutable_options.clock));
+
+  constexpr Statistics* statistics = nullptr;
+  constexpr bool use_fsync = false;
+  constexpr bool do_flush = false;
+  BlobLogWriter blob_log_writer(std::move(file_writer), immutable_options.clock,
+                                statistics, blob_file_number, use_fsync,
+                                do_flush);
+
+  constexpr bool has_ttl = false;
+  constexpr ExpirationRange expiration_range;
+  BlobLogHeader header(column_family_id, kNoCompression, has_ttl,
+                       expiration_range);
+  ASSERT_OK(blob_log_writer.WriteHeader(WriteOptions(), header));
+
+  uint64_t key_offset = 0;
+  uint64_t blob_offset = 0;
+  ASSERT_OK(blob_log_writer.AddRecord(WriteOptions(), "key0", "blob0",
+                                      &key_offset, &blob_offset));
+  ASSERT_OK(blob_log_writer.Sync(WriteOptions()));
+
+  constexpr size_t capacity = 10;
+  std::shared_ptr<Cache> backing_cache = NewLRUCache(capacity);
+
+  FileOptions file_options;
+  constexpr HistogramImpl* blob_file_read_hist = nullptr;
+  BlobFileCache blob_file_cache(backing_cache.get(), &immutable_options,
+                                &file_options, column_family_id,
+                                blob_file_read_hist, nullptr /*IOTracer*/);
+
+  const ReadOptions read_options;
+  std::unique_ptr<BlobFileReader> stale_reader;
+  ASSERT_OK(blob_file_cache.OpenBlobFileReaderUncached(
+      read_options, blob_file_number, &stale_reader,
+      /*allow_footer_skip_retry=*/true));
+
+  std::unique_ptr<BlobFileReader> initial_cached_reader;
+  ASSERT_OK(blob_file_cache.OpenBlobFileReaderUncached(
+      read_options, blob_file_number, &initial_cached_reader,
+      /*allow_footer_skip_retry=*/true));
+
+  CacheHandleGuard<BlobFileReader> cached_reader;
+  ASSERT_OK(blob_file_cache.RefreshBlobFileReader(
+      blob_file_number, &initial_cached_reader, &cached_reader));
+  ASSERT_NE(cached_reader.GetValue(), nullptr);
+  const uint64_t initial_file_size = cached_reader.GetValue()->GetFileSize();
+
+  ASSERT_OK(blob_log_writer.AddRecord(WriteOptions(), "key1", "blob1",
+                                      &key_offset, &blob_offset));
+  ASSERT_OK(blob_log_writer.Sync(WriteOptions()));
+
+  std::unique_ptr<BlobFileReader> fresh_reader;
+  ASSERT_OK(blob_file_cache.OpenBlobFileReaderUncached(
+      read_options, blob_file_number, &fresh_reader,
+      /*allow_footer_skip_retry=*/true));
+  ASSERT_GT(fresh_reader->GetFileSize(), initial_file_size);
+
+  CacheHandleGuard<BlobFileReader> refreshed_reader;
+  ASSERT_OK(blob_file_cache.RefreshBlobFileReader(
+      blob_file_number, &fresh_reader, &refreshed_reader));
+  ASSERT_NE(refreshed_reader.GetValue(), nullptr);
+  ASSERT_GT(refreshed_reader.GetValue()->GetFileSize(), initial_file_size);
+  BlobFileReader* const largest_reader = refreshed_reader.GetValue();
+
+  CacheHandleGuard<BlobFileReader> preserved_reader;
+  ASSERT_OK(blob_file_cache.RefreshBlobFileReader(
+      blob_file_number, &stale_reader, &preserved_reader));
+  ASSERT_NE(preserved_reader.GetValue(), nullptr);
+  ASSERT_EQ(preserved_reader.GetValue(), largest_reader);
+  ASSERT_EQ(stale_reader.get(), nullptr);
+}
+
 TEST_F(BlobFileCacheTest, GetBlobFileReader_IOError) {
   Options options;
   options.env = mock_env_.get();

db/blob/blob_file_partition_manager.cc

@@ -0,0 +1,854 @@
+//  Copyright (c) Meta Platforms, Inc. and affiliates.
+//  This source code is licensed under both the GPLv2 (found in the
+//  COPYING file in the root directory) and Apache 2.0 License
+//  (found in the LICENSE.Apache file in the root directory).
+
+#include "db/blob/blob_file_partition_manager.h"
+
+#include <array>
+#include <atomic>
+#include <cinttypes>
+#include <memory>
+#include <utility>
+
+#include "cache/cache_key.h"
+#include "cache/typed_cache.h"
+#include "db/blob/blob_contents.h"
+#include "db/blob/blob_file_cache.h"
+#include "db/blob/blob_file_completion_callback.h"
+#include "db/blob/blob_file_reader.h"
+#include "db/blob/blob_index.h"
+#include "db/blob/blob_log_writer.h"
+#include "db/version_set.h"
+#include "file/filename.h"
+#include "file/read_write_util.h"
+#include "file/writable_file_writer.h"
+#include "logging/logging.h"
+#include "rocksdb/file_system.h"
+#include "util/aligned_buffer.h"
+#include "util/compression.h"
+#include "util/mutexlock.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+namespace {
+
+class RoundRobinBlobFilePartitionStrategy : public BlobFilePartitionStrategy {
+ public:
+  using BlobFilePartitionStrategy::SelectPartition;
+
+  const char* Name() const override {
+    return "RoundRobinBlobFilePartitionStrategy";
+  }
+
+  uint32_t SelectPartition(uint32_t num_partitions,
+                           uint32_t /*column_family_id*/, const Slice& /*key*/,
+                           const Slice& /*value*/) override {
+    assert(num_partitions > 0);
+    return static_cast<uint32_t>(
+        next_partition_.fetch_add(1, std::memory_order_relaxed) %
+        static_cast<uint64_t>(num_partitions));
+  }
+
+ private:
+  std::atomic<uint64_t> next_partition_{0};
+};
+
+struct DirectWriteCompressionState {
+  CompressionOptions compression_opts;
+  // `working_area` must be released before its owning compressor.
+  std::unique_ptr<Compressor> compressor;
+  Compressor::ManagedWorkingArea working_area;
+};
+
+DirectWriteCompressionState& GetDirectWriteCompressionState(
+    CompressionType compression, const CompressionOptions& compression_opts) {
+  assert(compression <= kLastBuiltinCompression);
+
+  static thread_local std::array<DirectWriteCompressionState,
+                                 static_cast<size_t>(kLastBuiltinCompression) +
+                                     1>
+      compression_states;
+
+  auto& compression_state =
+      compression_states[static_cast<size_t>(compression)];
+  if (compression != kNoCompression &&
+      (compression_state.compressor == nullptr ||
+       compression_state.compression_opts != compression_opts)) {
+    // BDW compression settings are mutable, so rebuild the per-thread cached
+    // compressor when the latest published opts for this type change.
+    compression_state.working_area = Compressor::ManagedWorkingArea{};
+    compression_state.compressor.reset();
+    compression_state.compression_opts = compression_opts;
+    compression_state.compressor =
+        GetBuiltinV2CompressionManager()->GetCompressor(compression_opts,
+                                                        compression);
+    if (compression_state.compressor != nullptr) {
+      compression_state.working_area =
+          compression_state.compressor->ObtainWorkingArea();
+    }
+  }
+
+  return compression_state;
+}
+
+}  // namespace
+
+BlobFilePartitionManager::BlobFilePartitionManager(
+    uint32_t num_partitions,
+    std::shared_ptr<BlobFilePartitionStrategy> strategy,
+    FileNumberAllocator file_number_allocator, FileSystem* fs,
+    SystemClock* clock, Statistics* statistics, const FileOptions& file_options,
+    std::string db_path, std::string column_family_name,
+    uint64_t blob_file_size, bool use_fsync, BlobFileCache* blob_file_cache,
+    BlobFileCompletionCallback* blob_callback,
+    const std::vector<std::shared_ptr<EventListener>>& listeners,
+    FileChecksumGenFactory* file_checksum_gen_factory,
+    const FileTypeSet& checksum_handoff_file_types,
+    const std::shared_ptr<IOTracer>& io_tracer, std::string db_id,
+    std::string db_session_id, Logger* info_log)
+    : num_partitions_(num_partitions == 0 ? 1 : num_partitions),
+      strategy_(strategy != nullptr
+                    ? std::move(strategy)
+                    : std::make_shared<RoundRobinBlobFilePartitionStrategy>()),
+      file_number_allocator_(std::move(file_number_allocator)),
+      fs_(fs),
+      clock_(clock),
+      statistics_(statistics),
+      file_options_(file_options),
+      db_path_(std::move(db_path)),
+      column_family_name_(std::move(column_family_name)),
+      blob_file_size_(blob_file_size),
+      use_fsync_(use_fsync),
+      blob_file_cache_(blob_file_cache),
+      blob_callback_(blob_callback),
+      listeners_(listeners),
+      file_checksum_gen_factory_(file_checksum_gen_factory),
+      checksum_handoff_file_types_(checksum_handoff_file_types),
+      io_tracer_(io_tracer),
+      db_id_(std::move(db_id)),
+      db_session_id_(std::move(db_session_id)),
+      info_log_(info_log) {
+  partitions_.reserve(num_partitions_);
+  for (uint32_t i = 0; i < num_partitions_; ++i) {
+    partitions_.emplace_back(new Partition());
+  }
+}
+
+BlobFilePartitionManager::~BlobFilePartitionManager() {
+  if (blob_file_cache_ != nullptr) {
+    std::vector<uint64_t> tracked_file_numbers;
+    {
+      ReadLock lock(&file_partition_mutex_);
+      tracked_file_numbers.reserve(file_to_partition_.size());
+      for (const auto& entry : file_to_partition_) {
+        tracked_file_numbers.push_back(entry.first);
+      }
+    }
+
+    for (uint64_t file_number : tracked_file_numbers) {
+      // Dropping the CF or closing the DB can destroy the manager while active
+      // direct-write readers are still cached. Evict them before the CF-owned
+      // blob cache goes away so Close() does not retain obsolete footer-less
+      // readers for files that are no longer live.
+      blob_file_cache_->Evict(file_number);
+    }
+  }
+}
+
+void BlobFilePartitionManager::ResetPartitionState(Partition* partition) {
+  partition->writer.reset();
+  partition->file_number = 0;
+  partition->file_size = 0;
+  partition->blob_count = 0;
+  partition->total_blob_bytes = 0;
+  partition->garbage_blob_count = 0;
+  partition->garbage_blob_bytes = 0;
+  partition->column_family_id = 0;
+  partition->compression = kNoCompression;
+  partition->sync_required = false;
+}
+
+void BlobFilePartitionManager::AddFilePartitionMapping(uint64_t file_number,
+                                                       uint32_t partition_idx) {
+  WriteLock lock(&file_partition_mutex_);
+  file_to_partition_[file_number] = partition_idx;
+}
+
+void BlobFilePartitionManager::RemoveFilePartitionMapping(
+    uint64_t file_number) {
+  WriteLock lock(&file_partition_mutex_);
+  file_to_partition_.erase(file_number);
+}
+
+Status BlobFilePartitionManager::OpenNewBlobFile(Partition* partition,
+                                                 uint32_t column_family_id,
+                                                 CompressionType compression,
+                                                 uint32_t partition_idx) {
+  assert(partition != nullptr);
+  assert(!partition->writer);
+
+  const uint64_t blob_file_number = file_number_allocator_();
+  const std::string blob_file_path = BlobFileName(db_path_, blob_file_number);
+
+  AddFilePartitionMapping(blob_file_number, partition_idx);
+
+  if (blob_callback_ != nullptr) {
+    blob_callback_->OnBlobFileCreationStarted(
+        blob_file_path, column_family_name_,
+        /*job_id=*/0, BlobFileCreationReason::kFlush);
+  }
+
+  std::unique_ptr<FSWritableFile> file;
+  FileOptions writer_file_options = file_options_;
+  writer_file_options.open_contract = FileOpenContract::kNoReopenForWrite;
+  Status s = NewWritableFile(fs_, blob_file_path, &file, writer_file_options);
+  if (!s.ok()) {
+    RemoveFilePartitionMapping(blob_file_number);
+    return s;
+  }
+
+  const bool perform_data_verification =
+      checksum_handoff_file_types_.Contains(FileType::kBlobFile);
+  auto file_writer = std::make_unique<WritableFileWriter>(
+      std::move(file), blob_file_path, writer_file_options, clock_, io_tracer_,
+      statistics_, Histograms::BLOB_DB_BLOB_FILE_WRITE_MICROS, listeners_,
+      file_checksum_gen_factory_, perform_data_verification);
+
+  // This only drains WritableFileWriter's buffered bytes so readers can see
+  // each appended record promptly. Durability still comes from SyncAllOpenFiles
+  // or AppendFooter(), both of which call Sync().
+  constexpr bool kDoFlushEachRecord = true;
+  auto blob_log_writer = std::make_unique<BlobLogWriter>(
+      std::move(file_writer), clock_, statistics_, blob_file_number, use_fsync_,
+      kDoFlushEachRecord);
+
+  constexpr bool has_ttl = false;
+  constexpr ExpirationRange expiration_range{};
+  BlobLogHeader header(column_family_id, compression, has_ttl,
+                       expiration_range);
+  s = blob_log_writer->WriteHeader(WriteOptions(), header);
+  if (!s.ok()) {
+    RemoveFilePartitionMapping(blob_file_number);
+    return s;
+  }
+
+  partition->writer = std::move(blob_log_writer);
+  partition->file_number = blob_file_number;
+  partition->file_size = BlobLogHeader::kSize;
+  partition->blob_count = 0;
+  partition->total_blob_bytes = 0;
+  partition->garbage_blob_count = 0;
+  partition->garbage_blob_bytes = 0;
+  partition->column_family_id = column_family_id;
+  partition->compression = compression;
+  partition->sync_required = false;
+  return Status::OK();
+}
+
+Status BlobFilePartitionManager::SealActiveBlobFile(
+    const WriteOptions& write_options, Partition* partition,
+    SealedFile* sealed_file) {
+  assert(partition != nullptr);
+  assert(partition->writer);
+  assert(sealed_file != nullptr);
+
+  const uint64_t file_number = partition->file_number;
+  Status s =
+      FinalizeBlobFile(write_options, partition->writer.get(), file_number,
+                       partition->blob_count, partition->total_blob_bytes,
+                       &sealed_file->addition);
+  if (!s.ok()) {
+    RemoveFilePartitionMapping(file_number);
+    ResetPartitionState(partition);
+    return s;
+  }
+
+  sealed_file->garbage_blob_count = partition->garbage_blob_count;
+  sealed_file->garbage_blob_bytes = partition->garbage_blob_bytes;
+  ResetPartitionState(partition);
+  return Status::OK();
+}
+
+Status BlobFilePartitionManager::SealDeferredFile(
+    const WriteOptions& write_options, DeferredFile* deferred,
+    SealedFile* sealed_file) {
+  assert(deferred != nullptr);
+  assert(deferred->writer);
+  assert(sealed_file != nullptr);
+
+  Status s = FinalizeBlobFile(
+      write_options, deferred->writer.get(), deferred->file_number,
+      deferred->blob_count, deferred->total_blob_bytes, &sealed_file->addition);
+  if (!s.ok()) {
+    RemoveFilePartitionMapping(deferred->file_number);
+    return s;
+  }
+
+  sealed_file->garbage_blob_count = deferred->garbage_blob_count;
+  sealed_file->garbage_blob_bytes = deferred->garbage_blob_bytes;
+  deferred->writer.reset();
+  return Status::OK();
+}
+
+Status BlobFilePartitionManager::FinalizeBlobFile(
+    const WriteOptions& write_options, BlobLogWriter* writer,
+    uint64_t file_number, uint64_t blob_count, uint64_t total_blob_bytes,
+    BlobFileAddition* addition) {
+  assert(writer != nullptr);
+  assert(addition != nullptr);
+
+  BlobLogFooter footer;
+  footer.blob_count = blob_count;
+
+  std::string checksum_method;
+  std::string checksum_value;
+  Status s = writer->AppendFooter(write_options, footer, &checksum_method,
+                                  &checksum_value);
+  if (!s.ok()) {
+    return s;
+  }
+
+  if (blob_callback_ != nullptr) {
+    const std::string blob_file_path = BlobFileName(db_path_, file_number);
+    s = blob_callback_->OnBlobFileCompleted(
+        blob_file_path, column_family_name_, /*job_id=*/0, file_number,
+        BlobFileCreationReason::kFlush, s, checksum_value, checksum_method,
+        blob_count, total_blob_bytes);
+    if (!s.ok()) {
+      return s;
+    }
+  }
+
+  *addition =
+      BlobFileAddition(file_number, blob_count, total_blob_bytes,
+                       std::move(checksum_method), std::move(checksum_value));
+  return Status::OK();
+}
+
+void BlobFilePartitionManager::AddSealedFileGarbage(
+    const SealedFile& sealed_file, std::vector<BlobFileGarbage>* garbages) {
+  assert(garbages != nullptr);
+
+  if (sealed_file.garbage_blob_count == 0) {
+    assert(sealed_file.garbage_blob_bytes == 0);
+    return;
+  }
+
+  garbages->emplace_back(sealed_file.addition.GetBlobFileNumber(),
+                         sealed_file.garbage_blob_count,
+                         sealed_file.garbage_blob_bytes);
+}
+
+bool BlobFilePartitionManager::MarkPartitionGarbage(Partition* partition,
+                                                    uint64_t file_number,
+                                                    uint64_t blob_count,
+                                                    uint64_t blob_bytes) {
+  assert(partition != nullptr);
+
+  if (!partition->writer || partition->file_number != file_number) {
+    return false;
+  }
+
+  partition->garbage_blob_count += blob_count;
+  partition->garbage_blob_bytes += blob_bytes;
+  assert(partition->garbage_blob_count <= partition->blob_count);
+  assert(partition->garbage_blob_bytes <= partition->total_blob_bytes);
+  return true;
+}
+
+bool BlobFilePartitionManager::MarkDeferredFileGarbage(DeferredFile* deferred,
+                                                       uint64_t file_number,
+                                                       uint64_t blob_count,
+                                                       uint64_t blob_bytes) {
+  assert(deferred != nullptr);
+
+  if (deferred->file_number != file_number) {
+    return false;
+  }
+
+  deferred->garbage_blob_count += blob_count;
+  deferred->garbage_blob_bytes += blob_bytes;
+  assert(deferred->garbage_blob_count <= deferred->blob_count);
+  assert(deferred->garbage_blob_bytes <= deferred->total_blob_bytes);
+  return true;
+}
+
+bool BlobFilePartitionManager::MarkSealedFileGarbage(
+    std::vector<SealedFile>* sealed_files, uint64_t file_number,
+    uint64_t blob_count, uint64_t blob_bytes) {
+  assert(sealed_files != nullptr);
+
+  for (auto& sealed_file : *sealed_files) {
+    if (sealed_file.addition.GetBlobFileNumber() != file_number) {
+      continue;
+    }
+
+    sealed_file.garbage_blob_count += blob_count;
+    sealed_file.garbage_blob_bytes += blob_bytes;
+    assert(sealed_file.garbage_blob_count <=
+           sealed_file.addition.GetTotalBlobCount());
+    assert(sealed_file.garbage_blob_bytes <=
+           sealed_file.addition.GetTotalBlobBytes());
+    return true;
+  }
+
+  return false;
+}
+
+Status BlobFilePartitionManager::MaybePrepopulateBlobCache(
+    const BlobDirectWriteSettings& settings, const Slice& original_value,
+    uint64_t blob_file_number, uint64_t blob_offset) {
+  if (settings.blob_cache == nullptr ||
+      settings.prepopulate_blob_cache != PrepopulateBlobCache::kFlushOnly) {
+    return Status::OK();
+  }
+
+  FullTypedCacheInterface<BlobContents, BlobContentsCreator> blob_cache{
+      settings.blob_cache};
+  const OffsetableCacheKey base_cache_key(db_id_, db_session_id_,
+                                          blob_file_number);
+  const CacheKey cache_key = base_cache_key.WithOffset(blob_offset);
+  return blob_cache.InsertSaved(cache_key.AsSlice(), original_value, nullptr,
+                                Cache::Priority::BOTTOM,
+                                CacheTier::kVolatileTier);
+}
+
+uint32_t BlobFilePartitionManager::SelectWideColumnPartition(
+    uint32_t column_family_id, const Slice& key,
+    const WideColumns& columns) const {
+  return strategy_->SelectPartition(num_partitions_, column_family_id, key,
+                                    columns) %
+         num_partitions_;
+}
+
+Status BlobFilePartitionManager::WriteBlob(
+    const WriteOptions& write_options, uint32_t column_family_id,
+    CompressionType compression, const Slice& key, const Slice& value,
+    uint64_t* blob_file_number, uint64_t* blob_offset, uint64_t* blob_size,
+    const BlobDirectWriteSettings& settings, const uint32_t* partition_idx) {
+  assert(blob_file_number != nullptr);
+  assert(blob_offset != nullptr);
+  assert(blob_size != nullptr);
+
+  // Do compression before taking the partition mutex so large-value CPU work
+  // does not serialize writers. A concurrent file rollover can still cause
+  // this compressed buffer to be discarded below, which is acceptable on this
+  // non-hot failure/configuration-change path.
+  GrowableBuffer compressed_value;
+  Slice write_value = value;
+  if (compression != kNoCompression) {
+    auto& compression_state =
+        GetDirectWriteCompressionState(compression, settings.compression_opts);
+    if (compression_state.compressor == nullptr) {
+      return Status::NotSupported(
+          "Blob direct write compression type not supported");
+    }
+    Status s = LegacyForceBuiltinCompression(*compression_state.compressor,
+                                             &compression_state.working_area,
+                                             value, &compressed_value);
+    if (!s.ok()) {
+      return s;
+    }
+    write_value = Slice(compressed_value);
+  }
+
+  // Partition selection is based on the logical write inputs. In particular,
+  // strategies that inspect value contents or size see the original
+  // uncompressed value rather than `write_value`. The modulo here is
+  // intentional so custom strategies can return arbitrary hashed or sentinel
+  // values without violating the partition bounds.
+  const uint32_t selected_partition_idx =
+      partition_idx != nullptr
+          ? (*partition_idx % num_partitions_)
+          : (strategy_->SelectPartition(num_partitions_, column_family_id, key,
+                                        value) %
+             num_partitions_);
+
+  {
+    MutexLock lock(&mutex_);
+    Partition* partition = partitions_[selected_partition_idx].get();
+
+    auto seal_current_file = [&]() -> Status {
+      if (!partition->writer) {
+        return Status::OK();
+      }
+      SealedFile sealed_file;
+      Status s = SealActiveBlobFile(write_options, partition, &sealed_file);
+      if (s.ok()) {
+        current_generation_sealed_files_.push_back(std::move(sealed_file));
+      }
+      return s;
+    };
+
+    const uint64_t record_size =
+        BlobLogRecord::kHeaderSize + key.size() + write_value.size();
+    const uint64_t future_file_size =
+        partition->file_size + record_size + BlobLogFooter::kSize;
+
+    if (partition->writer &&
+        (partition->column_family_id != column_family_id ||
+         partition->compression != compression ||
+         (partition->blob_count > 0 && future_file_size > blob_file_size_))) {
+      Status s = seal_current_file();
+      if (!s.ok()) {
+        return s;
+      }
+    }
+
+    if (!partition->writer) {
+      Status s = OpenNewBlobFile(partition, column_family_id, compression,
+                                 selected_partition_idx);
+      if (!s.ok()) {
+        return s;
+      }
+    }
+
+    uint64_t key_offset = 0;
+    Status s = partition->writer->AddRecord(write_options, key, write_value,
+                                            &key_offset, blob_offset);
+    if (!s.ok()) {
+      return s;
+    }
+
+    partition->sync_required = true;
+    partition->blob_count += 1;
+    partition->total_blob_bytes += record_size;
+    partition->file_size = BlobLogHeader::kSize + partition->total_blob_bytes;
+
+    *blob_file_number = partition->file_number;
+    *blob_size = write_value.size();
+  }
+
+  Status prepopulate_s = MaybePrepopulateBlobCache(
+      settings, value, *blob_file_number, *blob_offset);
+  if (!prepopulate_s.ok() && info_log_ != nullptr) {
+    ROCKS_LOG_WARN(info_log_,
+                   "Failed to pre-populate direct-write blob cache entry: %s",
+                   prepopulate_s.ToString().c_str());
+  }
+
+  return Status::OK();
+}
+
+void BlobFilePartitionManager::RotateCurrentGeneration() {
+  MutexLock lock(&mutex_);
+
+  GenerationBatch batch;
+  batch.sealed_files = std::move(current_generation_sealed_files_);
+  current_generation_sealed_files_.clear();
+
+  for (auto& partition : partitions_) {
+    if (!partition->writer) {
+      continue;
+    }
+
+    DeferredFile deferred;
+    deferred.writer = std::move(partition->writer);
+    deferred.file_number = partition->file_number;
+    deferred.blob_count = partition->blob_count;
+    deferred.total_blob_bytes = partition->total_blob_bytes;
+    deferred.garbage_blob_count = partition->garbage_blob_count;
+    deferred.garbage_blob_bytes = partition->garbage_blob_bytes;
+    ResetPartitionState(partition.get());
+    batch.deferred_files.emplace_back(std::move(deferred));
+  }
+
+  pending_generations_.emplace_back(std::move(batch));
+}
+
+Status BlobFilePartitionManager::PrepareFlushAdditions(
+    const WriteOptions& write_options, size_t num_generations,
+    std::vector<BlobFileAddition>* additions,
+    std::vector<BlobFileGarbage>* garbages,
+    std::vector<std::vector<uint64_t>>* generation_blob_file_numbers) {
+  assert(additions != nullptr);
+  assert(garbages != nullptr);
+  additions->clear();
+  garbages->clear();
+  if (generation_blob_file_numbers != nullptr) {
+    generation_blob_file_numbers->clear();
+  }
+
+  MutexLock lock(&mutex_);
+  if (num_generations > pending_generations_.size()) {
+    return Status::Corruption(
+        "Missing blob direct write generation metadata for flush");
+  }
+
+  for (size_t i = 0; i < num_generations; ++i) {
+    GenerationBatch& batch = pending_generations_[i];
+    while (!batch.deferred_files.empty()) {
+      DeferredFile deferred = std::move(batch.deferred_files.front());
+      batch.deferred_files.pop_front();
+
+      SealedFile sealed_file;
+      Status s = SealDeferredFile(write_options, &deferred, &sealed_file);
+      if (!s.ok()) {
+        return s;
+      }
+      // Keep each successfully sealed file attached to the generation
+      // immediately. If a later seal or the flush job itself fails, retry must
+      // reuse these exact on-disk files instead of finalizing replacements.
+      batch.sealed_files.push_back(std::move(sealed_file));
+    }
+
+    std::vector<uint64_t>* generation_file_numbers = nullptr;
+    if (generation_blob_file_numbers != nullptr) {
+      generation_blob_file_numbers->emplace_back();
+      generation_file_numbers = &generation_blob_file_numbers->back();
+      generation_file_numbers->reserve(batch.sealed_files.size());
+    }
+
+    for (const auto& sealed_file : batch.sealed_files) {
+      additions->push_back(sealed_file.addition);
+      AddSealedFileGarbage(sealed_file, garbages);
+      if (generation_file_numbers != nullptr) {
+        generation_file_numbers->push_back(
+            sealed_file.addition.GetBlobFileNumber());
+      }
+    }
+  }
+
+  return Status::OK();
+}
+
+Status BlobFilePartitionManager::MarkBlobWriteAsGarbage(uint64_t file_number,
+                                                        uint64_t blob_count,
+                                                        uint64_t blob_bytes) {
+  if (blob_count == 0) {
+    assert(blob_bytes == 0);
+    return Status::OK();
+  }
+
+  MutexLock lock(&mutex_);
+  for (auto& partition : partitions_) {
+    if (MarkPartitionGarbage(partition.get(), file_number, blob_count,
+                             blob_bytes)) {
+      return Status::OK();
+    }
+  }
+
+  if (MarkSealedFileGarbage(&current_generation_sealed_files_, file_number,
+                            blob_count, blob_bytes)) {
+    return Status::OK();
+  }
+
+  for (auto& batch : pending_generations_) {
+    for (auto& deferred : batch.deferred_files) {
+      if (MarkDeferredFileGarbage(&deferred, file_number, blob_count,
+                                  blob_bytes)) {
+        return Status::OK();
+      }
+    }
+    if (MarkSealedFileGarbage(&batch.sealed_files, file_number, blob_count,
+                              blob_bytes)) {
+      return Status::OK();
+    }
+  }
+
+  const std::string message =
+      "Could not match failed blob direct-write rollback for file #" +
+      std::to_string(file_number);
+  if (info_log_ != nullptr) {
+    ROCKS_LOG_ERROR(info_log_, "%s", message.c_str());
+  }
+  return Status::Corruption(message);
+}
+
+void BlobFilePartitionManager::CommitPreparedGenerations(
+    size_t num_generations) {
+  MutexLock lock(&mutex_);
+  while (num_generations-- > 0 && !pending_generations_.empty()) {
+    pending_generations_.pop_front();
+  }
+}
+
+Status BlobFilePartitionManager::SyncAllOpenFiles(
+    const WriteOptions& write_options) {
+  MutexLock lock(&mutex_);
+  for (const auto& partition : partitions_) {
+    if (!partition->writer || !partition->sync_required) {
+      continue;
+    }
+    Status s = partition->writer->Sync(write_options);
+    if (!s.ok()) {
+      return s;
+    }
+    partition->sync_required = false;
+  }
+  return Status::OK();
+}
+
+void BlobFilePartitionManager::GetActiveBlobFileNumbers(
+    UnorderedSet<uint64_t>* file_numbers) const {
+  assert(file_numbers != nullptr);
+  ReadLock lock(&file_partition_mutex_);
+  for (const auto& entry : file_to_partition_) {
+    file_numbers->insert(entry.first);
+  }
+}
+
+void BlobFilePartitionManager::GetProtectedBlobFileNumbers(
+    UnorderedSet<uint64_t>* file_numbers) const {
+  assert(file_numbers != nullptr);
+  ReadLock lock(&file_partition_mutex_);
+  for (const auto& entry : protected_blob_file_refs_) {
+    file_numbers->insert(entry.first);
+  }
+}
+
+bool BlobFilePartitionManager::IsTrackedBlobFileNumber(
+    uint64_t file_number) const {
+  ReadLock lock(&file_partition_mutex_);
+  return file_to_partition_.find(file_number) != file_to_partition_.end() ||
+         protected_blob_file_refs_.find(file_number) !=
+             protected_blob_file_refs_.end();
+}
+
+void BlobFilePartitionManager::ProtectSealedBlobFileNumbers(
+    const std::vector<uint64_t>& file_numbers) {
+  if (file_numbers.empty()) {
+    return;
+  }
+
+  WriteLock lock(&file_partition_mutex_);
+  for (uint64_t file_number : file_numbers) {
+    ++protected_blob_file_refs_[file_number];
+  }
+}
+
+void BlobFilePartitionManager::UnprotectSealedBlobFileNumbers(
+    const std::vector<uint64_t>& file_numbers) {
+  if (file_numbers.empty()) {
+    return;
+  }
+
+  WriteLock lock(&file_partition_mutex_);
+  for (uint64_t file_number : file_numbers) {
+    auto it = protected_blob_file_refs_.find(file_number);
+    if (it == protected_blob_file_refs_.end()) {
+      if (info_log_ != nullptr) {
+        ROCKS_LOG_ERROR(info_log_,
+                        "Memtable blob protection underflow for file #%" PRIu64,
+                        file_number);
+      }
+      continue;
+    }
+
+    assert(it->second > 0);
+    --it->second;
+    if (it->second == 0) {
+      protected_blob_file_refs_.erase(it);
+      if (blob_file_cache_ != nullptr) {
+        // Once the last memtable-backed reference goes away, any cached reader
+        // for this file is either obsolete or can be reopened through the
+        // manifest-visible path. Evict it here so delayed protection does not
+        // leave an obsolete blob reader behind until DB close.
+        blob_file_cache_->Evict(file_number);
+      }
+    }
+  }
+}
+
+void BlobFilePartitionManager::RemoveFilePartitionMappings(
+    const std::vector<uint64_t>& file_numbers) {
+  if (file_numbers.empty()) {
+    return;
+  }
+
+  WriteLock lock(&file_partition_mutex_);
+  for (uint64_t file_number : file_numbers) {
+    file_to_partition_.erase(file_number);
+    if (blob_file_cache_ != nullptr) {
+      // In-flight direct-write reads may have cached a footer-less reader for
+      // this file before it was sealed. Drop it now so future manifest-visible
+      // reads reopen against the finalized on-disk size and footer state.
+      blob_file_cache_->Evict(file_number);
+    }
+  }
+}
+
+Status BlobFilePartitionManager::ResolveBlobDirectWriteIndex(
+    const ReadOptions& read_options, const Slice& user_key,
+    const BlobIndex& blob_idx, const Version* version,
+    BlobFileCache* blob_file_cache, FilePrefetchBuffer* prefetch_buffer,
+    PinnableSlice* blob_value, uint64_t* bytes_read) {
+  assert(blob_value != nullptr);
+
+  if (version != nullptr) {
+    // Only fall back when the blob file is still owned exclusively by the
+    // write path and therefore absent from Version metadata. Once Version
+    // knows the file number, every result from Version::GetBlob(), including
+    // corruption, must propagate directly.
+    if (blob_idx.HasTTL() || blob_idx.IsInlined() ||
+        version->storage_info()->GetBlobFileMetaData(blob_idx.file_number()) !=
+            nullptr) {
+      return version->GetBlob(read_options, user_key, blob_idx, prefetch_buffer,
+                              blob_value, bytes_read);
+    }
+  }
+
+  if (blob_file_cache == nullptr) {
+    return version != nullptr ? Status::Corruption("Invalid blob file number")
+                              : Status::NotFound();
+  }
+
+  if (read_options.read_tier == kBlockCacheTier) {
+    // The direct-write fallback below may need to open the blob file reader,
+    // which `kBlockCacheTier` forbids. Keep the normal Version-backed path
+    // above eligible for cache-only hits.
+    return Status::Incomplete("Cannot read blob(s): no disk I/O allowed");
+  }
+
+  Status s;
+  CacheHandleGuard<BlobFileReader> reader;
+  s = blob_file_cache->GetBlobFileReader(read_options, blob_idx.file_number(),
+                                         &reader,
+                                         /*allow_footer_skip_retry=*/true);
+  if (!s.ok()) {
+    return s;
+  }
+
+  std::unique_ptr<BlobContents> blob_contents;
+  s = reader.GetValue()->GetBlob(read_options, user_key, blob_idx.offset(),
+                                 blob_idx.size(), blob_idx.compression(),
+                                 prefetch_buffer, nullptr, &blob_contents,
+                                 bytes_read);
+  if (s.ok()) {
+    blob_value->PinSelf(blob_contents->data());
+    return s;
+  }
+
+  if (!s.IsCorruption()) {
+    return s;
+  }
+
+  reader.Reset();
+  blob_file_cache->Evict(blob_idx.file_number());
+
+  std::unique_ptr<BlobFileReader> fresh_reader;
+  s = blob_file_cache->OpenBlobFileReaderUncached(
+      read_options, blob_idx.file_number(), &fresh_reader,
+      /*allow_footer_skip_retry=*/true);
+  if (!s.ok()) {
+    return s;
+  }
+
+  std::unique_ptr<BlobContents> fresh_contents;
+  s = fresh_reader->GetBlob(read_options, user_key, blob_idx.offset(),
+                            blob_idx.size(), blob_idx.compression(),
+                            prefetch_buffer, nullptr, &fresh_contents,
+                            bytes_read);
+  if (s.ok()) {
+    blob_value->PinSelf(fresh_contents->data());
+    CacheHandleGuard<BlobFileReader> ignored;
+    blob_file_cache
+        ->RefreshBlobFileReader(blob_idx.file_number(), &fresh_reader, &ignored)
+        .PermitUncheckedError();
+  }
+
+  return s;
+}
+
+}  // namespace ROCKSDB_NAMESPACE

db/blob/blob_file_partition_manager.h

@@ -0,0 +1,294 @@
+//  Copyright (c) Meta Platforms, Inc. and affiliates.
+//  This source code is licensed under both the GPLv2 (found in the
+//  COPYING file in the root directory) and Apache 2.0 License
+//  (found in the LICENSE.Apache file in the root directory).
+#pragma once
+
+#include <cstdint>
+#include <deque>
+#include <functional>
+#include <memory>
+#include <string>
+#include <unordered_map>
+#include <vector>
+
+#include "db/blob/blob_file_addition.h"
+#include "db/blob/blob_file_garbage.h"
+#include "db/blob/blob_log_format.h"
+#include "db/blob/blob_write_batch_transformer.h"
+#include "port/port.h"
+#include "rocksdb/compression_type.h"
+#include "rocksdb/env.h"
+#include "rocksdb/file_system.h"
+#include "rocksdb/options.h"
+#include "rocksdb/rocksdb_namespace.h"
+#include "rocksdb/slice.h"
+#include "rocksdb/status.h"
+#include "util/hash_containers.h"
+namespace ROCKSDB_NAMESPACE {
+
+class BlobFileCache;
+class BlobFileCompletionCallback;
+class BlobIndex;
+class BlobLogWriter;
+class FilePrefetchBuffer;
+class IOTracer;
+class Logger;
+class PinnableSlice;
+class Version;
+class WritableFileWriter;
+struct FileOptions;
+struct ReadOptions;
+
+// Manages per-partition blob files for write-path blob direct write.
+//
+// The v1 design keeps each memtable switch as one FIFO generation batch.
+// Direct-write files for that memtable are sealed and registered when the
+// matching flush commits. The current implementation still serializes blob
+// appends through one manager mutex; follow-up PRs can add independent
+// per-partition locks to allow concurrent blob writes without changing the
+// generation/flush contract. Follow-up PRs can also broaden compatibility as
+// the feature matures.
+class BlobFilePartitionManager {
+ public:
+  using FileNumberAllocator = std::function<uint64_t()>;
+
+  // Creates the per-column-family manager for write-path blob files.
+  BlobFilePartitionManager(
+      uint32_t num_partitions,
+      std::shared_ptr<BlobFilePartitionStrategy> strategy,
+      FileNumberAllocator file_number_allocator, FileSystem* fs,
+      SystemClock* clock, Statistics* statistics,
+      const FileOptions& file_options, std::string db_path,
+      std::string column_family_name, uint64_t blob_file_size, bool use_fsync,
+      BlobFileCache* blob_file_cache, BlobFileCompletionCallback* blob_callback,
+      const std::vector<std::shared_ptr<EventListener>>& listeners,
+      FileChecksumGenFactory* file_checksum_gen_factory,
+      const FileTypeSet& checksum_handoff_file_types,
+      const std::shared_ptr<IOTracer>& io_tracer, std::string db_id,
+      std::string db_session_id, Logger* info_log);
+
+  // Evicts cached readers for manager-owned files during shutdown.
+  ~BlobFilePartitionManager();
+
+  // Appends one blob record to a partition file and returns the resulting
+  // BlobIndex location metadata.
+  Status WriteBlob(const WriteOptions& write_options, uint32_t column_family_id,
+                   CompressionType compression, const Slice& key,
+                   const Slice& value, uint64_t* blob_file_number,
+                   uint64_t* blob_offset, uint64_t* blob_size,
+                   const BlobDirectWriteSettings& settings,
+                   const uint32_t* partition_idx = nullptr);
+
+  // Selects the partition to use for all blob-backed columns of one PutEntity
+  // operation. The return value is already normalized to [0, num_partitions_).
+  uint32_t SelectWideColumnPartition(uint32_t column_family_id,
+                                     const Slice& key,
+                                     const WideColumns& columns) const;
+
+  // Move the current active partition files into the next immutable
+  // memtable-generation batch. Called from SwitchMemtable() while DB mutex is
+  // held. No I/O is performed here.
+  void RotateCurrentGeneration();
+
+  // Seal the first `num_generations` queued immutable generations and return
+  // all blob additions and initial-garbage updates that must be registered
+  // with the matching flush. Generations stay queued until
+  // CommitPreparedGenerations() is called. If `generation_blob_file_numbers`
+  // is non-null, it receives the sealed blob file numbers for each prepared
+  // generation in the same FIFO order.
+  Status PrepareFlushAdditions(const WriteOptions& write_options,
+                               size_t num_generations,
+                               std::vector<BlobFileAddition>* additions,
+                               std::vector<BlobFileGarbage>* garbages,
+                               std::vector<std::vector<uint64_t>>*
+                                   generation_blob_file_numbers = nullptr);
+
+  // Remove the first `num_generations` prepared immutable generations after
+  // their blob-file additions and garbage edits were committed to MANIFEST.
+  void CommitPreparedGenerations(size_t num_generations);
+
+  // In v1 flush-on-write mode each AddRecord flushes to the OS immediately, so
+  // there is nothing extra to do for the non-sync write path.
+  Status FlushAllOpenFiles(const WriteOptions& /*write_options*/) {
+    return Status::OK();
+  }
+
+  // Sync the active open blob files before a sync WAL write.
+  Status SyncAllOpenFiles(const WriteOptions& write_options);
+
+  // Returns blob files still owned by the manager and therefore not yet safe
+  // to consider obsolete.
+  void GetActiveBlobFileNumbers(UnorderedSet<uint64_t>* file_numbers) const;
+
+  // Returns sealed blob files that are still reachable through live memtables
+  // or old SuperVersions and therefore must not be purged yet.
+  void GetProtectedBlobFileNumbers(UnorderedSet<uint64_t>* file_numbers) const;
+
+  // Returns true when the blob file is still owned by the write path or
+  // protected by a live memtable / old SuperVersion.
+  bool IsTrackedBlobFileNumber(uint64_t file_number) const;
+
+  // Increments / decrements memtable-held protection on sealed blob files.
+  void ProtectSealedBlobFileNumbers(const std::vector<uint64_t>& file_numbers);
+  void UnprotectSealedBlobFileNumbers(
+      const std::vector<uint64_t>& file_numbers);
+
+  // Stops protecting file numbers that are now registered in MANIFEST.
+  void RemoveFilePartitionMappings(const std::vector<uint64_t>& file_numbers);
+
+  // Marks blob records from a failed transformed write as initial garbage for
+  // the target blob file while keeping the physical bytes in place.
+  // Returns Corruption if the manager can no longer match the blob file.
+  Status MarkBlobWriteAsGarbage(uint64_t file_number, uint64_t blob_count,
+                                uint64_t blob_bytes);
+
+  // Resolves a direct-write BlobIndex by consulting manifest-visible state
+  // first, then falling back to direct blob-file reads only when the target
+  // blob file is still write-path-owned and therefore not yet tracked by
+  // Version. Existing manifest-visible read results, including I/O failures,
+  // are returned directly rather than masked by fallback logic.
+  static Status ResolveBlobDirectWriteIndex(
+      const ReadOptions& read_options, const Slice& user_key,
+      const BlobIndex& blob_idx, const Version* version,
+      BlobFileCache* blob_file_cache, FilePrefetchBuffer* prefetch_buffer,
+      PinnableSlice* blob_value, uint64_t* bytes_read);
+
+ private:
+  struct Partition {
+    // Active writer for this partition, or null when the partition is idle.
+    std::unique_ptr<BlobLogWriter> writer;
+    // Metadata tracked while the active file remains open.
+    uint64_t file_number = 0;
+    uint64_t file_size = 0;
+    uint64_t blob_count = 0;
+    uint64_t total_blob_bytes = 0;
+    // Failed transformed writes already appended to this physical file.
+    uint64_t garbage_blob_count = 0;
+    uint64_t garbage_blob_bytes = 0;
+    uint32_t column_family_id = 0;
+    CompressionType compression = kNoCompression;
+    bool sync_required = false;
+  };
+
+  struct DeferredFile {
+    // Blob writer moved out of an active partition at memtable rotation.
+    std::unique_ptr<BlobLogWriter> writer;
+    // Metadata preserved until flush preparation seals the file.
+    uint64_t file_number = 0;
+    uint64_t blob_count = 0;
+    uint64_t total_blob_bytes = 0;
+    // Failed transformed writes already appended before the file is sealed.
+    uint64_t garbage_blob_count = 0;
+    uint64_t garbage_blob_bytes = 0;
+  };
+
+  struct SealedFile {
+    // Blob-file metadata ready to be added to MANIFEST.
+    BlobFileAddition addition;
+    // Initial unreachable bytes that should be registered as garbage
+    // alongside the file addition.
+    uint64_t garbage_blob_count = 0;
+    uint64_t garbage_blob_bytes = 0;
+  };
+
+  struct GenerationBatch {
+    // Files that were still open when the memtable became immutable.
+    std::deque<DeferredFile> deferred_files;
+    // Files already sealed for this memtable generation.
+    std::vector<SealedFile> sealed_files;
+  };
+
+  // Clears all active-file bookkeeping from a partition after sealing.
+  void ResetPartitionState(Partition* partition);
+  // Marks a blob file as manager-owned until it is committed or abandoned.
+  void AddFilePartitionMapping(uint64_t file_number, uint32_t partition_idx);
+  // Stops tracking a blob file after open/seal failure.
+  void RemoveFilePartitionMapping(uint64_t file_number);
+
+  // Opens a new active blob file for one partition.
+  Status OpenNewBlobFile(Partition* partition, uint32_t column_family_id,
+                         CompressionType compression, uint32_t partition_idx);
+  // Finalizes one active partition file and returns the MANIFEST addition.
+  Status SealActiveBlobFile(const WriteOptions& write_options,
+                            Partition* partition, SealedFile* sealed_file);
+  // Finalizes one deferred file that was carried over to flush preparation.
+  Status SealDeferredFile(const WriteOptions& write_options,
+                          DeferredFile* deferred, SealedFile* sealed_file);
+  // Appends the footer, reports file completion, and builds the resulting
+  // BlobFileAddition for one sealed blob file.
+  Status FinalizeBlobFile(const WriteOptions& write_options,
+                          BlobLogWriter* writer, uint64_t file_number,
+                          uint64_t blob_count, uint64_t total_blob_bytes,
+                          BlobFileAddition* addition);
+  // Appends one initial-garbage record to the flush output if any failed
+  // transformed writes targeted the sealed blob file.
+  static void AddSealedFileGarbage(const SealedFile& sealed_file,
+                                   std::vector<BlobFileGarbage>* garbages);
+  // Returns true if the active open file matches `file_number` and consumes
+  // the failed-write garbage accounting.
+  static bool MarkPartitionGarbage(Partition* partition, uint64_t file_number,
+                                   uint64_t blob_count, uint64_t blob_bytes);
+  // Returns true if the deferred flush-preparation file matches `file_number`
+  // and consumes the failed-write garbage accounting.
+  static bool MarkDeferredFileGarbage(DeferredFile* deferred,
+                                      uint64_t file_number, uint64_t blob_count,
+                                      uint64_t blob_bytes);
+  // Returns true if one already-sealed file matches `file_number` and
+  // consumes the failed-write garbage accounting.
+  static bool MarkSealedFileGarbage(std::vector<SealedFile>* sealed_files,
+                                    uint64_t file_number, uint64_t blob_count,
+                                    uint64_t blob_bytes);
+  // Seeds the blob cache with the original uncompressed value when configured.
+  Status MaybePrepopulateBlobCache(const BlobDirectWriteSettings& settings,
+                                   const Slice& original_value,
+                                   uint64_t blob_file_number,
+                                   uint64_t blob_offset);
+
+  // Configured partition fanout and write-selection strategy.
+  const uint32_t num_partitions_;
+  std::shared_ptr<BlobFilePartitionStrategy> strategy_;
+  // Shared services needed to create, track, and cache blob files.
+  FileNumberAllocator file_number_allocator_;
+  FileSystem* fs_;
+  SystemClock* clock_;
+  Statistics* statistics_;
+  FileOptions file_options_;
+  // Column-family context used for file naming and callbacks.
+  std::string db_path_;
+  std::string column_family_name_;
+  // File creation policy and shared blob-file infrastructure.
+  uint64_t blob_file_size_;
+  bool use_fsync_;
+  BlobFileCache* blob_file_cache_;
+  BlobFileCompletionCallback* blob_callback_;
+  std::vector<std::shared_ptr<EventListener>> listeners_;
+  FileChecksumGenFactory* file_checksum_gen_factory_;
+  FileTypeSet checksum_handoff_file_types_;
+  std::shared_ptr<IOTracer> io_tracer_;
+  // DB identity used when constructing cache keys.
+  std::string db_id_;
+  std::string db_session_id_;
+  Logger* info_log_;
+
+  // One active writer slot per configured partition.
+  std::vector<std::unique_ptr<Partition>> partitions_;
+
+  // Protects partition state and generation queues.
+  mutable port::Mutex mutex_;
+  // Files already sealed while the current mutable memtable was active.
+  std::vector<SealedFile> current_generation_sealed_files_;
+  // FIFO immutable memtable generations waiting to be flushed.
+  std::deque<GenerationBatch> pending_generations_;
+
+  // Tracks blob files still owned by active or deferred write-path state until
+  // MANIFEST commit publishes them.
+  std::unordered_map<uint64_t, uint32_t> file_to_partition_;
+  // Sealed direct-write files that remain reachable from live memtables, such
+  // as old SuperVersions serving lazy iterator reads after flush commit.
+  std::unordered_map<uint64_t, uint32_t> protected_blob_file_refs_;
+  // Protects file_to_partition_ and protected_blob_file_refs_.
+  mutable port::RWMutex file_partition_mutex_;
+};
+
+}  // namespace ROCKSDB_NAMESPACE

db/blob/blob_file_reader.cc

@@ -12,6 +12,7 @@
 #include "db/blob/blob_log_format.h"
 #include "file/file_prefetch_buffer.h"
 #include "file/filename.h"
+#include "file/read_write_util.h"
 #include "monitoring/statistics_impl.h"
 #include "options/cf_options.h"
 #include "rocksdb/file_system.h"
@@ -20,6 +21,7 @@
 #include "table/format.h"
 #include "table/multiget_context.h"
 #include "test_util/sync_point.h"
+#include "util/aligned_buffer.h"
 #include "util/compression.h"
 #include "util/stop_watch.h"
 
@@ -29,7 +31,7 @@ Status BlobFileReader::Create(
     const ImmutableOptions& immutable_options, const ReadOptions& read_options,
     const FileOptions& file_options, uint32_t column_family_id,
     HistogramImpl* blob_file_read_hist, uint64_t blob_file_number,
-    const std::shared_ptr<IOTracer>& io_tracer,
+    const std::shared_ptr<IOTracer>& io_tracer, bool skip_footer_validation,
     std::unique_ptr<BlobFileReader>* blob_file_reader) {
   assert(blob_file_reader);
   assert(!*blob_file_reader);
@@ -40,7 +42,8 @@ Status BlobFileReader::Create(
   {
     const Status s =
         OpenFile(immutable_options, file_options, blob_file_read_hist,
-                 blob_file_number, io_tracer, &file_size, &file_reader);
+                 blob_file_number, io_tracer, &file_size, &file_reader,
+                 /*skip_footer_size_check=*/skip_footer_validation);
     if (!s.ok()) {
       return s;
     }
@@ -61,7 +64,7 @@ Status BlobFileReader::Create(
     }
   }
 
-  {
+  if (!skip_footer_validation) {
     const Status s =
         ReadFooter(file_reader.get(), read_options, file_size, statistics);
     if (!s.ok()) {
@@ -76,9 +79,10 @@ Status BlobFileReader::Create(
         compression_type);
   }
 
-  blob_file_reader->reset(new BlobFileReader(
-      std::move(file_reader), file_size, compression_type,
-      std::move(decompressor), immutable_options.clock, statistics));
+  blob_file_reader->reset(
+      new BlobFileReader(std::move(file_reader), file_size, compression_type,
+                         std::move(decompressor), immutable_options.clock,
+                         statistics, !skip_footer_validation));
 
   return Status::OK();
 }
@@ -87,7 +91,8 @@ Status BlobFileReader::OpenFile(
     const ImmutableOptions& immutable_options, const FileOptions& file_opts,
     HistogramImpl* blob_file_read_hist, uint64_t blob_file_number,
     const std::shared_ptr<IOTracer>& io_tracer, uint64_t* file_size,
-    std::unique_ptr<RandomAccessFileReader>* file_reader) {
+    std::unique_ptr<RandomAccessFileReader>* file_reader,
+    bool skip_footer_size_check) {
   assert(file_size);
   assert(file_reader);
 
@@ -102,27 +107,18 @@ Status BlobFileReader::OpenFile(
 
   constexpr IODebugContext* dbg = nullptr;
 
-  {
-    TEST_SYNC_POINT("BlobFileReader::OpenFile:GetFileSize");
-
-    const Status s =
-        fs->GetFileSize(blob_file_path, IOOptions(), file_size, dbg);
-    if (!s.ok()) {
-      return s;
-    }
-  }
-
-  if (*file_size < BlobLogHeader::kSize + BlobLogFooter::kSize) {
-    return Status::Corruption("Malformed blob file");
-  }
-
   std::unique_ptr<FSRandomAccessFile> file;
+  FileOptions reader_file_opts = file_opts;
+
+  if (skip_footer_size_check && reader_file_opts.use_direct_reads) {
+    reader_file_opts.use_direct_reads = false;
+  }
 
   {
     TEST_SYNC_POINT("BlobFileReader::OpenFile:NewRandomAccessFile");
 
     const Status s =
-        fs->NewRandomAccessFile(blob_file_path, file_opts, &file, dbg);
+        fs->NewRandomAccessFile(blob_file_path, reader_file_opts, &file, dbg);
     if (!s.ok()) {
       return s;
     }
@@ -130,6 +126,24 @@ Status BlobFileReader::OpenFile(
 
   assert(file);
 
+  {
+    Status s = GetFileSizeFromOpenFileOrPath(
+        file.get(), fs, blob_file_path, file_size, dbg,
+        FileSizeFallback::kNotSupportedOnly,
+        []() { TEST_SYNC_POINT("BlobFileReader::OpenFile:GetFileSize"); });
+    if (!s.ok()) {
+      return s;
+    }
+  }
+
+  if (!skip_footer_size_check &&
+      *file_size < BlobLogHeader::kSize + BlobLogFooter::kSize) {
+    return Status::Corruption("Malformed blob file");
+  }
+  if (skip_footer_size_check && *file_size < BlobLogHeader::kSize) {
+    return Status::Corruption("Malformed blob file");
+  }
+
   if (immutable_options.advise_random_on_open) {
     file->Hint(FSRandomAccessFile::kRandom);
   }
@@ -153,7 +167,7 @@ Status BlobFileReader::ReadHeader(const RandomAccessFileReader* file_reader,
 
   Slice header_slice;
   Buffer buf;
-  AlignedBuf aligned_buf;
+  AlignedBuffer direct_io_buffer;
 
   {
     TEST_SYNC_POINT("BlobFileReader::ReadHeader:ReadFromFile");
@@ -163,7 +177,7 @@ Status BlobFileReader::ReadHeader(const RandomAccessFileReader* file_reader,
 
     const Status s =
         ReadFromFile(file_reader, read_options, read_offset, read_size,
-                     statistics, &header_slice, &buf, &aligned_buf);
+                     statistics, &header_slice, &buf, &direct_io_buffer);
     if (!s.ok()) {
       return s;
     }
@@ -204,7 +218,7 @@ Status BlobFileReader::ReadFooter(const RandomAccessFileReader* file_reader,
 
   Slice footer_slice;
   Buffer buf;
-  AlignedBuf aligned_buf;
+  AlignedBuffer direct_io_buffer;
 
   {
     TEST_SYNC_POINT("BlobFileReader::ReadFooter:ReadFromFile");
@@ -214,7 +228,7 @@ Status BlobFileReader::ReadFooter(const RandomAccessFileReader* file_reader,
 
     const Status s =
         ReadFromFile(file_reader, read_options, read_offset, read_size,
-                     statistics, &footer_slice, &buf, &aligned_buf);
+                     statistics, &footer_slice, &buf, &direct_io_buffer);
     if (!s.ok()) {
       return s;
     }
@@ -245,10 +259,11 @@ Status BlobFileReader::ReadFromFile(const RandomAccessFileReader* file_reader,
                                     const ReadOptions& read_options,
                                     uint64_t read_offset, size_t read_size,
                                     Statistics* statistics, Slice* slice,
-                                    Buffer* buf, AlignedBuf* aligned_buf) {
+                                    Buffer* buf,
+                                    AlignedBuffer* direct_io_buffer) {
   assert(slice);
   assert(buf);
-  assert(aligned_buf);
+  assert(direct_io_buffer);
 
   assert(file_reader);
 
@@ -265,15 +280,15 @@ Status BlobFileReader::ReadFromFile(const RandomAccessFileReader* file_reader,
 
   if (file_reader->use_direct_io()) {
     constexpr char* scratch = nullptr;
+    AlignedBufferAllocationContext direct_io_context{direct_io_buffer};
 
     s = file_reader->Read(io_options, read_offset, read_size, slice, scratch,
-                          aligned_buf, &dbg);
+                          &direct_io_context, &dbg);
   } else {
     buf->reset(new char[read_size]);
-    constexpr AlignedBuf* aligned_scratch = nullptr;
 
     s = file_reader->Read(io_options, read_offset, read_size, slice, buf->get(),
-                          aligned_scratch, &dbg);
+                          nullptr, &dbg);
   }
 
   if (!s.ok()) {
@@ -291,13 +306,14 @@ BlobFileReader::BlobFileReader(
     std::unique_ptr<RandomAccessFileReader>&& file_reader, uint64_t file_size,
     CompressionType compression_type,
     std::shared_ptr<Decompressor> decompressor, SystemClock* clock,
-    Statistics* statistics)
+    Statistics* statistics, bool has_footer)
     : file_reader_(std::move(file_reader)),
       file_size_(file_size),
       compression_type_(compression_type),
       decompressor_(std::move(decompressor)),
       clock_(clock),
-      statistics_(statistics) {
+      statistics_(statistics),
+      has_footer_(has_footer) {
   assert(file_reader_);
 }
 
@@ -312,7 +328,8 @@ Status BlobFileReader::GetBlob(
 
   const uint64_t key_size = user_key.size();
 
-  if (!IsValidBlobOffset(offset, key_size, value_size, file_size_)) {
+  if (!IsValidBlobOffset(offset, key_size, value_size, file_size_,
+                         has_footer_)) {
     return Status::Corruption("Invalid blob offset");
   }
 
@@ -335,7 +352,7 @@ Status BlobFileReader::GetBlob(
 
   Slice record_slice;
   Buffer buf;
-  AlignedBuf aligned_buf;
+  AlignedBuffer direct_io_buffer;
 
   bool prefetched = false;
 
@@ -365,7 +382,7 @@ Status BlobFileReader::GetBlob(
     const Status s =
         ReadFromFile(file_reader_.get(), read_options, record_offset,
                      static_cast<size_t>(record_size), statistics_,
-                     &record_slice, &buf, &aligned_buf);
+                     &record_slice, &buf, &direct_io_buffer);
     if (!s.ok()) {
       return s;
     }
@@ -428,7 +445,8 @@ void BlobFileReader::MultiGetBlob(
     const uint64_t offset = req->offset;
     const uint64_t value_size = req->len;
 
-    if (!IsValidBlobOffset(offset, key_size, value_size, file_size_)) {
+    if (!IsValidBlobOffset(offset, key_size, value_size, file_size_,
+                           has_footer_)) {
       *req->status = Status::Corruption("Invalid blob offset");
       continue;
     }
@@ -454,8 +472,15 @@ void BlobFileReader::MultiGetBlob(
 
   RecordTick(statistics_, BLOB_DB_BLOB_FILE_BYTES_READ, total_len);
 
+  if (read_reqs.empty()) {
+    if (bytes_read) {
+      *bytes_read = 0;
+    }
+    return;
+  }
+
   Buffer buf;
-  AlignedBuf aligned_buf;
+  AlignedBuffer direct_io_buffer;
 
   Status s;
   bool direct_io = file_reader_->use_direct_io();
@@ -478,8 +503,9 @@ void BlobFileReader::MultiGetBlob(
   IODebugContext dbg;
   s = file_reader_->PrepareIOOptions(read_options, opts, &dbg);
   if (s.ok()) {
+    AlignedBufferAllocationContext direct_io_context{&direct_io_buffer};
     s = file_reader_->MultiRead(opts, read_reqs.data(), read_reqs.size(),
-                                direct_io ? &aligned_buf : nullptr, &dbg);
+                                &direct_io_context, &dbg);
   }
   if (!s.ok()) {
     for (auto& req : read_reqs) {

db/blob/blob_file_reader.h

@@ -36,7 +36,20 @@ class BlobFileReader {
                        HistogramImpl* blob_file_read_hist,
                        uint64_t blob_file_number,
                        const std::shared_ptr<IOTracer>& io_tracer,
-                       std::unique_ptr<BlobFileReader>* reader);
+                       std::unique_ptr<BlobFileReader>* reader) {
+    return Create(immutable_options, read_options, file_options,
+                  column_family_id, blob_file_read_hist, blob_file_number,
+                  io_tracer, /*skip_footer_validation=*/false, reader);
+  }
+
+  // Allows opening in-flight direct-write blob files by optionally skipping
+  // footer validation.
+  static Status Create(
+      const ImmutableOptions& immutable_options,
+      const ReadOptions& read_options, const FileOptions& file_options,
+      uint32_t column_family_id, HistogramImpl* blob_file_read_hist,
+      uint64_t blob_file_number, const std::shared_ptr<IOTracer>& io_tracer,
+      bool skip_footer_validation, std::unique_ptr<BlobFileReader>* reader);
 
   BlobFileReader(const BlobFileReader&) = delete;
   BlobFileReader& operator=(const BlobFileReader&) = delete;
@@ -63,18 +76,22 @@ class BlobFileReader {
   uint64_t GetFileSize() const { return file_size_; }
 
  private:
+  // `has_footer` tracks whether offset validation should reserve footer space.
   BlobFileReader(std::unique_ptr<RandomAccessFileReader>&& file_reader,
                  uint64_t file_size, CompressionType compression_type,
                  std::shared_ptr<Decompressor> decompressor, SystemClock* clock,
-                 Statistics* statistics);
+                 Statistics* statistics, bool has_footer);
 
+  // `skip_footer_size_check` is used for direct-write files that are still
+  // missing their footer at open time.
   static Status OpenFile(const ImmutableOptions& immutable_options,
                          const FileOptions& file_opts,
                          HistogramImpl* blob_file_read_hist,
                          uint64_t blob_file_number,
                          const std::shared_ptr<IOTracer>& io_tracer,
                          uint64_t* file_size,
-                         std::unique_ptr<RandomAccessFileReader>* file_reader);
+                         std::unique_ptr<RandomAccessFileReader>* file_reader,
+                         bool skip_footer_size_check);
 
   static Status ReadHeader(const RandomAccessFileReader* file_reader,
                            const ReadOptions& read_options,
@@ -91,7 +108,7 @@ class BlobFileReader {
                              const ReadOptions& read_options,
                              uint64_t read_offset, size_t read_size,
                              Statistics* statistics, Slice* slice, Buffer* buf,
-                             AlignedBuf* aligned_buf);
+                             AlignedBuffer* direct_io_buffer);
 
   static Status VerifyBlob(const Slice& record_slice, const Slice& user_key,
                            uint64_t value_size);
@@ -110,6 +127,8 @@ class BlobFileReader {
   std::shared_ptr<Decompressor> decompressor_;
   SystemClock* clock_;
   Statistics* statistics_;
+  // False when the reader was opened before the blob file footer was written.
+  bool has_footer_;
 };
 
 }  // namespace ROCKSDB_NAMESPACE

db/blob/blob_file_reader_test.cc

@@ -7,10 +7,12 @@
 
 #include <cassert>
 #include <string>
+#include <utility>
 
 #include "db/blob/blob_contents.h"
 #include "db/blob/blob_log_format.h"
 #include "db/blob/blob_log_writer.h"
+#include "env/composite_env_wrapper.h"
 #include "env/mock_env.h"
 #include "file/filename.h"
 #include "file/read_write_util.h"
@@ -136,6 +138,167 @@ class BlobFileReaderTest : public testing::Test {
   std::unique_ptr<Env> mock_env_;
 };
 
+namespace {
+
+// Returns a stale path-level size for one target blob file while leaving the
+// opened file handle untouched. This lets the test verify
+// BlobFileReader::Create prefers the opened handle's size when the path stat
+// lags behind.
+class StalePathSizeFileSystem : public FileSystemWrapper {
+ public:
+  static const char* kClassName() { return "StalePathSizeFileSystem"; }
+
+  StalePathSizeFileSystem(const std::shared_ptr<FileSystem>& target,
+                          std::string stale_path)
+      : FileSystemWrapper(target), stale_path_(std::move(stale_path)) {}
+
+  const char* Name() const override { return kClassName(); }
+
+  IOStatus GetFileSize(const std::string& fname, const IOOptions& options,
+                       uint64_t* file_size, IODebugContext* dbg) override {
+    if (fname == stale_path_) {
+      *file_size = 0;
+      return IOStatus::OK();
+    }
+    return FileSystemWrapper::GetFileSize(fname, options, file_size, dbg);
+  }
+
+ private:
+  const std::string stale_path_;
+};
+
+// Makes opened blob file handles fail GetFileSize() with a hard error so the
+// reader path can verify that such errors are propagated directly.
+class FailingOpenFileSizeRandomAccessFile
+    : public FSRandomAccessFileOwnerWrapper {
+ public:
+  explicit FailingOpenFileSizeRandomAccessFile(
+      std::unique_ptr<FSRandomAccessFile>&& target)
+      : FSRandomAccessFileOwnerWrapper(std::move(target)) {}
+
+  IOStatus GetFileSize(uint64_t* /*result*/) override {
+    return IOStatus::IOError("open file size failed");
+  }
+};
+
+// Makes opened blob file handles report GetFileSize() as unsupported so the
+// reader path exercises its path-level fallback logic.
+class NotSupportedOpenFileSizeRandomAccessFile
+    : public FSRandomAccessFileOwnerWrapper {
+ public:
+  explicit NotSupportedOpenFileSizeRandomAccessFile(
+      std::unique_ptr<FSRandomAccessFile>&& target)
+      : FSRandomAccessFileOwnerWrapper(std::move(target)) {}
+
+  IOStatus GetFileSize(uint64_t* /*result*/) override {
+    return IOStatus::NotSupported("open file size not supported");
+  }
+};
+
+class OpenFileSizeFileSystem : public FileSystemWrapper {
+ public:
+  OpenFileSizeFileSystem(const std::shared_ptr<FileSystem>& target,
+                         std::string target_path)
+      : FileSystemWrapper(target), target_path_(std::move(target_path)) {}
+
+  IOStatus NewRandomAccessFile(const std::string& fname,
+                               const FileOptions& options,
+                               std::unique_ptr<FSRandomAccessFile>* result,
+                               IODebugContext* dbg) override {
+    std::unique_ptr<FSRandomAccessFile> file;
+    IOStatus s =
+        FileSystemWrapper::NewRandomAccessFile(fname, options, &file, dbg);
+    if (!s.ok()) {
+      return IOStatus(std::move(s));
+    }
+    if (fname == target_path_) {
+      *result = WrapTargetFile(std::move(file));
+    } else {
+      *result = std::move(file);
+    }
+    return IOStatus::OK();
+  }
+
+ protected:
+  virtual std::unique_ptr<FSRandomAccessFile> WrapTargetFile(
+      std::unique_ptr<FSRandomAccessFile>&& file) = 0;
+
+ private:
+  const std::string target_path_;
+};
+
+// Wraps the target blob file in FailingOpenFileSizeRandomAccessFile while
+// leaving all other files alone.
+class FailingOpenFileSizeFileSystem : public OpenFileSizeFileSystem {
+ public:
+  static const char* kClassName() { return "FailingOpenFileSizeFileSystem"; }
+
+  FailingOpenFileSizeFileSystem(const std::shared_ptr<FileSystem>& target,
+                                std::string target_path)
+      : OpenFileSizeFileSystem(target, std::move(target_path)) {}
+
+  const char* Name() const override { return kClassName(); }
+
+ private:
+  std::unique_ptr<FSRandomAccessFile> WrapTargetFile(
+      std::unique_ptr<FSRandomAccessFile>&& file) override {
+    return std::unique_ptr<FSRandomAccessFile>(
+        new FailingOpenFileSizeRandomAccessFile(std::move(file)));
+  }
+};
+
+// Wraps the target blob file in NotSupportedOpenFileSizeRandomAccessFile while
+// leaving all other files alone.
+class NotSupportedOpenFileSizeFileSystem : public OpenFileSizeFileSystem {
+ public:
+  static const char* kClassName() {
+    return "NotSupportedOpenFileSizeFileSystem";
+  }
+
+  NotSupportedOpenFileSizeFileSystem(const std::shared_ptr<FileSystem>& target,
+                                     std::string target_path)
+      : OpenFileSizeFileSystem(target, std::move(target_path)) {}
+
+  const char* Name() const override { return kClassName(); }
+
+ private:
+  std::unique_ptr<FSRandomAccessFile> WrapTargetFile(
+      std::unique_ptr<FSRandomAccessFile>&& file) override {
+    return std::unique_ptr<FSRandomAccessFile>(
+        new NotSupportedOpenFileSizeRandomAccessFile(std::move(file)));
+  }
+};
+
+// Forces the target blob file onto the NotSupported open-handle-size branch and
+// then fails the fallback path-level GetFileSize() call.
+class FailingFallbackPathSizeFileSystem
+    : public NotSupportedOpenFileSizeFileSystem {
+ public:
+  static const char* kClassName() {
+    return "FailingFallbackPathSizeFileSystem";
+  }
+
+  FailingFallbackPathSizeFileSystem(const std::shared_ptr<FileSystem>& target,
+                                    std::string target_path)
+      : NotSupportedOpenFileSizeFileSystem(target, target_path),
+        target_path_(std::move(target_path)) {}
+
+  const char* Name() const override { return kClassName(); }
+
+  IOStatus GetFileSize(const std::string& fname, const IOOptions& options,
+                       uint64_t* file_size, IODebugContext* dbg) override {
+    if (fname == target_path_) {
+      return IOStatus::IOError("fallback path size failed");
+    }
+    return FileSystemWrapper::GetFileSize(fname, options, file_size, dbg);
+  }
+
+ private:
+  const std::string target_path_;
+};
+
+}  // namespace
+
 TEST_F(BlobFileReaderTest, CreateReaderAndGetBlob) {
   Options options;
   options.env = mock_env_.get();
@@ -428,6 +591,211 @@ TEST_F(BlobFileReaderTest, CreateReaderAndGetBlob) {
   }
 }
 
+// Verifies Create() prefers the file size reported by the opened handle over a
+// stale path-level size query by returning a readable reader even when the path
+// stat intentionally reports 0 for the blob file.
+TEST_F(BlobFileReaderTest, CreateReaderUsesOpenedFileSizeWhenPathSizeIsStale) {
+  const std::string db_path = test::PerThreadDBPath(
+      mock_env_.get(),
+      "BlobFileReaderTest_CreateReaderUsesOpenedFileSizeWhenPathSizeIsStale");
+
+  constexpr uint64_t blob_file_number = 1;
+  const std::string blob_file_path = BlobFileName(db_path, blob_file_number);
+
+  auto stale_size_fs = std::make_shared<StalePathSizeFileSystem>(
+      mock_env_->GetFileSystem(), blob_file_path);
+  CompositeEnvWrapper stale_size_env(mock_env_.get(), stale_size_fs);
+
+  Options options;
+  options.env = &stale_size_env;
+  options.cf_paths.emplace_back(db_path, 0);
+  options.enable_blob_files = true;
+
+  ImmutableOptions immutable_options(options);
+
+  constexpr uint32_t column_family_id = 1;
+  constexpr bool has_ttl = false;
+  constexpr ExpirationRange expiration_range;
+  constexpr char key[] = "key";
+  constexpr char blob[] = "blob";
+
+  uint64_t blob_offset = 0;
+  uint64_t blob_size = 0;
+  WriteBlobFile(immutable_options, column_family_id, has_ttl, expiration_range,
+                expiration_range, blob_file_number, key, blob, kNoCompression,
+                &blob_offset, &blob_size);
+
+  std::unique_ptr<BlobFileReader> reader;
+  ReadOptions read_options;
+  read_options.verify_checksums = false;
+  ASSERT_OK(BlobFileReader::Create(
+      immutable_options, read_options, FileOptions(), column_family_id,
+      /*blob_file_read_hist=*/nullptr, blob_file_number, nullptr /*IOTracer*/,
+      &reader));
+  ASSERT_NE(reader, nullptr);
+
+  std::unique_ptr<BlobContents> value;
+  uint64_t bytes_read = 0;
+  ASSERT_OK(reader->GetBlob(read_options, key, blob_offset, blob_size,
+                            kNoCompression,
+                            /*prefetch_buffer=*/nullptr,
+                            /*allocator=*/nullptr, &value, &bytes_read));
+  ASSERT_NE(value, nullptr);
+  ASSERT_EQ(value->data(), Slice(blob));
+  ASSERT_EQ(bytes_read, blob_size);
+}
+
+// Verifies Create() falls back to the path-level GetFileSize() call when the
+// opened file handle reports that size queries are unsupported.
+TEST_F(BlobFileReaderTest,
+       CreateReaderFallsBackToPathSizeWhenOpenedFileSizeNotSupported) {
+  const std::string db_path = test::PerThreadDBPath(
+      mock_env_.get(),
+      "BlobFileReaderTest_"
+      "CreateReaderFallsBackToPathSizeWhenOpenedFileSizeNotSupported");
+
+  constexpr uint64_t blob_file_number = 1;
+  const std::string blob_file_path = BlobFileName(db_path, blob_file_number);
+
+  auto not_supported_open_file_size_fs =
+      std::make_shared<NotSupportedOpenFileSizeFileSystem>(
+          mock_env_->GetFileSystem(), blob_file_path);
+  CompositeEnvWrapper not_supported_open_file_size_env(
+      mock_env_.get(), not_supported_open_file_size_fs);
+
+  Options options;
+  options.env = &not_supported_open_file_size_env;
+  options.cf_paths.emplace_back(db_path, 0);
+  options.enable_blob_files = true;
+
+  ImmutableOptions immutable_options(options);
+
+  constexpr uint32_t column_family_id = 1;
+  constexpr bool has_ttl = false;
+  constexpr ExpirationRange expiration_range;
+  constexpr char key[] = "key";
+  constexpr char blob[] = "blob";
+
+  uint64_t blob_offset = 0;
+  uint64_t blob_size = 0;
+  WriteBlobFile(immutable_options, column_family_id, has_ttl, expiration_range,
+                expiration_range, blob_file_number, key, blob, kNoCompression,
+                &blob_offset, &blob_size);
+
+  std::unique_ptr<BlobFileReader> reader;
+  ReadOptions read_options;
+  read_options.verify_checksums = false;
+  ASSERT_OK(BlobFileReader::Create(
+      immutable_options, read_options, FileOptions(), column_family_id,
+      /*blob_file_read_hist=*/nullptr, blob_file_number, nullptr /*IOTracer*/,
+      &reader));
+  ASSERT_NE(reader, nullptr);
+
+  std::unique_ptr<BlobContents> value;
+  uint64_t bytes_read = 0;
+  ASSERT_OK(reader->GetBlob(read_options, key, blob_offset, blob_size,
+                            kNoCompression,
+                            /*prefetch_buffer=*/nullptr,
+                            /*allocator=*/nullptr, &value, &bytes_read));
+  ASSERT_NE(value, nullptr);
+  ASSERT_EQ(value->data(), Slice(blob));
+  ASSERT_EQ(bytes_read, blob_size);
+}
+
+// Verifies Create() propagates a real error from the opened file handle's
+// GetFileSize() rather than masking it with any path-level fallback.
+TEST_F(BlobFileReaderTest, CreateReaderPropagatesOpenedFileSizeError) {
+  const std::string db_path = test::PerThreadDBPath(
+      mock_env_.get(),
+      "BlobFileReaderTest_CreateReaderPropagatesOpenedFileSizeError");
+
+  constexpr uint64_t blob_file_number = 1;
+  const std::string blob_file_path = BlobFileName(db_path, blob_file_number);
+
+  auto failing_open_file_size_fs =
+      std::make_shared<FailingOpenFileSizeFileSystem>(
+          mock_env_->GetFileSystem(), blob_file_path);
+  CompositeEnvWrapper failing_open_file_size_env(mock_env_.get(),
+                                                 failing_open_file_size_fs);
+
+  Options options;
+  options.env = &failing_open_file_size_env;
+  options.cf_paths.emplace_back(db_path, 0);
+  options.enable_blob_files = true;
+
+  ImmutableOptions immutable_options(options);
+
+  constexpr uint32_t column_family_id = 1;
+  constexpr bool has_ttl = false;
+  constexpr ExpirationRange expiration_range;
+  constexpr char key[] = "key";
+  constexpr char blob[] = "blob";
+
+  uint64_t blob_offset = 0;
+  uint64_t blob_size = 0;
+  WriteBlobFile(immutable_options, column_family_id, has_ttl, expiration_range,
+                expiration_range, blob_file_number, key, blob, kNoCompression,
+                &blob_offset, &blob_size);
+
+  std::unique_ptr<BlobFileReader> reader;
+  ReadOptions read_options;
+  Status s = BlobFileReader::Create(
+      immutable_options, read_options, FileOptions(), column_family_id,
+      /*blob_file_read_hist=*/nullptr, blob_file_number, nullptr /*IOTracer*/,
+      &reader);
+  ASSERT_TRUE(s.IsIOError());
+  ASSERT_EQ(reader, nullptr);
+}
+
+// Verifies Create() also propagates errors from the path-level fallback size
+// query when the opened file handle only reports NotSupported for GetFileSize.
+TEST_F(
+    BlobFileReaderTest,
+    CreateReaderPropagatesFallbackPathSizeErrorWhenOpenedFileSizeNotSupported) {
+  const std::string db_path =
+      test::PerThreadDBPath(mock_env_.get(),
+                            "BlobFileReaderTest_"
+                            "CreateReaderPropagatesFallbackPathSizeErrorWhenOpe"
+                            "nedFileSizeNotSupported");
+
+  constexpr uint64_t blob_file_number = 1;
+  const std::string blob_file_path = BlobFileName(db_path, blob_file_number);
+
+  auto failing_fallback_path_size_fs =
+      std::make_shared<FailingFallbackPathSizeFileSystem>(
+          mock_env_->GetFileSystem(), blob_file_path);
+  CompositeEnvWrapper failing_fallback_path_size_env(
+      mock_env_.get(), failing_fallback_path_size_fs);
+
+  Options options;
+  options.env = &failing_fallback_path_size_env;
+  options.cf_paths.emplace_back(db_path, 0);
+  options.enable_blob_files = true;
+
+  ImmutableOptions immutable_options(options);
+
+  constexpr uint32_t column_family_id = 1;
+  constexpr bool has_ttl = false;
+  constexpr ExpirationRange expiration_range;
+  constexpr char key[] = "key";
+  constexpr char blob[] = "blob";
+
+  uint64_t blob_offset = 0;
+  uint64_t blob_size = 0;
+  WriteBlobFile(immutable_options, column_family_id, has_ttl, expiration_range,
+                expiration_range, blob_file_number, key, blob, kNoCompression,
+                &blob_offset, &blob_size);
+
+  std::unique_ptr<BlobFileReader> reader;
+  ReadOptions read_options;
+  Status s = BlobFileReader::Create(
+      immutable_options, read_options, FileOptions(), column_family_id,
+      /*blob_file_read_hist=*/nullptr, blob_file_number, nullptr /*IOTracer*/,
+      &reader);
+  ASSERT_TRUE(s.IsIOError());
+  ASSERT_EQ(reader, nullptr);
+}
+
 TEST_F(BlobFileReaderTest, Malformed) {
   // Write a blob file consisting of nothing but a header, and make sure we
   // detect the error when we open it for reading
@@ -849,7 +1217,6 @@ class BlobFileReaderIOErrorTest
 
 INSTANTIATE_TEST_CASE_P(BlobFileReaderTest, BlobFileReaderIOErrorTest,
                         ::testing::ValuesIn(std::vector<std::string>{
-                            "BlobFileReader::OpenFile:GetFileSize",
                             "BlobFileReader::OpenFile:NewRandomAccessFile",
                             "BlobFileReader::ReadHeader:ReadFromFile",
                             "BlobFileReader::ReadFooter:ReadFromFile",

db/blob/blob_garbage_meter.cc

@@ -8,69 +8,50 @@
 #include "db/blob/blob_index.h"
 #include "db/blob/blob_log_format.h"
 #include "db/dbformat.h"
+#include "db/wide/wide_column_serialization.h"
 
 namespace ROCKSDB_NAMESPACE {
 
 Status BlobGarbageMeter::ProcessInFlow(const Slice& key, const Slice& value) {
-  uint64_t blob_file_number = kInvalidBlobFileNumber;
-  uint64_t bytes = 0;
-
-  const Status s = Parse(key, value, &blob_file_number, &bytes);
-  if (!s.ok()) {
-    return s;
-  }
-
-  if (blob_file_number == kInvalidBlobFileNumber) {
-    return Status::OK();
-  }
-
-  flows_[blob_file_number].AddInFlow(bytes);
-
-  return Status::OK();
+  return ProcessFlow(key, value, /*is_inflow=*/true);
 }
 
 Status BlobGarbageMeter::ProcessOutFlow(const Slice& key, const Slice& value) {
-  uint64_t blob_file_number = kInvalidBlobFileNumber;
-  uint64_t bytes = 0;
-
-  const Status s = Parse(key, value, &blob_file_number, &bytes);
-  if (!s.ok()) {
-    return s;
-  }
-
-  if (blob_file_number == kInvalidBlobFileNumber) {
-    return Status::OK();
-  }
-
-  // Note: in order to measure the amount of additional garbage, we only need to
-  // track the outflow for preexisting files, i.e. those that also had inflow.
-  // (Newly written files would only have outflow.)
-  auto it = flows_.find(blob_file_number);
-  if (it == flows_.end()) {
-    return Status::OK();
-  }
-
-  it->second.AddOutFlow(bytes);
-
-  return Status::OK();
+  return ProcessFlow(key, value, /*is_inflow=*/false);
 }
 
-Status BlobGarbageMeter::Parse(const Slice& key, const Slice& value,
-                               uint64_t* blob_file_number, uint64_t* bytes) {
+Status BlobGarbageMeter::GetBlobReferenceDetails(const ParsedInternalKey& ikey,
+                                                 const BlobIndex& blob_index,
+                                                 uint64_t* blob_file_number,
+                                                 uint64_t* bytes) {
   assert(blob_file_number);
   assert(*blob_file_number == kInvalidBlobFileNumber);
   assert(bytes);
   assert(*bytes == 0);
 
-  ParsedInternalKey ikey;
-
-  {
-    constexpr bool log_err_key = false;
-    const Status s = ParseInternalKey(key, &ikey, log_err_key);
-    if (!s.ok()) {
-      return s;
-    }
+  if (blob_index.IsInlined() || blob_index.HasTTL()) {
+    return Status::Corruption("Unexpected TTL/inlined blob index");
   }
+  if (blob_index.IsSameFile()) {
+    return Status::OK();
+  }
+
+  *blob_file_number = blob_index.file_number();
+  *bytes =
+      blob_index.size() +
+      BlobLogRecord::CalculateAdjustmentForRecordHeader(ikey.user_key.size());
+
+  return Status::OK();
+}
+
+Status BlobGarbageMeter::ParseBlobIndexReference(const ParsedInternalKey& ikey,
+                                                 const Slice& value,
+                                                 uint64_t* blob_file_number,
+                                                 uint64_t* bytes) {
+  assert(blob_file_number);
+  assert(*blob_file_number == kInvalidBlobFileNumber);
+  assert(bytes);
+  assert(*bytes == 0);
 
   if (ikey.type != kTypeBlobIndex) {
     return Status::OK();
@@ -85,16 +66,73 @@ Status BlobGarbageMeter::Parse(const Slice& key, const Slice& value,
     }
   }
 
-  if (blob_index.IsInlined() || blob_index.HasTTL()) {
-    return Status::Corruption("Unexpected TTL/inlined blob index");
+  return GetBlobReferenceDetails(ikey, blob_index, blob_file_number, bytes);
+}
+
+void BlobGarbageMeter::AddFlow(uint64_t blob_file_number, uint64_t bytes,
+                               bool is_inflow) {
+  if (is_inflow) {
+    flows_[blob_file_number].AddInFlow(bytes);
+    return;
   }
 
-  *blob_file_number = blob_index.file_number();
-  *bytes =
-      blob_index.size() +
-      BlobLogRecord::CalculateAdjustmentForRecordHeader(ikey.user_key.size());
+  // Note: in order to measure the amount of additional garbage, we only need to
+  // track the outflow for preexisting files, i.e. those that also had inflow.
+  // (Newly written files would only have outflow.)
+  auto it = flows_.find(blob_file_number);
+  if (it != flows_.end()) {
+    it->second.AddOutFlow(bytes);
+  }
+}
 
-  return Status::OK();
+Status BlobGarbageMeter::ProcessFlow(const Slice& key, const Slice& value,
+                                     bool is_inflow) {
+  ParsedInternalKey ikey;
+
+  {
+    constexpr bool log_err_key = false;
+    const Status s = ParseInternalKey(key, &ikey, log_err_key);
+    if (!s.ok()) {
+      return s;
+    }
+  }
+
+  uint64_t blob_file_number = kInvalidBlobFileNumber;
+  uint64_t bytes = 0;
+  if (Status s =
+          ParseBlobIndexReference(ikey, value, &blob_file_number, &bytes);
+      !s.ok()) {
+    return s;
+  }
+
+  if (blob_file_number != kInvalidBlobFileNumber) {
+    AddFlow(blob_file_number, bytes, is_inflow);
+    return Status::OK();
+  }
+
+  return ProcessEntityBlobReferences(ikey, value, is_inflow);
+}
+
+Status BlobGarbageMeter::ProcessEntityBlobReferences(
+    const ParsedInternalKey& ikey, const Slice& value, bool is_inflow) {
+  if (ikey.type != kTypeWideColumnEntity) {
+    return Status::OK();
+  }
+
+  return WideColumnSerialization::ForEachBlobFileNumber(
+      value, [&](const BlobIndex& blob_index) -> Status {
+        uint64_t file_number = kInvalidBlobFileNumber;
+        uint64_t blob_bytes = 0;
+        if (const Status s = GetBlobReferenceDetails(ikey, blob_index,
+                                                     &file_number, &blob_bytes);
+            !s.ok()) {
+          return s;
+        }
+        if (file_number != kInvalidBlobFileNumber) {
+          AddFlow(file_number, blob_bytes, is_inflow);
+        }
+        return Status::OK();
+      });
 }
 
 }  // namespace ROCKSDB_NAMESPACE

db/blob/blob_garbage_meter.h

@@ -15,6 +15,8 @@
 
 namespace ROCKSDB_NAMESPACE {
 
+class BlobIndex;
+struct ParsedInternalKey;
 class Slice;
 
 // A class that can be used to compute the amount of additional garbage
@@ -93,8 +95,19 @@ class BlobGarbageMeter {
   }
 
  private:
-  static Status Parse(const Slice& key, const Slice& value,
-                      uint64_t* blob_file_number, uint64_t* bytes);
+  static Status GetBlobReferenceDetails(const ParsedInternalKey& ikey,
+                                        const BlobIndex& blob_index,
+                                        uint64_t* blob_file_number,
+                                        uint64_t* bytes);
+  static Status ParseBlobIndexReference(const ParsedInternalKey& ikey,
+                                        const Slice& value,
+                                        uint64_t* blob_file_number,
+                                        uint64_t* bytes);
+
+  void AddFlow(uint64_t blob_file_number, uint64_t bytes, bool is_inflow);
+  Status ProcessFlow(const Slice& key, const Slice& value, bool is_inflow);
+  Status ProcessEntityBlobReferences(const ParsedInternalKey& ikey,
+                                     const Slice& value, bool is_inflow);
 
   std::unordered_map<uint64_t, BlobInOutFlow> flows_;
 };

db/blob/blob_garbage_meter_test.cc

@@ -11,10 +11,23 @@
 #include "db/blob/blob_index.h"
 #include "db/blob/blob_log_format.h"
 #include "db/dbformat.h"
+#include "db/wide/wide_column_serialization.h"
 #include "test_util/testharness.h"
 
 namespace ROCKSDB_NAMESPACE {
 
+namespace {
+
+BlobIndex MakeBlobIndex(uint64_t file_number, uint64_t offset, uint64_t size) {
+  std::string encoded;
+  BlobIndex::EncodeBlob(&encoded, file_number, offset, size, kNoCompression);
+  BlobIndex blob_index;
+  EXPECT_OK(blob_index.DecodeFrom(encoded));
+  return blob_index;
+}
+
+}  // namespace
+
 TEST(BlobGarbageMeterTest, MeasureGarbage) {
   BlobGarbageMeter blob_garbage_meter;
 
@@ -138,6 +151,24 @@ TEST(BlobGarbageMeterTest, PlainValue) {
   ASSERT_TRUE(blob_garbage_meter.flows().empty());
 }
 
+TEST(BlobGarbageMeterTest, SameFileBlobIndex) {
+  constexpr char user_key[] = "user_key";
+  constexpr SequenceNumber seq = 123;
+
+  const InternalKey key(user_key, seq, kTypeBlobIndex);
+  const Slice key_slice = key.Encode();
+
+  std::string value;
+  BlobIndex::EncodeBlob(&value, kCurrentFileBlobIndexFileNumber,
+                        /*offset=*/123, /*size=*/456, kNoCompression);
+
+  BlobGarbageMeter blob_garbage_meter;
+
+  ASSERT_OK(blob_garbage_meter.ProcessInFlow(key_slice, value));
+  ASSERT_OK(blob_garbage_meter.ProcessOutFlow(key_slice, value));
+  ASSERT_TRUE(blob_garbage_meter.flows().empty());
+}
+
 TEST(BlobGarbageMeterTest, CorruptInternalKey) {
   constexpr char corrupt_key[] = "i_am_corrupt";
   const Slice key_slice(corrupt_key);
@@ -188,6 +219,67 @@ TEST(BlobGarbageMeterTest, InlinedTTLBlobIndex) {
   ASSERT_NOK(blob_garbage_meter.ProcessOutFlow(key_slice, value_slice));
 }
 
+TEST(BlobGarbageMeterTest, WideColumnEntity) {
+  constexpr char user_key[] = "entity_key";
+  constexpr SequenceNumber seq = 123;
+  const InternalKey key(user_key, seq, kTypeWideColumnEntity);
+  const Slice key_slice = key.Encode();
+
+  const std::vector<std::pair<std::string, std::string>> columns = {
+      {"", "default_inline"},
+      {"blob_attr", "blob_inline"},
+      {"ttl", "00000001"},
+  };
+
+  const BlobIndex default_blob = MakeBlobIndex(4, 123, 555);
+  const BlobIndex attr_blob = MakeBlobIndex(5, 456, 777);
+
+  std::string inflow_value;
+  ASSERT_OK(WideColumnSerialization::SerializeV2(
+      columns, {{0, default_blob}, {1, attr_blob}}, inflow_value));
+
+  std::string outflow_value;
+  ASSERT_OK(WideColumnSerialization::SerializeV2(columns, {{0, default_blob}},
+                                                 outflow_value));
+
+  BlobGarbageMeter blob_garbage_meter;
+  ASSERT_OK(blob_garbage_meter.ProcessInFlow(key_slice, inflow_value));
+  ASSERT_OK(blob_garbage_meter.ProcessOutFlow(key_slice, outflow_value));
+
+  const auto& flows = blob_garbage_meter.flows();
+  ASSERT_EQ(flows.size(), 2);
+
+  constexpr size_t kUserKeySize = sizeof(user_key) - 1;
+  const uint64_t default_bytes =
+      default_blob.size() +
+      BlobLogRecord::CalculateAdjustmentForRecordHeader(kUserKeySize);
+  const uint64_t attr_bytes =
+      attr_blob.size() +
+      BlobLogRecord::CalculateAdjustmentForRecordHeader(kUserKeySize);
+
+  {
+    const auto it = flows.find(default_blob.file_number());
+    ASSERT_NE(it, flows.end());
+    ASSERT_EQ(it->second.GetInFlow().GetCount(), 1);
+    ASSERT_EQ(it->second.GetInFlow().GetBytes(), default_bytes);
+    ASSERT_EQ(it->second.GetOutFlow().GetCount(), 1);
+    ASSERT_EQ(it->second.GetOutFlow().GetBytes(), default_bytes);
+    ASSERT_FALSE(it->second.HasGarbage());
+  }
+
+  {
+    const auto it = flows.find(attr_blob.file_number());
+    ASSERT_NE(it, flows.end());
+    ASSERT_EQ(it->second.GetInFlow().GetCount(), 1);
+    ASSERT_EQ(it->second.GetInFlow().GetBytes(), attr_bytes);
+    ASSERT_EQ(it->second.GetOutFlow().GetCount(), 0);
+    ASSERT_EQ(it->second.GetOutFlow().GetBytes(), 0);
+    ASSERT_TRUE(it->second.HasGarbage());
+    ASSERT_EQ(it->second.GetGarbageCount(), 1);
+    ASSERT_EQ(it->second.GetGarbageBytes(), attr_bytes);
+  }
+}
+
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {