codebase-memory-mcp/.github/workflows/pr.yml

# PR validation: security gates + lint + the test legs this PR's changes need.
# `changes` runs scripts/ci/select-lanes.sh on the PR's file list; every lane
# job below runs only when selected (unknown paths, workflows, the harness,
# Makefile.cbm and vendored code select everything). Builds, smoke and soak at
# full breadth stay maintainer-driven via the dry-run workflow_dispatch. Branch
# protection requires `dco` + `ci-ok` — a single stable summary context that
# fails unless every PR stage succeeded and every selected lane ran green.
name: PR

on:
  pull_request:
    branches: [main]

permissions:
  contents: read
  pull-requests: read

# A new push to the PR supersedes the running validation — cancel it
# instead of stacking zombie pipelines.
concurrency:
  group: pr-${{ github.event.pull_request.number }}
  cancel-in-progress: true


jobs:
  security:
    needs: [changes]
    uses: ./.github/workflows/_security.yml
    # No job in _security.yml reads code scanning or the Actions API any more
    # (the CodeQL gate moved to codeql.yml), so the token stays read-only.
    permissions:
      contents: read
    secrets: inherit
    with:
      lanes: ${{ needs.changes.outputs.lanes }}

  lint:
    needs: [changes]
    if: ${{ contains(fromJSON(needs.changes.outputs.lanes), 'lint') || contains(fromJSON(needs.changes.outputs.lanes), 'lint-mem') }}
    uses: ./.github/workflows/_lint.yml
    with:
      lanes: ${{ needs.changes.outputs.lanes }}

  # ── Smart CI, PR CI only: run only the tests this change can reach.
  #    Active only while the repository variable TEST_IMPACT_MODE is `gate`;
  #    unset (shadow) every test runs and test-impact-shadow only reports.
  #    test-impact-select.sh --gate narrows only on a well-formed answer: a
  #    failure of this job, a run-all answer or a lane that selects nothing
  #    leaves `test` without a selection, and every test runs. ci-ok does not
  #    need this job: when it breaks, PRs cost more minutes, never coverage.
  #    Dry runs and releases never pass a selection.
  select-tests:
    needs: [changes]
    if: ${{ vars.TEST_IMPACT_MODE == 'gate' }}
    runs-on: ubuntu-24.04
    timeout-minutes: 30
    permissions:
      contents: read
      actions: read # the merge base's team-artifact bundle (test-impact-artifact.yml)
    outputs:
      selection: ${{ steps.select.outputs.selection }}
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          # Full history: the engine certifies the history it diffs and refuses
          # a shallow clone, so a depth-2 checkout always answered `all`.
          fetch-depth: 0
          persist-credentials: false
      - name: Build this checkout's engine
        run: |
          sudo apt-get update && sudo apt-get install -y zlib1g-dev
          scripts/build.sh CC=gcc CXX=g++
      - name: Select the tests
        id: select
        env:
          GH_TOKEN: ${{ github.token }}
        run: scripts/ci/test-impact-select.sh --binary build/c/codebase-memory-mcp --out "$RUNNER_TEMP/test-impact" --gate "$RUNNER_TEMP/test-selection"
      - name: Upload the selection
        if: ${{ steps.select.outputs.selection == 'narrowed' }}
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
        with:
          name: test-selection
          path: ${{ runner.temp }}/test-selection
          if-no-files-found: error
          overwrite: true
      # The selection a veto discarded (written only when one did): evidence
      # for the vetoes, never read by a test leg.
      - name: Upload the vetoed selection
        if: ${{ !cancelled() && steps.select.outcome != 'skipped' }}
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
        with:
          name: test-selection-vetoed
          path: ${{ runner.temp }}/test-impact/selection-vetoed.json
          if-no-files-found: ignore
          overwrite: true

  test:
    needs: [changes, lint, select-tests]
    if: ${{ !cancelled() && needs.lint.result == 'success' }}
    uses: ./.github/workflows/_test.yml
    with:
      lanes: ${{ needs.changes.outputs.lanes }}
      # Smart CI: only a successful select-tests run that narrowed passes a
      # selection; skipped (shadow), failed, run-all and nothing-selected
      # answers all run every test.
      test_selection: ${{ needs.select-tests.result == 'success' && needs.select-tests.outputs.selection == 'narrowed' && 'test-selection' || '' }}
      # Perf assertions are timing-sensitive on shared runners; they stay in
      # dry runs and releases where a flaky red does not block a merge.
      skip_perf: true
      # Iteration speed: split C-suite legs across parallel shard jobs
      # (coverage re-proven every run by the shard-completeness job). The PR
      # profile shards only arm64 (x2); x86 and Windows run whole, since each
      # shard re-pays its setup. Dry runs and releases keep x3 / x2.
      shard_suites: true
      shard_profile: pr

  # ── Which lanes does this PR need? The changed files go through
  #    scripts/ci/select-lanes.sh (tier, lanes, full, reasons); docs / CI /
  #    test-only PRs stay fast, anything it cannot place runs everything. ──
  changes:
    runs-on: ubuntu-latest
    timeout-minutes: 5
    outputs:
      tier: ${{ steps.select.outputs.tier }}
      full: ${{ steps.select.outputs.full }}
      lanes: ${{ steps.select.outputs.lanes }}
      smoke: ${{ steps.select.outputs.smoke }}
      codeql: ${{ steps.select.outputs.codeql }}
    steps:
      - name: List the changed files
        env:
          GH_TOKEN: ${{ github.token }}
          PR: ${{ github.event.pull_request.number }}
          REPO: ${{ github.repository }}
        run: |
          # The full .diff endpoint rejects large-but-valid PRs at 20k lines.
          # The paginated files endpoint remains filename-only for this gate.
          # A rename lists its old path too, as the fallback's --no-renames
          # diff does: moving a file away is a change to where it was.
          if FILES=$(gh api --paginate "repos/$REPO/pulls/$PR/files?per_page=100" --jq '.[] | .filename, (.previous_filename // empty)'); then
            echo "file list: pulls/files API"
          else
            # The files endpoint renders the diff server-side too, and answers
            # 422 "this diff is taking too long to generate" for a PR that
            # regenerates a vendored parser (#2246: a 42 MB sql/parser.c) --
            # so does the compare endpoint. The PR merge ref's first parent is
            # the base, so a name-only diff across it is the same file list.
            # depth 2 + blob:none fetches commits and trees only (well under
            # 1 MB, under a second); nothing from the PR is checked out or run.
            echo "::notice::pulls/files API failed; file list taken from refs/pull/$PR/merge"
            CHANGES_GIT="$RUNNER_TEMP/changes.git"
            git init -q --bare "$CHANGES_GIT"
            git --git-dir="$CHANGES_GIT" fetch -q --depth=2 --filter=blob:none \
              "https://github.com/$REPO" "refs/pull/$PR/merge"
            FILES=$(git --git-dir="$CHANGES_GIT" diff --name-only --no-renames FETCH_HEAD^1 FETCH_HEAD)
          fi
          printf '%s\n' "$FILES" > "$RUNNER_TEMP/changed-files.txt"
          cat "$RUNNER_TEMP/changed-files.txt"

      # The selector is PR code: it runs in its own step, after the
      # token-bearing one, from a checkout that keeps no credentials.
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          persist-credentials: false

      # The pulls/files API lists at most 3000 files: the PR's own count lets
      # the selector notice a cut list and select everything instead.
      - name: Select lanes
        id: select
        env:
          CHANGED_FILES: ${{ github.event.pull_request.changed_files }}
        run: |
          scripts/ci/select-lanes.sh --format github-output --expect-files "$CHANGED_FILES" "$RUNNER_TEMP/changed-files.txt" > "$RUNNER_TEMP/selection.txt"
          cat "$RUNNER_TEMP/selection.txt"
          cat "$RUNNER_TEMP/selection.txt" >> "$GITHUB_OUTPUT"
          # pr-smoke's legs, each tagged with its lane (kept in the job name,
          # where ci-ok reads it back); the job gets the selected ones.
          SMOKE='[{"os":"ubuntu-latest","lane":"smoke-ubuntu"},{"os":"macos-15","lane":"smoke-mac"},{"os":"windows-latest","lane":"smoke-win"}]'
          LANES=$(sed -n 's/^lanes=//p' "$RUNNER_TEMP/selection.txt")
          echo "smoke={\"include\":$(jq -c --argjson sel "$LANES" 'map(select(.lane | IN($sel[])))' <<< "$SMOKE")}" >> "$GITHUB_OUTPUT"
          # The CodeQL legs, tagged the same way (codeql.yml's full matrix has
          # the same entries, so PR and main share a configuration).
          CODEQL='[{"language":"c-cpp","build-mode":"manual","lane":"codeql-cpp"},{"language":"javascript-typescript","build-mode":"none","lane":"codeql-js"},{"language":"python","build-mode":"none","lane":"codeql-py"},{"language":"actions","build-mode":"none","lane":"codeql-actions"}]'
          echo "codeql={\"include\":$(jq -c --argjson sel "$LANES" 'map(select(.lane | IN($sel[])))' <<< "$CODEQL")}" >> "$GITHUB_OUTPUT"

  # ── Light smoke: build the PRODUCTION binary and run the core smoke on the
  #    three RELIABLE NATIVE platforms. Catches built-binary regressions at PR
  #    time (e.g. the Windows CreateProcess argv-quoting class that previously
  #    only surfaced in the release dry run). The full broad/emulated smoke stays
  #    in the dry run. Only the selected legs run (ubuntu from the extraction
  #    tier up, all three for platform/packaging changes); every leg gates.
  #    The maintained local wrappers add a race-free release fixture so the
  #    download/checksum/install/update phases run here as well. ──
  pr-smoke:
    needs: [changes]
    if: ${{ !cancelled() && needs.changes.result == 'success' && (contains(fromJSON(needs.changes.outputs.lanes), 'smoke-ubuntu') || contains(fromJSON(needs.changes.outputs.lanes), 'smoke-mac') || contains(fromJSON(needs.changes.outputs.lanes), 'smoke-win')) }}
    strategy:
      fail-fast: false
      matrix: ${{ fromJSON(needs.changes.outputs.smoke) }}
    runs-on: ${{ matrix.os }}
    timeout-minutes: 30
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1

      # SMOKE_REQUIRE_WRAPPERS=1 needs a Go toolchain for the Go wrapper; the
      # macOS runner images have none on PATH.
      - uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7.0.0
        with:
          go-version-file: pkg/go/go.mod
          cache: false

      - name: Install deps (Ubuntu)
        if: matrix.os == 'ubuntu-latest'
        run: sudo apt-get update && sudo apt-get install -y zlib1g-dev ccache

      - name: Install ccache (macOS)
        if: matrix.os == 'macos-15'
        run: command -v ccache >/dev/null 2>&1 || brew install ccache

      - uses: msys2/setup-msys2@66cd2cce69caa17b53920067426061ca1de3a884 # v2
        if: matrix.os == 'windows-latest'
        with:
          msystem: CLANG64
          path-type: inherit
          install: >-
            mingw-w64-clang-x86_64-clang
            mingw-w64-clang-x86_64-zlib
            mingw-w64-clang-x86_64-python3
            mingw-w64-clang-x86_64-ccache
            make
            coreutils
            curl
            zip
            unzip

      # Verified compiler cache — content-keyed, stale hits impossible by
      # construction (see scripts/env.sh).
      - name: Compiler cache (content-verified)
        uses: actions/cache@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6.1.0
        with:
          path: ${{ github.workspace }}/.ccache
          key: ccache-smoke-${{ matrix.os }}-${{ github.ref }}-${{ github.sha }}
          restore-keys: |
            ccache-smoke-${{ matrix.os }}-${{ github.ref }}-

      - name: Build prod + smoke (Ubuntu)
        if: matrix.os == 'ubuntu-latest'
        run: |
          scripts/build.sh CC=gcc CXX=g++
          scripts/smoke-local.sh "$(pwd)/build/c/codebase-memory-mcp"
        env:
          CCACHE_DIR: ${{ github.workspace }}/.ccache
          CCACHE_MAXSIZE: 1000M
          SMOKE_REQUIRE_WRAPPERS: "1"

      - name: Build prod + smoke (macOS)
        if: matrix.os == 'macos-15'
        run: |
          scripts/build.sh CC=cc CXX=c++
          codesign --sign - --force build/c/codebase-memory-mcp
          scripts/smoke-local.sh "$(pwd)/build/c/codebase-memory-mcp"
        env:
          CCACHE_DIR: ${{ github.workspace }}/.ccache
          CCACHE_MAXSIZE: 1000M
          SMOKE_REQUIRE_WRAPPERS: "1"

      - name: Build prod + smoke (Windows)
        if: matrix.os == 'windows-latest'
        shell: msys2 {0}
        run: |
          scripts/build.sh CC=clang CXX=clang++
          SMOKE_ARCH=amd64 bash test-infrastructure/vm/vm-smoke.sh
        env:
          CCACHE_DIR: ${{ github.workspace }}/.ccache
          CCACHE_MAXSIZE: 1000M
          SMOKE_REQUIRE_WRAPPERS: "1"

  # ── Waste sanitizer, REPORT-ONLY: memory + CPU waste on this repository and
  #    the per-function scaling lane (Linux). Deliberately NOT in ci-ok's needs:
  #    it uploads evidence and prints the ratchet, it never blocks a merge. Only
  #    PRs whose selection holds the memwaste lane (memory core, memwaste
  #    tooling, full runs) pay for it. ──
  memwaste:
    needs: [changes]
    if: ${{ !cancelled() && needs.changes.result == 'success' && contains(fromJSON(needs.changes.outputs.lanes), 'memwaste') }}
    uses: ./.github/workflows/_memwaste.yml
    with:
      mode: pr

  # ── Test-impact SHADOW, REPORT-ONLY: predicts from the code graph which C
  #    suites this change can affect (the latest released binary indexes the
  #    PR checkout; detect_changes at depth 15 against the base, mapped to
  #    suites, with the selector's run-all triggers) and uploads the
  #    prediction, so graph-based selection can be judged against what really
  #    failed before anything gates on it. continue-on-error and NOT in
  #    ci-ok's needs: it can never block or pass a merge. ──
  test-impact-shadow:
    runs-on: ubuntu-latest
    timeout-minutes: 45
    continue-on-error: true
    permissions:
      contents: read
      actions: read # the merge base's team-artifact bundle (test-impact-artifact.yml)
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          # Full history: the engine (test-impact select) certifies the history
          # it diffs and refuses a shallow clone ("Graft or shallow state
          # prevents history certification"), which made every shadow answer
          # run_all (GIT_UNAVAILABLE). About 400 MiB, this job only.
          fetch-depth: 0
          persist-credentials: false
      - name: Fetch the latest released binary
        id: release
        continue-on-error: true
        env:
          GH_TOKEN: ${{ github.token }}
        run: scripts/ci/test-impact-shadow.sh --fetch-release "$RUNNER_TEMP/cbm-bin"
      - name: Build the binary from this checkout (download failed)
        if: ${{ steps.release.outcome == 'failure' }}
        run: |
          sudo apt-get update && sudo apt-get install -y zlib1g-dev
          scripts/build.sh CC=gcc CXX=g++
          mkdir -p "$RUNNER_TEMP/cbm-bin"
          cp build/c/codebase-memory-mcp "$RUNNER_TEMP/cbm-bin/"
      - name: Predict the impacted suites
        run: scripts/ci/test-impact-shadow.sh --binary "$RUNNER_TEMP/cbm-bin/codebase-memory-mcp" --out "$RUNNER_TEMP/test-impact"
      # The engine (test-impact select) is this checkout's own code: the
      # released binary predates it, so the checkout is built either way.
      - name: Build this checkout's engine
        if: ${{ steps.release.outcome != 'failure' }}
        run: |
          sudo apt-get update && sudo apt-get install -y zlib1g-dev
          scripts/build.sh CC=gcc CXX=g++
      - name: Select with the team artifact (engine shadow)
        env:
          GH_TOKEN: ${{ github.token }}
        run: scripts/ci/test-impact-select.sh --binary build/c/codebase-memory-mcp --out "$RUNNER_TEMP/test-impact"
      - name: Upload the prediction
        if: ${{ always() }}
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
        with:
          name: test-impact-prediction
          path: |
            ${{ runner.temp }}/test-impact/prediction.json
            ${{ runner.temp }}/test-impact/selection.json
            ${{ runner.temp }}/test-impact/changed-files.txt
            ${{ runner.temp }}/test-impact/selection-engine.json
            ${{ runner.temp }}/test-impact/summary-engine.md
          if-no-files-found: warn

  # ── Docs-only and non-product PRs (tiers T0a, T0b) run the static contract
  #    steps of the canonical test leg on their own: they police docs
  #    surfaces (language counts, version metadata), test-infrastructure/ and
  #    lint config too. Every test leg runs the same steps. ──
  contracts:
    needs: [changes]
    if: ${{ !cancelled() && needs.changes.result == 'success' && contains(fromJSON(needs.changes.outputs.lanes), 'contracts') }}
    runs-on: ubuntu-latest
    timeout-minutes: 20
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
      - name: Contract steps only
        run: scripts/test.sh --contracts-only

  # ── CodeQL PR scan (lanes codeql-cpp / -js / -py / -actions): one leg per
  #    language whose first-party scope this PR changes; each analyses that
  #    language's whole first-party scope, c-cpp with the vendored objects
  #    compiled before tracing (except vendored files the PR changes), and
  #    reports alerts on the changed lines (diff-informed). A job of this run,
  #    needed by ci-ok -- nothing waits for a separate CodeQL run. Push to
  #    main, dry run and release run the full scan (codeql.yml). ──
  codeql:
    needs: [changes]
    if: ${{ !cancelled() && needs.changes.result == 'success' && (contains(fromJSON(needs.changes.outputs.lanes), 'codeql-cpp') || contains(fromJSON(needs.changes.outputs.lanes), 'codeql-js') || contains(fromJSON(needs.changes.outputs.lanes), 'codeql-py') || contains(fromJSON(needs.changes.outputs.lanes), 'codeql-actions')) }}
    uses: ./.github/workflows/codeql.yml
    # codeql.yml uploads results and reads alerts: a called workflow cannot
    # request more than its caller grants (startup failure otherwise).
    permissions:
      contents: read
      security-events: write
      actions: read
    with:
      light: true
      matrix: ${{ needs.changes.outputs.codeql }}

  ci-ok:
    # The one required context (besides dco) — fails unless every PR stage
    # succeeded, so matrix renames can never silently deadlock merges, and
    # every lane the selector chose ran green: `skipped` is OK only for a lane
    # nobody selected (a selected lane that skipped is a silent gate loss).
    # memwaste and the test-impact shadow are report-only and not needed here.
    needs: [changes, security, lint, test, pr-smoke, contracts, codeql]
    if: ${{ always() }}
    runs-on: ubuntu-latest
    timeout-minutes: 5
    # actions:read lists this run's jobs, the lane-level truth behind `needs`.
    permissions:
      contents: read
      actions: read
    steps:
      # Needs a checkout (unlike before): the gate logic lives in the canonical
      # scripts/ci entry, not inline YAML (venue-parity contract).
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
      - name: List this run's jobs
        env:
          GH_TOKEN: ${{ github.token }}
        run: |
          gh api --paginate "repos/$GITHUB_REPOSITORY/actions/runs/$GITHUB_RUN_ID/jobs?filter=latest&per_page=100" \
            --jq '.jobs[] | {name, status, conclusion}' > "$RUNNER_TEMP/jobs.jsonl"
      - name: All PR stages and every selected lane must have succeeded
        env:
          RESULTS: ${{ toJSON(needs) }}
          LANES: ${{ needs.changes.outputs.lanes }}
          JOBS: ${{ runner.temp }}/jobs.jsonl
        run: scripts/ci/require-all-green.sh