diff --git a/.github/workflows/pre-commit.yml b/.github/workflows/pre-commit.yml index 871478951..10424c063 100644 --- a/.github/workflows/pre-commit.yml +++ b/.github/workflows/pre-commit.yml @@ -122,12 +122,11 @@ jobs: # `package = false` so member packages and their # `[project.scripts]` entry points would be skipped. run: uv sync --all-packages --group dev - # Cache prek's hook environments. The `lychee` hook is - # `language: rust` with `additional_dependencies: [cli:lychee:…]`, - # so prek `cargo install`s lychee into a hook env under - # `~/.cache/prek` — a multi-minute compile. Caching that dir - # reuses the built binary (and every other hook env) across runs; - # keyed on the prek config so a hook/version change busts it. + # Cache prek's hook environments, so each run reuses the built envs + # instead of installing every hook's dependencies again; keyed on the + # prek config so a hook/version change busts it. (lychee, the one + # `language: rust` hook, is not among them in CI: it runs from a + # prebuilt binary, see "Install lychee" below.) # # Kept small for `ubuntu-slim` (#2175), whose single vCPU has to # download and unpack the archive on every PR run. The archive had @@ -147,16 +146,21 @@ jobs: env: SEGMENT_DOWNLOAD_TIMEOUT_MINS: "2" with: + # `prek/*`, not `prek`: the action tars each matched path whole, so an + # exclusion only drops a path the glob matched on its own. With the + # bare directory, `!…/cache` would match nothing and stay in. path: | - ~/.cache/prek + ~/.cache/prek/* !~/.cache/prek/cache - # `v2`: a fresh prefix, so the first run builds a lean archive instead - # of restoring the old ~3.5 GB one (download caches and stale envs - # included) through `restore-keys`. Bump it whenever the archive - # must start over rather than carry the previous one forward. - key: prek-v2-${{ runner.os }}-${{ hashFiles('.pre-commit-config.yaml') }} + # `v3` (`v2` was the cache-size fix): CI stopped building lychee's + # Rust env (see "Install lychee" below). `prek cache gc` keeps envs + # of every hook the config names, so under the old prefix + # `restore-keys` would carry that env forward for good; a new prefix + # starts the archive without it. Bump it whenever the archive must + # start over rather than carry the previous one forward. + key: prek-v3-${{ runner.os }}-${{ hashFiles('.pre-commit-config.yaml') }} restore-keys: | - prek-v2-${{ runner.os }}- + prek-v3-${{ runner.os }}- # There is deliberately no lychee result cache here. `.lychee.toml` # sets `offline = true`, so the hook resolves only in-repo paths and # anchors and never fetches a URL — there are no results to cache. @@ -276,16 +280,39 @@ jobs: # # Cheap enough to do on both events: `.lychee.toml` sets # `offline = true`, so this resolves in-repo paths and `#anchor` - # fragments only and fetches no URL. The hook is `language: rust` - # (no Docker daemon, no service container), so it runs on - # `ubuntu-slim` as happily as on `ubuntu-latest`, and the prek - # hook-env cache above already holds the compiled binary. + # fragments only and fetches no URL. + # + # CI runs lychee's prebuilt release binary rather than the hook: + # the hook is `language: rust`, so prek would `cargo install` it, a + # multi-minute compile on `ubuntu-slim`'s one vCPU and the biggest + # thing in the hook-env cache (#2175). The release tarball is pinned + # by version and SHA-256 (the digest GitHub reports for the asset); + # the static musl build runs on both runner images. The arguments are + # the hook's own, from `.pre-commit-config.yaml`: keep the two in + # step. Contributors still run the hook locally through prek. + - name: Install lychee (pinned prebuilt release) + if: ${{ !cancelled() }} + env: + LYCHEE_VERSION: "0.24.2" + LYCHEE_SHA256: "73657a111819a30c47c08352896796f23d64e4eb2b3ed39b6d32149241566fc5" + run: | + dir="$RUNNER_TEMP/lychee" + mkdir -p "$dir/bin" + curl -sSfL --retry 3 -o "$dir/lychee.tar.gz" \ + "https://github.com/lycheeverse/lychee/releases/download/lychee-v${LYCHEE_VERSION}/lychee-x86_64-unknown-linux-musl.tar.gz" + echo "${LYCHEE_SHA256} $dir/lychee.tar.gz" | sha256sum --check --strict + tar -xzf "$dir/lychee.tar.gz" -C "$dir" + bin="$(find "$dir" -path "$dir/bin" -prune -o -type f -name lychee -print | head -n 1)" + test -n "$bin" + install -m 0755 "$bin" "$dir/bin/lychee" + echo "$dir/bin" >> "$GITHUB_PATH" + "$dir/bin/lychee" --version # # `!cancelled()` rather than `always()`: a link break and a hook # failure in the step above are independent, and a contributor # should see both in one run instead of fixing one to discover the # other. `always()` would also run this on a cancelled job. - - name: Run prek (lychee — whole repo, both events) + - name: Run lychee (whole repo, both events) if: ${{ !cancelled() }} # GITHUB_TOKEN lets lychee authenticate its github.com link # checks — unauthenticated requests get rate-limited (429) once a @@ -295,7 +322,7 @@ jobs: # kept so flipping that back on needs no workflow change. env: GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} - run: prek run lychee --color=always --all-files + run: lychee --config .lychee.toml --no-progress . # Claude Code mods: `claude plugin validate --strict` and `claude plugin # test` on every plugin whose `hooks/hooks.json` declares `modules`. The # hook is `stages: [manual]` so contributors' commits never fetch the diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml index 4b8090084..266bad3c0 100644 --- a/.pre-commit-config.yaml +++ b/.pre-commit-config.yaml @@ -225,14 +225,16 @@ repos: # whole repo (so a renamed link target is caught no matter which # file references it); `files:` only gates *whether* the whole-repo # scan fires, i.e. it runs when any doc file changes and always on - # `prek run --all-files`. That gate is why CI gives this hook a step - # of its own (`prek run lychee --all-files`, both events) instead of + # `prek run --all-files`. That gate is why CI gives lychee a step + # of its own (whole repo, both events) instead of # letting it ride along with the pull-request run's `--from-ref` # scope: a PR that breaks a link by renaming a non-doc file matches # no `.md`, so the hook would never fire. `.lychee.toml` sets # `offline = true`, so no external URL is fetched and the # `.lycheecache` result cache stays empty — CI carries no cache step - # for it. + # for it. CI runs a pinned prebuilt lychee with this hook's `args` + # rather than the hook itself (no cargo build on the runner, see + # `.github/workflows/pre-commit.yml`): change the args in both places. - repo: local hooks: - id: lychee