Repository navigation
chore(biome): enforce block statements - #679
Merged
Merged
Conversation
Turns on style/useBlockStatements as an error and marks its fix safe, so the pre-commit `biome check --write` adds the braces instead of only reporting them. Closes #666. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e
Mechanical sweep for the new useBlockStatements rule: `biome lint --write --only=style/useBlockStatements .` then `biome format --write .`. 703 sites in 134 files; with whitespace and braces stripped, every file is identical to before. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e
goosewobbler
pushed a commit
that referenced
this pull request
Oct 22, 2026
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e
Contributor
|
Standing release PR: #614 · v0.42.0 queued (4 packages) · open 82d 14m · ✅ ready to merge Release Preview — v0.41.2 → v0.41.3
These changes will be added to the release PR (#614) when merged: Changelogmonorepo 0.41.2 → 0.41.3
After merge — predicted release
Updated automatically by ReleaseKit |
6 of 19 tasks
Merged
3 of 19 tasks
goosewobbler
pushed a commit
that referenced
this pull request
Oct 22, 2026
Same sweep as #679, run on this branch so merging main resolves cleanly. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e
goosewobbler
pushed a commit
that referenced
this pull request
Oct 22, 2026
Same sweep as #679, run on this branch so merging main resolves cleanly. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e
goosewobbler
pushed a commit
that referenced
this pull request
Oct 22, 2026
Same sweep as #679, run on this branch so merging main resolves cleanly. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e
goosewobbler
pushed a commit
that referenced
this pull request
Oct 22, 2026
main now enforces style/useBlockStatements (#679). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e
goosewobbler
pushed a commit
that referenced
this pull request
Oct 22, 2026
main now enforces style/useBlockStatements (#679). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e
7 of 20 tasks
goosewobbler
added a commit
that referenced
this pull request
Oct 22, 2026
* docs: add commands and gotchas to AGENTS.md, and drop CLAUDE.md AGENTS.md keeps the architecture invariants and gains what an agent can't infer from the tree: the checks to run before pushing (there's no pre-push hook), why a one-package run can test stale dist/, that a package's own lint script skips ESLint, which suites only CI runs, and three gotchas: Turborepo's strict env mode drops opt-in variables such as RELEASEKIT_NOTES_E2E, Vitest hides output when it detects an agent, and Biome finds no files in a worktree under .claude/. It also lists the conventions the linters don't check. Claude Code reads AGENTS.md itself when there's no CLAUDE.md, and the claude-code-action pin installs a version that does, so the CLAUDE.md that only imported AGENTS.md goes. CONTRIBUTING.md asked for Node 18, while the packages need 22, and said there's no ESLint. It also claimed `pnpm --filter ... build` builds a package's dependencies, linked a BOOTSTRAP.md that no longer exists, and showed a package layout most packages don't match. The PR template and the commit scopes gain @releasekit/release, and the bug report's Node placeholder moves to a supported version. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * docs(agents): point the braces rule at #666 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * docs(agents): say which suites pnpm test skips Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * docs(agents): drop the braces rule now that Biome enforces it (#679) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * docs(contributing): require pnpm 12 and drop the .nvmrc aside Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * docs(agents): trim commands to what an agent can't infer; drop the any rule #682 makes noExplicitAny an error, so the rule no longer needs prose. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * docs(agents): drop the Vitest version from the reporter gotcha Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e --------- Co-authored-by: Claude <noreply@anthropic.com>
3 of 19 tasks
goosewobbler
added a commit
that referenced
this pull request
Oct 22, 2026
* chore(biome): flag console in package source Package code logs through the logger: core's, or version's own utils/logging.ts. Biome's noConsole now enforces that in packages/*/src, except the loggers themselves and the CLI command layer (command.ts, commands/), which owns the terminal: the JSON result on stdout and the final error message on stderr. The version package's error paths dumped stack traces and git output with console.error next to log(..., 'error'); they now go through log, which writes the error level to stderr in both modes, so the stream is unchanged. Three outputs that aren't diagnostics keep console behind a suppression: version's --json result, preview's dry-run comment, and the RELEASEKIT_DEBUG prompt dump (folded into one write, same bytes), which must print regardless of log level or --quiet. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * test(version): stop the strategy unit tests running real cargo versionStrategies.spec.ts mocks node:fs, so the Cargo.lock lookup in syncCargoLockfile finds a lock and the real sync shells out to `cargo update --workspace --offline` — 78 times per run. On a cold CI runner that pushes "should update all packages to the same version" past its 5s timeout; it failed that way on #668 and #671. Mock the sync like the other collaborators. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * test(version): drop the comment on the cargoLock mock Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * style: add braces to every single-statement body Same sweep as #679, run on this branch so merging main resolves cleanly. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * chore(biome): shorten the noConsole override comment and suppression reasons Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * chore(biome): reword the noConsole override and debug-dump suppression Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e --------- Co-authored-by: Claude <noreply@anthropic.com>
7 of 19 tasks
goosewobbler
added a commit
that referenced
this pull request
Oct 23, 2026
…618) * test(notes): make the LLM e2e infra-tolerant and close a false green First half of #589 — the spec change that has to land before the check can block merges. The workflow job and the OLLAMA_API_KEY secret are still outstanding. `isProviderUnreachable` splits the two things a real-provider check can fail on. Only an LLMError the provider positively classified as transient (timeout, 429, 5xx, socket) counts as infra; a non-retryable 4xx is a malformed request, and so is anything that isn't an LLMError — assertion failures and TypeErrors included. An unclassified LLMError fails too: every provider sets the flag explicitly, so an absent one means something unrecognised happened, and a visible red beats a silent skip. The predicate is a separate module with its own always-running spec, since it decides whether the gated check can go red at all. Found while verifying the skip path: the chunk-boundary test passed against an unreachable server in 20ms. enhanceAndCategorize never throws on a provider failure — its per-chunk fallback preserves the input entries — so asserting entry counts holds just as well with no provider at all. The test could not have detected the thing #589 wants it to guard. Two changes close that. A beforeAll probe establishes reachability once, so an absent host skips rather than being inferred from a later assertion; and the chunk test now requires that some descriptions actually differ from their inputs, which the fallback never produces. Verified against a dead port: both tests skip, where one previously reported a false pass. The positive path is unverified locally (no Ollama here) — it needs a run against a reachable model before the check is made blocking. Part of #589. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FADRmu9uhtCk3StMoqwV27 * test(notes): skip on a mid-run outage instead of reporting a regression From Greptile review of #618, and correct: the up-front probe cannot see an outage that starts after it. enhanceAndCategorize swallows retryable provider errors and falls back to the input entries verbatim, so a host that dies mid-test produces exactly the same observable as a model that declined to rewrite — and the new rewrite assertion would have failed the check as a product regression. "Nothing was rewritten" now re-probes before being called a regression. The provider is the only thing that can disambiguate, since the fallback is silent by design. Verified both directions against mock Ollama servers, because a fix here can fail either way — too eager and every regression becomes a silent skip: - Answers the probe, then 503s: skips with "SKIPPED mid-run: provider became unreachable", where it previously failed red. - Stays up but returns unparseable output: still fails red ("expected 0 to be greater than 0"), which is the regression the check exists to catch. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FADRmu9uhtCk3StMoqwV27 * test(notes): count rewrites per chunk so a partial outage can't pass From Greptile's re-review of #618, and right: chunks fall back independently, so a host that dies after the first chunk left ~30 rewrites in place, kept the total above zero, and sailed past the re-probe into a green check — a quieter version of the false green this PR set out to remove. Counting per chunk closes it. Either chunk coming back untouched now triggers the re-probe, which is the correct reading for a test whose subject is the boundary: the point is that both sides of it were enhanced, not that 30 of 35 entries were. Chunk size and entry count now derive from LLM_DEFAULTS rather than a hardcoded 35, so the case keeps straddling the boundary if that constant moves. Verified with a mock that answers the probe, serves the first chunk correctly, then 503s: skips, where before the fix it passed. The healthy-but-broken mock still fails red. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FADRmu9uhtCk3StMoqwV27 * test(notes): require transport evidence to skip, and retry like production Addresses the review findings on #618. isProviderUnreachable no longer accepts retryable:true on its own. Providers wrap any unexpected error as retryable, so a TypeError in provider code, a 200 HTML page, a null JSON body, a stream:false regression and a scheme-less base URL all skipped green. An LLMError with no cause is still accepted, because the provider classified it from an HTTP status or its own timeout. When there is a cause, the chain must contain a retryable status, a socket/undici code, or an abort/timeout. New spec cases cover a TypeError cause, a SyntaxError cause, and undici's fetch failed -> ECONNREFUSED. The e2e now wraps the provider in withRetry with isRetryableLLMError, as the pipeline does, so one 503/429/reset/timeout is absorbed and not reported as a regression. Errors that outlive the retries are recorded. An untouched chunk skips only when every recorded error is a proven outage. Otherwise it fails, and the message names the chunk and the errors with their cause chains, or says no provider error was recorded. The up-front probe and the re-probe are removed: the recorded errors answer the same question, and the maxTokens: 1 probe could fail a thinking model red. A 45s per-call timeout and a per-test breaker bound a mid-run hang to one retry budget (about 140s), which fits under the 240s test timeout. Under CI, OLLAMA_BASE_URL must be set. Every unreachable skip emits a ::warning:: annotation. Stale comments are corrected, and the no-op callProvider wrapper around enhanceAndCategorize is removed. Verified against a fake Ollama. Healthy and one-503 runs pass. HTML, null JSON, an injected provider TypeError, a removed stream:false and scheme-less URLs fail red with the cause chain. A dead port skips with ::warning::. A hang after the first chunk skips after about 138s per test. CI=1 with no base URL fails. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * test(notes): make the skip annotation parse under CI and trim unreachable predicate branches Third-review follow-ups: - Under CI vitest colours the `stdout | <test>` header, leaving escape codes in front of `::warning`, so the runner never parsed it as an annotation — every skip passed silently. Emit it on its own line, and pass the reason to ctx.skip(). - isProviderUnreachable: drop the status-in-cause and abort/timeout-name branches and the undici headers/body timeout codes — the Ollama provider classifies statuses and its own timeout itself, so they were unreachable; add EHOSTUNREACH/EHOSTDOWN, which production retries. Document the inherent blind spot (a provider-classified 5xx or timeout is trusted). - Assert no provider error was recorded when both chunks were rewritten; de-duplicate recorded errors by description; correct the breaker comment; size the test timeout for a chunk that needed corrective calls before a hang. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * test(notes): drop a pass-path assertion that can't fail; note the per-test skip decision Fourth-review nits: any recorded provider error already ends its chunk in the verbatim fallback, so asserting none on the pass path guarded nothing. Document that the skip-vs-red call is made per test rather than per chunk. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * test(notes): drop low-value e2e guards Remove the depth bound on the cause-chain walk (error causes aren't cyclic in practice) and the de-duplication of recorded errors by description, which only trimmed repeated lines from a failure message. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * test(notes): drop the e2e breaker and host-unreachable codes Without the breaker a full hang skips both tests inside the 330s timeout (measured 276s and 138s). EHOSTUNREACH/EHOSTDOWN aren't in production's NETWORK_ERROR_CODES; if one occurs the test goes red, the safe direction. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * style: add braces to the new single-statement bodies main now enforces style/useBlockStatements (#679). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * test(notes): tighten LLM e2e comments Keep the non-obvious why; drop restated mechanics and examples the code already shows. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * test(notes): cut LLM e2e comments to the non-obvious Drop comments that restate test names, test data, assertions or self-explanatory helpers; shorten the rest. The unclassified-error test now uses Ollama's real message instead of a comment pointing at it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * test(notes): call the LLM e2e what it is: a live Ollama test It calls the notes tasks directly with an OllamaProvider, which makes it an integration test against a live service, not an e2e (test/e2e drives the packed CLI). Rename it to ollama-live.spec.ts and RELEASEKIT_NOTES_E2E[_MODEL] to RELEASEKIT_NOTES_LIVE[_MODEL]. Its outage predicate assumes the Ollama provider's error shapes (SDK providers attach HTTP errors as a cause), and this spec is its only user, so it moves into the spec as isOllamaUnreachable. Its tests stay outside the opt-in suite so they run in every gate. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
goosewobbler
added a commit
that referenced
this pull request
Oct 23, 2026
…crate (#674) * fix(version): sync Cargo.lock on every Cargo.toml bump path The #496 lock sync only ran from updateCargoFiles, so crates bumped through version.groups (grouped and ungrouped) or the per-package processor left their own entry in the committed Cargo.lock at the old version. The sync now runs inside updateCargoVersion, the one function every version-step Cargo.toml write goes through. The sync also stops shelling out to `cargo update --workspace --offline`, which exits 101 on a runner with an empty CARGO_HOME (no cached crates.io index) and needs a Rust toolchain the release job may not have. Core's new syncCargoLockVersion edits the lock text the way cargo does for a path package's version change: the entry's version line plus any "name old-version" dependency references. Registry and git entries are never touched, and a lock where the edit would change how cargo tells same-named packages apart is left alone with a warning. Publish's prepare-stage sync uses the same helper. Closes #650 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * fix(version): sync every committed Cargo.lock that records a bumped crate #650 now asks that no committed Cargo.lock needs a follow-up sync PR after a release. - Locks that use a crate through a path dependency are synced too: for each bumped crate, every Cargo.lock tracked by git with a sourceless entry of that name is updated, leaving registry and git entries of the name alone. The crate's own lock is still synced even when untracked. The tracked locks are listed once per run through the git seam (new Git.lsFiles). - updateCargoVersion records each bump and VersionEngine.run syncs the locks once the strategy is done, so every strategy is covered, a lock shared by several crates is rewritten once, and every edit is worked out before the first write. - A lock that can't be synced fails the run with CargoLockSyncError (CARGO_LOCK_SYNC_ERROR, VERSION_ERROR exit family) instead of warning. A dry run works out the same edits, so it fails too, and records them as pending writes that release's flush writes with the manifests. - Publish's prepare stage syncs and stages every such lock, so the direct-commit flow commits fixture locks as well. It stays non-fatal: in the standing-PR flow it runs after the release PR has merged. - The real-cargo integration test runs `releasekit version` with no cargo on PATH and an empty CARGO_HOME for version.groups, the per-package processor and updateCargoFiles, and checks that `cargo metadata --locked` accepts every such lock unchanged. The syncCargoLockfile export of @releasekit/version (added for the integration test in #498) is removed: the sync is now per run. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * fix(core): keep cargo's order when a lock bump passes another version Cargo orders same-name [[package]] entries by semver and a package's dependency references by the version as a plain string, so "dupe 10.0.0" sorts before "dupe 2.0.0" (checked with cargo 1.97). Rewriting a path crate's version in place left both in their old order whenever the new version passed a same-named registry or git entry, and the next cargo run would reorder the lock. setCargoLockVersion now moves the edited entry, and the run of references around each edited reference, back to cargo's order. The real-cargo byte-for-byte test bumps crate-b from 0.1.0 to 0.2.0 past git copies at 0.1.5 (entry order) and 0.10.0 (reference order). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * style: add braces to every single-statement body Same sweep as #679, run on this branch so merging main resolves cleanly. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * fix(version): sync a lock entry only for the crate it records, never an ignored lock Crates in different workspaces can share a name (every `tauri init` app is `app`), and every bump was applied to every lock with a path entry of that name, so one app's lock took another app's version. A crate's own lock still takes the crate's bump whatever version it records. In any other lock, the entry follows a bump only when something in that lock depends on it (an entry nothing depends on is that lock's own root or member, a different crate) and it is at the crate's old or new version; bumps record the manifest's pre-bump version for this. Same-named bumps to different versions that fit one entry fail the run. A crate's own lock that git ignores is no longer synced or staged: the direct-commit release's `git add` refused it and failed the release. `Git` gains `isIgnored` (`check-ignore -q`); the planner now does its own discovery, so `findCommittedCargoLockfiles` is gone. Also: version's syncCargoLockfiles returns void, the unused `git` parameters and FakeGit's trackedFiles seed are dropped, the duplicate comment in VersionEngine.run is removed, and tests pin numeric pre-release ordering and the several-path-entries refusal. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * docs: document CARGO_LOCK_SYNC_ERROR and how lock entries are matched docs/cli.md lists the code (category `cargo-lock-sync`, exit code 8). docs/rust.md states which lock entries follow a bump and covers every reason the sync refuses a lock, not only the other-source one. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * docs: keep the exit-code table to families; shorten the lock-sync bullet Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * fix(version): treat a lock beside a crate's manifest as its own findCargoLockfile doesn't read an ancestor workspace's `exclude`, so a crate excluded from it (a Tauri app's src-tauri, say) got the ancestor's lock as its own, and its real lock was matched as another workspace's: its root entry, which nothing depends on, was skipped and left stale. A lock beside the manifest can only be the crate's own (or an unused stray), so it now counts as one. When the manifest records no version (`version.workspace = true`, or none), the old version now comes from the crate's own lock, so other locks are still matched on it; only without an own lock does any version match. Tests pin both, the any-version fallback, publish reporting the replaced version, and the versioned-reference match in core. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * docs(rust): narrow the same-name claim, note the residual case and exit codes Only other apps' own crates of the same name are told apart; a different same-named crate that a lock depends on at the bumped crate's old or new version is not, so the unsupported list says so and how to avoid it. The lock-sync error exits 8 from `release` and `standing-pr` but 1 from `releasekit version`. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * docs(rust): shorten the same-name limitation Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * fix(core): find a crate's Cargo.lock by cargo's workspace-root rule findCargoLockfile took the first ancestor `[workspace]` as the crate's root without reading its `exclude`, so a crate excluded from the repo's workspace treated that workspace's lock as its own, where every bump is applied: two excluded Tauri apps named `app` failed the run as ambiguous in a root lock that has no `app`, and bumping one excluded `app` set the root workspace's own member `app` to its version. It now follows cargo: a crate whose manifest has a `[workspace]` is its own root; otherwise the nearest ancestor workspace that doesn't exclude it is (an `exclude` path prefix not overridden by a `members` one, both literal); with none, the crate is its own root. Only the root's lock is returned, which also drops the nearest-lock fallback for standalone crates (cargo never reads a non-root ancestor's lock). Checked against cargo 1.97. `exclude` is read with smol-toml, now a core dependency. The round-2 "lock beside the manifest" own-lock candidate is dropped: the new rule returns it whenever it is the crate's lock, and a member's stray lock is unused by cargo. The planner's doc now notes that any version matches when neither the manifest nor the crate's own lock records the old one. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * docs(rust): note the any-version case of the same-name limitation Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * test(core): pin whole-component exclude matching; note the unfollowed workspace pointer Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * refactor: cut Cargo.lock sync comments to the non-obvious Drop doc comments that restate a name or signature, issue numbers and history; shorten the rest to the cargo behaviour the code relies on. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(core): sync another workspace's Cargo.lock only if it reaches the crate by path Name and version can't identify a path crate: two workspaces can each have a `core` 0.1.0, and bumping one rewrote the other's lock, breaking `--locked` builds. A lock other than the crate's own now takes a bump only if its workspace reaches the crate's directory through path dependencies (target-specific, `workspace = true`, `[patch]`; dev-dependencies of its own members only), as cargo resolves them. Drops the version heuristic and `fromVersion`. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(core): follow package.workspace pointers, and only the root's [patch] Cargo honours [patch] only in the workspace root's manifest, so following it from every visited manifest could rewrite an unrelated same-named entry. And a member outside its workspace's directory points back with package.workspace; without following it, its real lock went stale. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(core): refuse to edit a Cargo.lock format newer than v4 Also document that `[patch]` in `.cargo/config.toml` isn't followed. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(core): stop following [patch] when syncing Cargo.lock Cargo drops a patch the bumped version no longer satisfies, so following it could write a lock cargo never would. Support is tracked in #706. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude <noreply@anthropic.com> Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
goosewobbler
added a commit
that referenced
this pull request
Oct 23, 2026
…pipe (#616) * feat(cli): envelope the version and publish commands, unwrap on the pipe Completes the envelope scope of #544. `version --json` and `publish --json` were the two commands still printing a bare payload; both now emit the uniform envelope and accept `--output <path>` for parity with the orchestration commands. `notes` has no JSON output, so the surface is now uniform. The coupling this was deferred for is the pipe: `version | notes | publish` carries JSON between processes, so enveloping the producer means teaching the consumers to unwrap on input. Both do, via `unwrapEnvelope` in core — called once inside `parseVersionOutput` and `parseInput` so the string, file, and stdin entries are all covered. A bare `VersionOutput` still passes through, so a hand-assembled file or an older releasekit keeps working. Piping a failed stage forward now fails with that stage's message rather than a schema complaint about missing fields, which pointed at the wrong thing. Two things fell out of doing this honestly: `errorEnvelope` takes optional `data` and `changed`. Enveloping publish's failure path would otherwise have dropped `partialOutput` — what already landed before the failure — leaving the enveloped output less informative than the prose it replaced, and reporting `changed: false` for a partial publish that did change state. Publishes are idempotent and retried, so that is exactly what the retry reads. `failedStage` has no envelope field, so it is named in the error message rather than dropped. Readers that ignore `data` when `status` is `error` are unaffected, so schemaVersion is unchanged. The publish-changed predicate moved to `@releasekit/publish` as `publishDidChange`. The publish CLI and `standing-pr publish` must answer `changed` the same way, and the logic is subtle enough (tags, `git.pushed`, and already-existing GitHub releases all report success without anything having published) that two copies would drift. It is also now tolerant of absent registry arrays: it runs on the partial output of a failed pipeline, where throwing would replace the publish error the caller needs. Note `version` now exits with the mapped code from the documented table rather than always 1 — in practice only `GIT_ERROR`, which becomes 7. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FADRmu9uhtCk3StMoqwV27 * fix(cli): complete the exit-code table and validate envelopes on unwrap Both from Greptile review of #616; both were real. **Exit classification.** The code → exit-code map only held the nine family names, but packages throw specific codes — NOT_GIT_REPO, TAG_ALREADY_EXISTS, PACKAGES_NOT_FOUND, NPM_AUTH_ERROR and the rest — so almost every real failure fell through to exit 1 while docs/cli.md advertised a mapping. Automation could not tell a git failure from a config one. Every code the version, notes, and publish packages throw is now mapped to its family, with a test that fails when a new code isn't, since the degradation is silent. The family assignment is a judgement call in two places worth a second opinion: workspace/package-discovery failures (PACKAGES_NOT_FOUND, WORKSPACE_ERROR) are treated as config errors on the reasoning that the `packages` globs matched nothing, and manifest-parse failures (CARGO_TOML_ERROR, PUBSPEC_YAML_ERROR) likewise. **Envelope validation.** `unwrapEnvelope` accepted anything with the three keys, so an envelope from a newer producer, an unrecognised status, or a success envelope with a null payload all flowed through — the last surfacing in `notes` as a null-property error about `changelogs` rather than an input error. It now rejects each with a message naming the actual cause. This is what schemaVersion was for; nothing had been reading it. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FADRmu9uhtCk3StMoqwV27 * fix(cli): report publish commits as changes and exit publish by code family Review follow-ups for the envelope work: - `publish --json` reported `changed: false` after creating the release commit and tags whenever no registry published (GitHub-release-only, `--skip-publish`). `publishDidChange` now also counts `git.committed`, which is only set for a commit this run created; `standing-pr publish` never commits (`skipGitCommit`), so its `changed` is unaffected. - `publish` exited 9/1 regardless of the error code, contradicting the code-to-exit table. It now exits via `exitCodeForError`, like `version`: config failures exit 2, input failures 3, pipeline failures still 9. - `version --output` with a relative path landed inside `--project-dir` (the command chdirs first). Resolve it against the invocation directory up front, as `release --output` does. - docs: the `version | notes | publish` pipe never worked (`notes` does not pass the VersionOutput on); document chaining through a file. Correct the `changed`/retry wording and the exit-code example. Update the version and publish READMEs for the envelope and `--output`. - release: reuse core's `writeEnvelope` instead of a private copy. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * fix(cli): correct the chaining, retry and exit-code docs; move the --output test into CI Second-review follow-ups: - docs: the file-based `version` -> `notes` -> `publish` chain left the changelog out of the release commit (publish commits only the version files plus what is staged). Document staging it; verified end to end. - docs: a re-run of the standalone `publish` in the same tree stops at the git-commit stage, so drop the blanket "a retry is safe". - docs: the exit-code examples named codes `version` never throws (TAG_ALREADY_EXISTS / GIT_PROCESS_ERROR are unreachable). Use codes `publish` actually emits: input -> 3, config -> 2, pipeline -> 9. - test: the `--output` integration test lived in version's integration suite, which CI never runs. Replace it with a unit test in test/unit/cli.spec.ts (fails against the pre-fix command). - publishDidChange docstring: `committed` is also set in a dry run; the dry-run guard is what keeps that out of `changed`. - version README: the flag table listed --strict-reachable, --verbose and --quiet, which the command doesn't have, and omitted several it does. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * docs(cli): scope the exit-code contract to the commands it holds for Third-review follow-ups: - The specific-code / family-exit contract holds for version, notes and publish, but `release` and `standing-pr` report errors raised inside their bundled version/notes/publish stages as GENERAL_ERROR (exit 1): the release bundle carries several copies of core's ReleaseKitError, so the instanceof check misses errors from the other copies. Say so rather than claim otherwise; the cause predates this PR and is fixed separately. - `data` on error: describe the actual exception (a publish pipeline failure carries partial output even when nothing landed). - core README: VersionOutput is now the envelope's `data`. - Restore unit tests for releaseDidPublish, the standing-pr `changed` wiring, dropped when the predicate moved to @releasekit/publish. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * fix(cli): don't crash the error path on an unwritable --output; scope the notes exit claim Fourth-review follow-ups: - docs/cli.md claimed the family exit-code contract for `notes`, but notes exits 1 for config errors and most non-parse failures; only its input-parse failures exit 3. Say exactly that. - An unwritable `--output` path made the result write throw, then the error path threw again writing its envelope to the same file, crashing with a stack and no envelope. writeEnvelope still throws for a result write (the command must fail), but notes a failed error-envelope write on stderr instead of throwing out of the error handler. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * docs(cli): describe notes' parse check precisely; name a lost --output result Fifth-review nits: - notes only checks that its input has a `changelogs` array, so "a schema-invalid object" overstated the exit-3 case; say what it checks. - When the result write to --output fails, throw "Could not write the result to <path>: ..." instead of the bare ENOENT, so stderr says the command's work happened but its result never reached the file. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * test: drop low-value review additions Remove the releaseDidPublish unit tests (a one-line wrapper over publishDidChange, which publish's own tests cover) and rethrow the original error when a result can't be written to --output instead of re-wrapping it — the ENOENT already names the path. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * test: drop the duplicate publish exit test and the notes exit clause INPUT→3 runs the same exit-code path as CONFIG→2, which stays as the guard against hard-coded exits. The cli.md notes clause is trimmed to the commands the table holds for. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * docs(cli): release and standing-pr now report a stage's own error code #651 is merged, so an error raised inside their version, notes or publish stage keeps its code instead of degrading to GENERAL_ERROR. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * refactor(cli): share one error-envelope writer and trim review leftovers - core: add writeErrorEnvelope next to writeEnvelope; version, publish and release's emitError now call it instead of each building the same single-error envelope. - core: isEnvelope is no longer exported — only unwrapEnvelope needs it, and the version test helper now reads output through unwrapEnvelope, the same boundary notes and publish cross. - publishDidChange: drop the `?.` / `?? []` guards. Every PublishOutput it sees comes from runPipeline's ctx.output (the only PipelineError is built from it), whose registry arrays are always initialised. - Drop comments that repeat the unwrapEnvelope JSDoc or another comment, the old-behaviour note on publish's exit, and the "pipe's contract is unchanged" claim (older consumers reject the envelope). - Braces on the one-line ifs and loop this PR added; give publishDidChange its own export section so runPipeline keeps its header. - Drop the "older producer" unwrap test: it used the current schemaVersion, so it duplicated the success-envelope case. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * test(cli): cover the version and publish failure paths through the commands - version --json: an engine error reaches stdout as an error envelope with its own code and exits by its family (INVALID_CONFIG -> 2); --stable with --prerelease prints an INPUT_ERROR envelope and exits 3. - publish --output: a PipelineError writes its envelope, partial output in `data`, to the file and nothing to stdout. - publish --input holding a failed upstream envelope runs the real parseInput and exits 3 with INPUT_PARSE_ERROR. It replaces the "registry failure" test, which rejected runPipeline with a bare NPM_PUBLISH_ERROR that runPipeline never throws (it wraps every stage error in a PipelineError). - Exit-code coverage moves from a hardcoded list in core to each package's own error enums (PublishErrorCode; VersionErrorCode, GitErrorCode and StrictReachableError), so a new unmapped code now fails a test. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * docs: point --json stability at schemaVersion; keep the version README to this PR's flags - troubleshooting: "the JSON schema is stable across minor versions" no longer holds — this release changes the version/publish --json shape. Say what is versioned instead: the envelope's schemaVersion. - version README: revert the flag-table rewrite that rode along (short aliases, --stable, --allow-first-bump, --sync, --config, --include-prerequisites, and dropping --strict-reachable / --verbose / --quiet). Keep only the --json description and the new --output row. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * style: add braces to every single-statement body Same sweep as #679, run on this branch so merging main resolves cleanly. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * fix(cli): suppress per line instead of exempting cli.ts; test --output alone The --json print takes a one-line noConsole suppression and the unwritable error-envelope note goes through the logger, so core/src/cli.ts no longer needs a file-wide exemption. A version --output test now fails if --output alone stops collecting the result. Inline releaseDidPublish, drop a guard writeEnvelope already applies, and trim two comments. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * test(version): write --output through the real JSON writer Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * docs(core): trim the envelope doc comments; drop a stale e2e helper note Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * docs: tighten the envelope docs and comments Fix the failed-upstream wording in docs/cli.md (it's an input error that names the upstream code), say plainly that a partial publish can't be resumed yet, and cut or shorten the comments that restated code or docs. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * docs: link to the chaining section from architecture, simplify envelope comments Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
goosewobbler
added a commit
that referenced
this pull request
Oct 23, 2026
* test(notes): add the structured enhance-and-categorize eval case Completes the assertion scope of #542. The harness shipped with the free-text release-notes path; the structured path is the case it was built to take next, and the one that unlocks the category-distribution assertion the issue lists. Two assertions the free-text case can't reach: `checkCategoryDistribution` — grouping has to discriminate, not merely be legal. The task validator already rejects category names outside the configured set, so a model that drops every entry into one bucket passes validation and still produces a useless grouping. Catches too few populated categories, an emitted-but-empty category, and one category swallowing most of the entries. The share bound is skipped below four entries, where a lopsided split isn't evidence of anything. `checkNoEntryLoss` — every entry survives categorization exactly once, catching both drops and duplication across buckets. `loadGoldenCase` is now generic in its context type, since the tasks take different ones — release notes a ReleaseNotesContext, this an EnhanceContext & CategorizeContext. Also adds unit tests for the assertion module. These are the regression net, so each needs a case that actually trips it; without one the eval would pass every run while catching nothing. Verified the drift signal fires on the new case: editing the enhance-and-categorize system prompt changes the cache key and the run fails with the re-record hint rather than silently replaying stale output. Recording the new case left the existing release-notes fixture untouched, confirming the per-case refresh. Part of #542. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FADRmu9uhtCk3StMoqwV27 * test(notes): give the eval categories their required descriptions From Greptile review of #617, and a real defect in the fixture. `LLMCategory.description` is required, and `buildCategorySection` renders each category as `- "<name>": <description>`. The fixture omitted them, so the prompt the eval measured against contained `- "New": undefined` — a shape production never produces. The case was pinning malformed categorization guidance. Re-recorded from a clean cache, since the corrected prompt changes the key. That also drops the entry orphaned under the old key — the clean-slate procedure the harness header documents, now exercised for the first time. The release-notes fixture is untouched, confirming per-case recording. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FADRmu9uhtCk3StMoqwV27 * test(notes): check entry identity in the eval and use a production-shaped golden input Review follow-ups for the structured eval case: - `checkNoEntryLoss` compared totals only, so a dropped entry and a duplicated one cancelled out (a groupByCategory mutation that did exactly that passed the eval). It now checks identity: every enhanced entry sits in exactly one bucket once, no bucket holds anything else, and enhancement kept the golden entry count. The same mutation now fails with "entry 0 was categorized 2 times / entry 1 was dropped". - The golden input kept raw `feat(api): ...` subjects with no `scope`, a shape neither commit parser produces, so the pinned prompt wasn't the one production builds. Use the parsed shape (bare subject, `scope`, `originalType`, no empty `issueIds`) and re-key the cache from the hand-authored recorded response. - A cache miss on a corrective retry now says the recorded response was rejected, instead of advising a re-record that would fail identically. - Note that the case exercises the text -> JSON path, not the structured-output branch. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * test(notes): tighten the eval harness and fixture after second review - Record mode answered corrective retries with the canned markdown, caching orphan entries that replay then served, so a rejected recording failed later on a misleading distribution assertion. Both offline modes now refuse a corrective retry at the source. - checkNoEntryLoss reports grouping that copies entries as a broken identity assumption instead of N phantom "dropped" entries. - Restrict the golden case's scopes to api/config, so grouping has to use the scope-validated entries: grouping the pre-validation entries now fails the eval (it passed before). Prompt and cache key unchanged. - Retarget the prefix-leak guard to a leading `word(scope):` / `word:` prefix; the old pattern only matched tokens the prompt never contains. - checkCategoryDistribution: drop the possibly-undefined reduce seed (TS18048 under tsconfig.test.json). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * test(notes): only call bucketed entries copies when their content matches Third-review follow-ups: - checkNoEntryLoss's "copies" diagnosis matched on counts alone, so bucketing the raw, un-enhanced input (a real regression) was reported as a harmless copy and the developer told to relax the check. It now requires each stray to be bucketed once and to match a distinct missing entry by content; anything else reads as loss. Partial copies are named as copies; unrelated objects and duplicated copies are not. - Assert directly that no disallowed scope reaches a bucket, so grouping the unvalidated response fails on its own, not only via identity. - Build the prefix-leak pattern from the golden case's own type, originalType and scope labels, so a legitimate lead-in ("CLI: ...") in a live run isn't flagged. - checkCategoryDistribution's doc listed two failure modes; it checks three. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * test(notes): close the last copy-detection and scope-check gaps Fourth-review nits: - Cover the length guard in isOneToOneCopy: one copied entry plus one genuinely dropped entry must read as loss, not as a clean copy that hides the drop. - Compare copies by key-sorted JSON, so a copy rebuilt with its keys in another order is still recognised. - Match scopes the way production validates them (case-insensitive unless the config says otherwise), so a live model's "API" for an allowed "api" is not a false failure. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * test(notes): drop low-value eval additions Remove the copy diagnosis in checkNoEntryLoss (isOneToOneCopy, canonicalJson and their spec cases) — the check fails either way, it only changed the message for a refactor that doesn't exist — the regex escaping of the fixture's fixed labels, and the case-insensitive scope match, which only mattered for live-model runs. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * style: add braces to the new single-statement bodies main now enforces style/useBlockStatements (#679). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HQN3v9Z4f1mu3xK4Z7K44e * test(notes): tighten eval comments Keep the non-obvious why; drop restated mechanics and examples the code already shows. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * test(notes): trim eval comments and escape the prefix-leak labels Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Implements #666: every
if/else/loop body is a block, enforced by Biome.Two commits, kept apart so the mechanical one is easy to skip in review:
chore(biome): enforce block statements—biome.jsoncturnsstyle/useBlockStatementsfrom"off"to:Marking the fix safe means the pre-commit
biome check --write(lint-staged) adds the braces itself instead of only reporting them, so contributors and agents don't have to.style: add braces to every single-statement body— the sweep:biome lint --write --only=style/useBlockStatements .thenbiome format --write .. 703 sites in 134 files (src, tests andscripts/). Nothing else changes: with whitespace and{/}stripped, every one of the 134 files is identical tomain.eslint.config.jshas nocurlyrule, so nothing conflicts with it.Merging this before other open PRs
The sweep touches most source files, so open branches will conflict after it lands. A branch resolves cleanly by running the same fix before merging
main, so both sides already carry identical braces and only real changes conflict:I'll do that on the PRs I'm driving (#616, #668, #673, #674) once this merges.
Follow-up:
.git-blame-ignore-revsThe sweep should be listed in
.git-blame-ignore-revssogit blameskips it, but the entry has to be the commit that lands onmain— with a squash merge that SHA only exists after merging. I'll add the file with the merged SHA as a one-line follow-up.Related Issue(s)
Closes #666
Type of Change
feat— New featurefix— Bug fixdocs— Documentation changerefactor— Code refactoringtest— Adding or updating testschore— Maintenance, dependencies, toolingScope
@releasekit/version@releasekit/notes@releasekit/publish@releasekit/config@releasekit/coreChecklist
pnpm lint)pnpm test)pnpm typecheck)Testing Notes
biome lint --only=style/useBlockStatements .reports nothing; before it, 703 errors.mainversion and this branch's version are identical once whitespace and braces are removed.pnpm build,pnpm lint(29 warnings, same asmain),pnpm typecheckpass.pnpm test:coverage(what CI runs): 15/15 tasks pass with thresholds enforced.pnpm --filter @releasekit/integration-tests test: 63 passed.Generated by Claude Code
The PR appears safe to merge.Summary
Enables
useBlockStatementsas an error inbiome.jsoncand marks its fix safe. Adds braces across source files, tests, and scripts.Reviews (1) · Last reviewed commit: "style: add braces to every single-statem..." · Reviewed by Greptile