Repository navigation
fix(review-depth): gate-class review-exclude and default-base; harness pins class as code (GST-972) - #3959
Merged
Conversation
…s pins class as code (GST-972) lib/review-exclude.ts and lib/default-base.ts join GATE_PATH, so an edit to either plans gate class at full depth. scripts/harness-pins/** classifies as code before the test/ rule and joins DATA_CONFIG, so the unconsumed-data rule cannot move a pin to docs. Repins lib/review-depth.ts under replay GST-972. Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…te class adds Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
|
😎 This pull request was merged. |
…hs and harness pins as code (GST-972) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
…followup-review-depth Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> # Conflicts: # test/fixtures/review-pins/files/lib/review-depth.ts.json
…followup-review-depth Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> # Conflicts: # test/fixtures/review-pins/files/lib/review-depth.ts.json
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What changed
gstack-review-plan on a scratch commit: a harness helper plus its pin now plans the code-class readers, and a
review-exclude.tsedit now plans gate class at full depth. (Image render failed on this machine, PR_SHOTS_RENDER_FAILED, so the text version stands in.)Before/after as text: gstack-review-plan on a scratch commit: a harness helper plus its pin now plans the code-class readers, and a review-exclude.ts edit now plans gate class at full depth
Why
GST-972 item 4: an edit to
lib/review-exclude.ts(which files reviewers never see) orlib/default-base.ts(the base every diff is cut from) planned light review with no gate class. Since #3903 (a1da77e) split harness pins into one file per harness file, a harness helper edit with its pin planned as tests with no reviewers at all. Both edits now get the review they had, or should have had.Failing-first proof
Four new tests in
test/review-depth.test.tsfail at main 1ccab0d (4 fail, 0 pass) and pass on this branch (4 pass). Logs:handoffs/GSTA-202/evidence-GST-972/followup/failing-first-{main,branch}.txtin the GStack Minder drive.No-loss replay (GST-972)
GST-972 follow-up no-loss check): the GST-967 history grid over the four changed cases, 1,152 planReview cases. Stand-in paths reproduce main's plans at 1ccab0d byte for byte (sha798e56fa…). On the real paths the branch only raises class and depth (review-exclude and default-base code/light to gate/full; helper+pin tests/light to code/light; pin alone docs/light to code/light), and drops no reviewer. One effort change: in 56 cases (28 per module), light's batched reader carried the Red Team prompt at full effort; full depth runs the checklists at the specialist effort (medium) with the Red Team as its own full-effort reader. That's 168 checklist slots, the same split every gate-class diff already gets.collectCommit, scope {} and gate scan 0): 0 changed plans. The three commits touching these paths (a1da77e, 8d276f9, 14d7414) were already gate/full for other reasons.lib/review-depth.tsrepinned under replayGST-972;test/review-packets-removed.test.tspasses (9/9).Blast radius
Review planning only:
classifyPath,isGatePathandisDataConfigfeedclassifyDiff, the consumer scan inlib/review-collect.tsand the keep rules inlib/review-exclude.ts. Every change only raises a class or keeps a file in review, and the replay below shows no plan on the last 100 main commits changed (script ran: 28,800 plans compared).Tests
Targeted: 25 review/ship test files, 790 pass, 0 fail.
bun run typecheckclean;typecheck:testratchet: none new.Open notes
Approvals
Approved: yes (record r-6e3cbbcdec0c). The full approval record stays private: this repository is public or its visibility is unknown.
Captain grant, recorded verbatim (a-9d54beebf4ba): "review_loop_grant for #3959 at 30d1016: the post-Step-11 doc fix-check counts (GStack Minder, 2026-11-03)."
Captain grant, recorded verbatim (a-ec43de3234bf): "review_loop_grant by GStack Minder 2026-11-03: one scoped fix-check at 4bdd9d2 on docs/review-policy.md (main's auto-merge), the review-depth.ts.json pin and the GSTA-258-C+GST-972 replay entry. No further rounds, within the $31 cap."
Main merge: bf2cab3 merged once under GStack Minder's rule 19 grant. The only conflict was the
lib/review-depth.tspin, repinned at c7e7d71f under the combined replayGSTA-258-C+GST-972(main's bytes plus exactly the GST-972 lines). The scoped fix-check at 4bdd9d2 found nothing.Summary
Two review-depth bugs, both proven on main 1ccab0d before the fix.
lib/review-exclude.tsandlib/default-base.tsjoinGATE_PATH, so an edit plans gate class at full depth.scripts/harness-pins/**classifies as code before thetest/rule, and joinsDATA_CONFIG. TheclassifyPathchange alone wasn't enough. On a real scratch commit, the collector's unused-data rule still turned the pin into docs, because no code names a pin file by basename. Pins outsidetest/planned as docs for the same reason.docs/review-policy.mdanddocs/ship-policy.mdnow list both modules as gate paths and the pins as code and config.Failing-first proof. Four new tests in
test/review-depth.test.tsfail at 1ccab0d (4 fail, 0 pass) and pass on this branch.No-loss replay (GST-972), under GST-967's bar. The committed check runs the GST-967 history grid over the four changed cases, 1,152 planReview cases. Stand-in paths reproduce main's plans at 1ccab0d byte for byte. On the real paths the branch only raises class and depth and drops no reviewer. Over the last 100 first-parent main commits to 1ccab0d, 100 commits × 288 histories (28,800 plans), 0 plans changed. The three commits touching these paths (a1da77e, 8d276f9, 14d7414) were already gate/full.
The shift covers 56 cases and 168 checklist slots: light's batched reader carried the Red Team prompt at full effort, and full depth runs those checklists at medium with the Red Team as its own full-effort reader.
lib/review-depth.tsis repinned under replayGST-972.Test Coverage
All new code paths have test coverage: the four failing-first tests plus the no-loss check (
test/review-depth.test.ts).Tests ran: free 1566 (files, carried from Free Tests run 41701136631 at tree 6287ee3e3712)
Pre-Landing Review
No blocking issues. One informational finding was skipped as out of scope (below).
REVIEW_LOOP: passes=1 blocking_found=0 deferred=0 noted=1 fixed=0 scope=frozen ledgers_read=2 fix_checks=2 dispatches=4
Gate edits: none detected (4 listed, 2 read; patterns: RH-1,RH-3,RH-4,RH-12,RH-13,RH-14,RH-15,RH-16; unlisted idioms: none)
Outside review: completed, no findings (openai/gpt-6.1-sol, host subagent).
Review notes
bin/gstack-diff-scope,bin/gstack-gate-diffandlib/gate-diff/*produce the auth, migration and gate-scan signals, but an edit to them still plans at light depth. This gap predates the PR and is outside the brief. It's a follow-up candidate, with its own no-loss repin.scripts/harness-version.jsonedit planned before v1.94.20.0 test(review): one file per review and harness pin, so independent repins merge cleanly (GST-1084) #3903, so it costs review time and loses no coverage.test/segment", but only the pins underscripts/harness-pins/test/do.Exploratory QA
Smoke:
bin/gstack-review-planfrom main and from this branch on three scratch commits off 1ccab0d. A helper plus its pin goes from tests with 0 dispatches to code with 2. Edits tolib/review-exclude.tsandlib/default-base.tsgo from code/light to gate/full with 4 dispatches. All pass on current inputs.Design Review
No frontend files changed, so design review was skipped.
Eval Results
No prompt-related files changed, so evals were skipped.
Review bot comments
No review bot comments.
Plan Completion
Plan completion audit: not run (no plan is bound to this branch and no docs/designs/ file matches).
TODOS
No TODO items completed in this PR.
Documentation
Scope: I compared the diff against 1ccab0d (lib/review-depth.ts, docs/review-policy.md, docs/ship-policy.md, the changelog fragment) with a repo-wide grep of markdown and templates for
review-exclude,default-base,harness-pinsand gate-path lists.Result: The parent's edits are accurate against the code. docs/review-policy.md lists
harness-pinsas code class and as a data-rule config file, and listslib/review-exclude.tsandlib/default-base.tsamong the machinery gate paths. docs/ship-policy.md updates thegate-pathandconfigkept-reason rows. The other mentions (docs/TESTING_INTERNALS.md, docs/troubleshooting.md, docs/test-audit-2026-10.md, review-policy.md's history paragraph at line 436) describe pin storage or thereview_exclude_pathsdigest, not the path classes, so they need no change. I made no edits.Debt: None found.
Documentation exception: accepted by GStack Minder 2026-11-03 via GSTA-36 via parent relay: docs/review-policy.md (main's auto-merge plus the GST-972 lines) and the review-depth.ts.json pin, read by the scoped fix-check at 4bdd9d2 with 0 findings, no docs re-audit
Test plan
bun test test/review-depth.test.ts -t "GST-972": 5 pass (4 failing-first tests fail at 1ccab0d)bun test test/review-packets-removed.test.ts: 9 passbun run typecheckclean;bun run typecheck:test: none newGSTA-258-C+GST-972; typecheck clean, 289 focused tests pass on the merged tree