Repository navigation
Bump the pinned openRepoShape submodule to 0fa73b7, past the park/resume carve - #5
Conversation
…ume carve openRepoShape's `main` advanced past the park/resume carve (opensoft/openRepoShape#94, slice S2 of #92): `park` and `resume` left its tree for this repository's own copies. Bump `contracts/openreposhape-pin.yaml`'s `commit:` and `digests.tree_sha256`, and the `upstream/openRepoShape` gitlink, together in one commit, per the README's "Bumping the pin" procedure — the edit lands before the `git add`, so the moved gitlink never lands paired with the old pin file. The digest is recomputed with the pinned checkout's own `scripts/repo_shape.py`, never adjusted to fit. `tests/test_upstream_pin.py`'s pre-carve witness entry for `park` is deleted here too, as its own docstring said to: 0fa73b7's tree no longer carries the file `test_park_resume_commands.py` used to diff against. Lane: openreposhape-2 (openRepoShape-2) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
There was a problem hiding this comment.
Sorry @brettheap, you've used your own review budget of 250,000 diff characters for the last 7 days.
You can request another review in 4 hours and 32 minutes by commenting @sourcery-ai review. Upgrade to get a review now.
|
@codex review |
Reviewer's guide (collapsed on small PRs)Reviewer's GuideThe PR updates the openRepoShape submodule and lockfile from d090a7a to main-tip 0fa73b7, recomputes the tree digest, and removes the obsolete park fixture-presence assertion now that the upstream carve has deleted that file. The implementation preserves the remaining fixture paths and verifies lockstep pinning through the existing test suite. File-Level Changes
Assessment against linked issues
Possibly linked issues
Tips and commandsInteracting with Sourcery
Customizing Your ExperienceAccess your dashboard to:
Getting Help
|
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
There was a problem hiding this comment.
🟢 Approval recommended
The pin and digest update are internally consistent, and the upstream-content verification test is correctly updated to reflect the removed park file.
Pull request overview
This PR bumps the pinned upstream/openRepoShape submodule and its recorded tree digest to commit 0fa73b7…, which is after the upstream change that removed park/resume from openRepoShape (since those commands now live in this repository). It also updates the pin-verification test to stop asserting presence of park in the upstream tree, while preserving the remaining fixture-file integrity checks.
Changes:
- Update
contracts/openreposhape-pin.yamlto pin openRepoShape at0fa73b7…and refreshdigests.tree_sha256. - Update
tests/test_upstream_pin.pyto remove"park"from the verified upstream file list and adjust the docstring rationale accordingly.
File summaries
| File | Description |
|---|---|
tests/test_upstream_pin.py |
Removes the park path from the “pinned commit carries what the suite reads” list and updates the explanatory docstring to match the post-carve upstream tree. |
contracts/openreposhape-pin.yaml |
Advances the pinned commit SHA and recorded tree digest for the openRepoShape submodule. |
Review details
- Files reviewed: 3/3 changed files
- Comments generated: 0
- Review effort level: Lite
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
…did not (F-X18, A11 Addendum 4 ruling 13) `lane-start:1027` and `lanes-edit.sh:4794` both cited act 0 — the register veto this branch builds its second layer beside — as `opensoft/brett-wip#5` @`95e7a4c`. Measured in the live checkout: ``` $ git merge-base --is-ancestor 3719d97 origin/main && echo YES # YES $ git merge-base --is-ancestor 95e7a4c origin/main && echo YES # (nothing) $ gh pr view 5 -R opensoft/brett-wip --json headRefOid,mergeCommit,mergedAt headRefOid 95e7a4c… mergeCommit 3719d97… mergedAt 2026-09-13T19:14:37Z ``` `#5` merged SQUASHED, so its pre-merge head is not an ancestor of `origin/main` at all: a comment citing it names code that never landed, in the one place a reader goes to check that this branch does not re-implement act 0's veto. A sha in a comment is a claim about history, and this one was false. Ruling 13, ratified *"a11 addendum 4 yes"*: *"Act 0 is cited by its merged sha `3719d97` everywhere; `95e7a4c` stops being quoted as what landed."* SPEC rev 6 §6/§13 already carries it on the text's side, and `workBenches#71`'s skill already said `3719d97`. Both comments now say `3719d97` AND still name `95e7a4c` — as the thing that did not land, with the measurement, so the next reader who finds the draft head in an older document knows which of the two it is. **Test** — `tests/test_repo_hygiene.py` walks every tracked file for the citation FORM (`opensoft/brett-wip#5` @`<sha>`), which is the shape that asserts *"this is act 0"*, and requires the sha to be the merge commit. Naming the draft head in prose stays legal, because saying which sha did not land is the useful half. **The citation WRAPS, and the first version of this test missed the one that did.** `lane-start` writes it across two comment lines, so the continuation's `#` sits between the two halves of the pattern; a line-by-line match found only `lanes-edit.sh`'s one-line citation and PASSED the mutation. Measured, then fixed: the text is joined on comment continuations before matching, and the test also asserts it found at least one citation at all — so a change to the citation form fails loudly rather than silently testing nothing. **Mutation** — `lane-start`'s citation put back to `95e7a4c`: **1 failed**, `assert not {'lane-start': ['95e7a4c']}`. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Lane: openreposhape-2 (openRepoShape-2)
Claim: #2 (comment)
Closes #2
Bumps
contracts/openreposhape-pin.yamland theupstream/openRepoShapegitlinkfrom
d090a7a141048b214f0420a8c7afbfff62472d1ato0fa73b7df704266589dcc27afca020c0127fa36f— openRepoShape'smainafteropensoft/openRepoShape#94 (slice S2 of opensoft/openRepoShape#92), which moved
parkandresumeinto this repository and deleted them from that one.Verified before pinning
gh api repos/opensoft/openRepoShape/compare/main...0fa73b7df704266589dcc27afca020c0127fa36f -q .status→
identical:0fa73b7IS openRepoShape'smaintip, so it is a commit onmainthe next fetch can always reach, never a branch commit a squash-merge orphans.test_park_resume_commands.pyreads did not change betweenthe two commits:
git -C upstream/openRepoShape diff --stat d090a7a 0fa73b7 -- templates/ scripts/repo_shape.pyis empty (confirmed;parkandresumeare theonly relevant deletions, and neither is under
templates/orscripts/repo_shape.py).Followed the README's "Bumping the pin" procedure literally
The procedure held exactly as written — no README changes were needed. The digest
was recomputed with the pinned checkout's own
scripts/repo_shape.py, never adjustedto fit, and the edit landed before the
git addso the gitlink and the pin file movedtogether in one commit.
The one deliberate extra edit, in the same commit
tests/test_upstream_pin.py'stest_the_pinned_commit_carries_what_this_suite_readsparametrized
"park"as "the pre-carve witness" for reviewers of the T1 carve. Its owndocstring named the exit: delete that one entry in the same commit that bumps the pin
past the deletion, because
0fa73b7's tree no longer carries the file. Deleted here,and only that entry — the other eight paths in the parametrize list are untouched, and
the docstring's title and body are updated to say why, in one added line.
Tests
python3 -m pytest tests -q→ 140 passed(141 at the old pin; exactly one fewer — the deleted
parkparametrize case).git clone --recurse-submodulesagainstthe real, pushed remote): the submodule checks out at
0fa73b7on its own, andpython3 -m pytest tests -q→ 140 passed, exit 0 — the lockstep test (git ls-tree HEADvs. the working-tree pin file) is what CI itself will run.skipped, exit 0, every skip naming
git submodule update --init upstream/openRepoShape(88 skipped at the old pin — again exactly one fewer).git ls-tree HEAD -- upstream/openRepoShape→160000 commit 0fa73b7df704266589dcc27afca020c0127fa36f, matchingcontracts/openreposhape-pin.yaml'scommit:exactly.bash -nonpark,resume,openRepoTools: all three OK, all three untouched —git diff --stat main -- park resume openRepoToolsis empty.🤖 Generated with Claude Code
Summary by Sourcery
Advance the openRepoShape dependency past the park/resume carve while keeping the repository pin metadata and validation tests in sync.
Enhancements:
Tests: