Repository navigation
fix(apps): pin the adopted apps at commits carrying the wheel's fixes - #31343
Conversation
The five adopted former built-ins were pinned at repository snapshots cut before several fixes landed in the wheel copy, so a host that adopted ran code missing fixes a built-in host had. Each repository now carries those fixes at a patch release; pin those commits. The wheel built-in's version moves to the same patch release, because a first-adoption pin must equal it (D12): otherwise every host that has not adopted yet would refuse the new pin with version_mismatch.
|
Catalog pair: kirodotdev/KiroCrewApps#68 moves the five catalog |
UX Review (Opus 5.5) — ⏭️ skippedRevision |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Nothing blocks: this only bumps the pinned data, with matching wheel versions under the D12 equality rule, and it takes nothing away. [DESIGN-REVIEWED] 475fb27 Description read[DESCRIPTION-READ] 3f052862b6f84059f17b4848b549df1e15c60c8c036d572cf6db91337f2310d3 That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
First Principles Review (Opus 5.5) — ✅ PASSPremise-level review of First-Principles-Verdict: PASS Nothing blocks. This is a data-only fix: each pin moves to a commit that carries the wheel copy's fixes, and each wheel version bump is required by the existing What this change shipsInventory (6 items) — 6 justifiedIntent: a host that adopts one of the five former built-in apps should run a copy with every fix the built-in copy has. This is a FIX.
[FIRST-PRINCIPLES-REVIEWED] 475fb27 Description read[DESCRIPTION-READ] 70164a8a4604c9681624b1ae8722efcb39e3909e08fd936e5194a4bb2a561401 That sha256 names the PR description this verdict read. Recompute it with |
Opus 5.5 Review — ✅ no blocking findingsReviewed Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
buluoray
left a comment
There was a problem hiding this comment.
🤖 Kiro Crew Auto-Pipeline [operator: buluoray]:
APPROVE.
What it does for users: hosts that adopt one of the five former built-ins now get the same fixes as hosts still on the built-in copy. That covers the papyrus Windows userspace TeX probe, the non-ASCII group names and field labels in personal-shopper, the file-explorer search and Windows path fixes, the design-critique translations and accessible names, and the pptx-maker engine install fixes. The wheel app.json versions move up one patch, so first adoption is not refused with version_mismatch.
Each pin was checked against its app repository through the GitHub API:
- personal-shopper e3d71dec00. The commit exists and is the head of main. It is 1 commit ahead of the old pin and 0 behind. app.json at that commit says 0.1.1. CI: 6 checks, all success.
- papyrus 6ca7bc697f. Exists, head of main, 1 ahead and 0 behind. app.json says 0.2.1. 6 of 6 success.
- file-explorer e68646984d. Exists, head of main, 1 ahead and 0 behind. app.json says 1.0.1. 6 of 6 success.
- design-critique 3618805505. Exists, head of main, 1 ahead and 0 behind. app.json says 1.0.1. 7 of 7 success.
- pptx-maker 1c47c5099d. Exists, head of main, 2 ahead (one is a merge commit) and 0 behind. app.json says 0.3.1. 6 of 6 success.
Fixes carried. I downloaded each pinned archive and checked every fix the body lists against it:
- All 21 listed fixes have a row in that app's PORTING.md, keyed by the core merge commit: 2 for personal-shopper, 1 for papyrus, 7 for file-explorer, 7 for design-critique and 4 for pptx-maker.
- I compared the lines each fix PR added to the wheel copy with the pinned tree. Many added lines match exactly. The rest differ only for porting reasons. The app repositories use their own i18n namespaces: t('historyTab.search_label') where core has i18nT('apps.personalShopper.historyTab.search_label'). They replace core helpers with local ones: relativeToRoot moves to ui/src/lib/platform.ts, and safe_read_prefix and FileTooLargeError come through the host bridge with a fallback. Some lines were reformatted, such as the casefold group lookup in backend/store.py:432.
- I opened the remaining differences by hand. design-critique's AskLayer question_label (line 89) and the step-range and screen fallbacks are all present. pptx-maker's start button is disabled={!engineReady} where core has disabled={startChat.isPending || !engineReady}. The engineReady gate is the #30185 fix, and the app's startChat is a plain function, not a mutation, so it has no isPending.
- No fixes appear to be missing. I listed every core commit since each old pin's date that touches the five wheel copies. Everything not in PORTING.md is test-only, a core-independence refactor, or (for #27451, file-explorer aria-expanded) already in ui/src/TreeNode.tsx:99.
The repo's own verifier, run over the network at this head, reports ok for all five: tree and archive digests match, and the pinned-tree checks pass.
Tests: test/test_builtin_adoption_pins.py gave 47 passed. Mutation: putting papyrus app.json back to 0.2.0 reddened test_every_shipped_pin_matches_its_wheel_built_in_version ("assert '0.2.1' == '0.2.0'"), and the tree was clean afterwards.
CI on this head: 77 success, 16 skipped, 0 failing. PR Readiness is success. Design and First Principles both PASS. No reviews and no inline threads. The merge-tree against main exits 0.
Source-only: I did not boot a host to exercise first adoption or the catalog update path for an already-adopted host.
Follow-up: #31351 (the App repo: line check) would fail this PR's PR Hygiene on any later run, because this PR edits five wheel-copy app.json files. If this PR runs CI again after #31351 merges, add one App repo: line per app naming the pinned commit. That is a body edit, with no push.
Problem / Motivation
Goal: A host that adopts one of the five former built-ins runs a copy that has every fix the built-in copy has.
The adoption pins named repository snapshots cut before several fixes landed in the wheel copy: personal-shopper's non-ASCII group names (#30535) and field labels (#30493), papyrus's Windows userspace compiler probe (#30438), seven file-explorer fixes (#27870, #27995, #28649, #29854, #30177, #30505, #30445), seven design-critique translation and a11y fixes (#27392 through #30299), and four pptx-maker engine fixes (#28972, #28737, #30185, #31094). A host that adopted ran without them while a host still on the built-in had them.
Why it matters
Adoption is meant to be invisible to the user. Instead, adopting silently took fixes away: a Windows Papyrus user with only a userspace TeX saw "no compiler found" again, and a pptx-maker engine install could re-resolve to an mcp major it cannot import.
Not a goal
What changed (motivation → approach → change)
Each app repository now carries those fixes at a patch release, with green CI on ubuntu, macOS and Windows. This PR pins those commits: each row is the output of
scripts/adoption_pin.py compute, andadoption_pin.py verifyreports all fiveok(git tree and codeload archive digests match, and the pinned-tree checks pass against the shipped built-in).The wheel built-in's
app.jsonversion moves to the same patch release. A first-adoption pin must equal the wheel version (D12,pinned_tree_refusal). Without the bump, every host that has not adopted yet would block the new pin withversion_mismatch, a code that is never retried for that pin.How each host reaches the new code:
Backwards compatibility
Compatible: same names, repositories,
legacyNavandminAppSdkLevel. Each version moves one patch up, so neither the record-version check norwould_downgraderefuses it.Removes nothing: only the pinned commit and the matching wheel version change.
Tests
test_the_shipped_pin_file_validates_and_pins_exactly_the_shipped_tabletakes the five new rows;test_every_shipped_pin_matches_its_wheel_built_in_versionholds the bump.Proof: PROVEN test/test_builtin_adoption_pins.py::test_the_shipped_pin_file_validates_and_pins_exactly_the_shipped_table (head 475fb27)
Determinism: no test logic changes, only the pinned data table.
Locally: the seven adoption test files, integration boot included, give 226 passed and 3 skipped. The related backend set gives 4399 passed and 13 skipped. black, subprocess-encoding, isort, flake8, comment-history, brand and line-cap gates are clean.
Manual verification
scripts/adoption_pin.py verifywas run over the network: all five rowsok.Related Issues
Part of #26967
Sibling sweep
Swept: git log --since=2026-11-01 -- src/kiro_crew/apps/builtins/{personal_shopper,papyrus,file_explorer,design_critique,pptx_maker} website/src/apps/{personal-shopper,papyrus,file-explorer,design-critique,pptx-maker}, checked against each repository commit's list of carried fixes
Sibling: src/kiro_crew/apps/adoption-pins.json -- changed here
Sibling: src/kiro_crew/apps/builtins/personal_shopper/app.json -- changed here
Sibling: src/kiro_crew/apps/builtins/papyrus/app.json -- changed here
Sibling: src/kiro_crew/apps/builtins/file_explorer/app.json -- changed here
Sibling: src/kiro_crew/apps/builtins/design_critique/app.json -- changed here
Sibling: src/kiro_crew/apps/builtins/pptx_maker/app.json -- changed here
Pattern harvest
Rule candidate: review-prompt
Pattern: a change under the wheel copy of an app with an adoption pin also needs the same fix in that app's repository and a pin bump; otherwise adopted hosts lose it
Checklist