Repository navigation
fix: backport 0.12.1 batch 43 (10 user-reported fixes) - #32641
Conversation
When a safety policy blocks a shell call that chains several commands (git add && git commit && git push), nothing in it runs, including the steps before the one that was refused. The host notice named the rule but not that, so the model could assume the commit had landed. The policy notice, and the fallback recovery prompt a harness without steer gets, now say no part of the call ran, a chained command's earlier steps included. No matcher changes. Closes #25248 (cherry picked from commit bc3d9cc)
…31248) AppStorage.get() read each value file with a bare read_text() and turned any OSError into None. On Windows a read that overlaps a concurrent set()'s atomic replace raises a sharing violation, so get() answered "absent" for a key that is present, and a get-or-default caller then overwrote the live value with its default. _read now goes through read_bytes_with_retry, as its sibling _read_digest_key already does, so that window is retried off the event loop. An absent or damaged file still reads as None; the write and delete paths are unchanged. Builds on #26839. Closes #26827 (cherry picked from commit f7078c0)
The Add abilities success banner offered Open Approvals as a stock Btn: a transparent fill with a --border edge. On the banner's translucent ok tint that edge measures 1.08:1 (light) and 1.21:1 (dark), so the control only read as a button on hover. It now rests on --bg-elevated with a --border-strong edge; hover is unchanged. The shared Btn is not touched. Fixes #30745 ## Pattern harvest Not generalizable: one call site places a default Btn on a tinted banner; other banners already style their buttons per site. (cherry picked from commit 93c7820) (cherry picked from commit 909165ce7518d6c5bbc1f0f887a8e0255f33381a) Original author: Matthew Briones (#31924).
…nstall (#31955) A service never opens a browser, so the install success block is the only guide a new user gets. Add an 'Open: kirocrew token' line ahead of the status/logs/remove hints on both systemd and launchd. Fixes #31911 (cherry picked from commit 1109563) (cherry picked from commit ba85997898619677046baee4dfbca42f5fc4215f) Original author: Zezhen Xu (#31955).
(cherry picked from commit 1ab5673)
) A composer the user had resized by hand kept its height when the list-indent hint row appeared. The text column shrank below the textarea's 44px floor, so the textarea overflowed onto the hint and the typed prompt and the hint were painted in the same place. With the plain textarea, the manually sized box now never goes below what its rows need: the wrapper takes a min-content floor and the text column carries the textarea's 44px, so a row shown inside the box grows it instead of sitting under the text. A taller manual height and the auto-sized composer keep their sizes. The rich-text editor, which never shows the hint, keeps min-h-0: its draft would otherwise count toward the floor and grow the box line by line. Closes #25166 (cherry picked from commit 4a0b139)
…32177) On a busy host the keyed mount-source sweep's host-wide pin scan took longer than the whole 10 s pass budget. The scan runs lazily at the first directory candidate inside that budget, so the next budget check ended the pass and each pass reclaimed one batch, fewer entries than the readiness probes stage between passes. Take the scan's own duration off the budget clock, as the legacy pass already does by starting its clock after its scan. The scan still runs at most once per pass and only when a directory candidate exists, and the directory gate is unchanged. Closes #16887 (cherry picked from commit 727392d)
…2132) * fix(portability): export a data home spelled with the \\?\ prefix * fix(portability): keep fenced files out under a \\?\ data home The fence builds its targets from the root's own spelling, so ask it about that spelling as well as the folded real path. Refs #27855 (cherry picked from commit 6f63fe7)
Sensitive Change ReviewSensitive files changed:
How to clear this check:
Only the author's comment counts this way. A reviewer can instead |
UX Review (Opus 5.5) — ✅ PASS (a BLOCK with no Trigger and no Outcome does not block)UX-level review of UX-Verdict: PASS cannot evaluate: missing screenshots of the two dashboard fixes, the restyled "Open Approvals" button and the composer list-indent hint. The PR attaches no image, so the blind read was not performed. Evidence gaps
Notes
[UX-REVIEWED] 15b2303 Description read[DESCRIPTION-READ] e377d5c0e94780935236b16f94e9804406e5dd0e49586296d5e41a20c85a890c That sha256 names everything this verdict read: the PR description, plus the 3 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 |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Nothing blocks. These are ten clean cherry-picks of fixes that already merged on Notes
[DESIGN-REVIEWED] 15b2303 Description read[DESCRIPTION-READ] e37060669e1b2d316b2d24f6fe1b8ed5077c32fcd8392f91482b56e0861c4021 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 This is a backport of ten fixes already merged on Notes
What this change shipsInventory (10 items) — 10 justifiedIntent: give 0.12.x users ten user-reported fixes that already merged on
[FIRST-PRINCIPLES-REVIEWED] 15b2303 Description read[DESCRIPTION-READ] a60a30a8d024bc3a6c0e5b9df07ba7c546ce019972df655801fb041571053355 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: |
Problem / Motivation
Goal: Bring ten small user-reported fixes that merged on
mainafterv0.12.0ontorelease/0.12.0, one commit each, in a PR of its own.Why it matters
Each fix closes an issue a user filed. Each faulty line was checked on
release/0.12.0before the pick; the right-hand column names it.deny_notice.py:57slack/transport_dispatch.py:1469apps/app_storage.py:179dashboard/handlers/worktree.py:841crew/WorkerKitPane.tsx:227service install, nothing says how to open the dashboard.service/controller.py:136%in a path breaks the pi gate launcher.acp/harness/pi.py:559ChatInput.tsx:1440sandbox_mount_sweep.py:811\\?\prefix cannot be exported.portability_export.py:325Not a goal
main; nothing is changed only here.CHANGELOG.mdchange.What changed (motivation → approach → change)
Each main squash commit was cherry-picked with
-xontorelease/0.12.0@04891384c7, in merge order onmain, one commit each. Every pick applied cleanly; every file has added and removed lines identical to its main commit.bc3d9cc31e4995ebc893406e8c6e7dde7020dda5f7078c0f6cfe5e1a4c7f678769e234fab603e90a93c7820f07d285a144cf1109563a0ebf26029c6a1ab5673b1749554457874a0b1392d4e49a84b5ff727392d0d0bb0bad8f55\\?\prefix6f63fe7b6d15b2303cee#31924 and #31955 carried a co-author trailer with an internal address, which the push guard refuses, so those two commits are authored with the public identity and name their original author in the message; the trailer line is removed. The other eight keep their original author.
builder-security checked: no bind, listener, route, auth mechanism or storage grant changes.
Backwards compatibility
Compatible: each change is the one already merged on
main, and its compatibility note applies unchanged.Removes nothing: no setting, key or caller is removed.
Tests
Proof: not provable -- other: a backport of main commits whose proofs ran on their own PRs
release/*, so the related set was run here: every touched test file plus every test file that imports a touched module, 46 files,-n 4 --dist loadgroup. 2056 passed, 67 skipped, 0 failed.tsc -p tsconfig.app.json --noEmitpasses.vitest run ChatInput WorkerKitPane: 75 files, 1060 passed, 3 failed. The 3 are inChatInput.approval.test.tsxand fail the same way on a cleanrelease/0.12.0@04891384c7.Manual verification
N/A - each fix was verified on its own main PR. The 0.12.1 RC soak is the end-to-end check.
Screenshots / video
Why no screenshot: the two frontend picks (#31924, #32311) carry the same rendered change as their main PRs, whose screenshots show it.
Related Issues
Backport to the 0.12 line for 0.12.1 of #30950, #31206, #31248, #31414, #31924, #31955, #32142, #32311, #32177 and #32132. Source issues: #25248, #27147, #26827, #22137, #30745, #31911, #27851, #25166, #16887, #27855.
Sibling sweep
Siblings: none -- each source PR carried its own sibling sweep; this PR only carries those fixes.
Pattern harvest
Not generalizable: this PR only carries fixes already on
main. The source PRs have their own harvest.Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)Contribution License Agreement
Contributions are accepted under the project's Apache-2.0 license; see the Licensing section of CONTRIBUTING.md.