Repository navigation
fix(release): forward-port crash, hang and fail-open fixes to 0.12 - #29386
Conversation
Sensitive Change ReviewSensitive files changed:
How to clear this check:
Only the author's comment counts this way. A reviewer can instead |
First Principles Review (Opus 5.5) — ✅ PASSPremise-level review of First-Principles-Verdict: PASS Nothing blocks. All five picks are fixes already merged on main and linked by issue. The one reversed pin, the "soft fail-open" comment in What this change shipsInventory (8 items) — 8 justifiedIntent: let 0.12.0 RC users get main's hang, launch-crash, macOS sub-agent and governance fail-open fixes. FIX.
[FIRST-PRINCIPLES-REVIEWED] 7c0f212 Description read[DESCRIPTION-READ] a0512e684a98dbd7ff6e0c89c35e9a7246f020b925c061c302fb2511cc77f4aa That sha256 names the PR description this verdict read. Recompute it with |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS These are byte-identical picks of fixes already merged on main. The one conflict resolution keeps the fail-closed wiring, and a mutation test pins it. Notes
[DESIGN-REVIEWED] 7c0f212 Description read[DESCRIPTION-READ] d94d58e7b33e2d5dacd079734834f92615b6166239ea9d4dd4afe0d02323a15b 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 |
UX Review (Opus 5.5) — ✅ PASSUX-level review of UX-Verdict: PASS Nothing to check. This PR adds or changes no dashboard controls, labels or states. It brings backend, packaging, dependency and test fixes over to the release branch, so having no screenshots leaves no evidence gap. [UX-REVIEWED] 7c0f212 Description read[DESCRIPTION-READ] ba35c95de7c051aef97d2ecd2116b445c5c82e383923d04f6bb5deccd8a6e305 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 |
Cherry-picks (git cherry-pick -x), in this order: - fix(i18n): raise the i18next floor to 26.4.4 so an unclosed nesting marker cannot hang the tab (#22326) (cherry picked from commit 7efc766) - fix(desktop): make the Linux packages depend on libgbm (#28980) (cherry picked from commit 068b96b) - fix(platform): read a macOS all-zombie process group as empty in pgroup_exists (#27615) (cherry picked from commit e1b9965) - fix(platform): read a macOS all-zombie group's killpg EPERM as an emptied group (#27734) (cherry picked from commit 7d4da6b) - fix(governance): judge cron turns by their session and fail closed on unattended turns (#27745) (cherry picked from commit 3fa3e80) Conflict, hand-resolved: - #27745, src/kiro_crew/slack/gateway_runtime/cron_turn.py (both stream calls): on main the added session_key=/unattended=True lines sit next to on_resume=, which #27508 added and 0.12 lacks. Took the two added keyword arguments at both sites and left on_resume= out, so #27508 is not pulled in. Every other file matches the source.
Opus 5.5 Review — 🔴 changes requested (blocking)Reviewed The cron governance change sends the gate the cron session key but not the agent, so a governance profile bound to that agent never applies to its cron turns. BLOCKING — src/kiro_crew/slack/gateway_runtime/cron_turn.py:265 [BLOCK-MERGE] 7c0f212 Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
e92de71 to
7c0f212
Compare
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: |
This Opus verdict on |
7b02f20 to
7d20584
Compare
7d20584 to
73e5ec2
Compare
73e5ec2 to
7c0f212
Compare
Supersedes the disposition posted earlier for this head. |
|
/ai-review override scope 7c0f212 infra-failure: the earlier Security Scope Review run on this head failed only because a GitHub API rate limit stopped it reading PR comments (https://github.com/kirodotdev/KiroCrew/actions/runs/41742630150); the rerun on the same head passed. |
Human judgment recorded@buluoray recorded a infra-failure release (class
This decision applies only to this commit. A new push requires a new judgment. |
Problem / Motivation
Goal: Bring the wide-impact crash, hang and fail-open fixes from main to
release/0.12.0so the next 0.12.0 RC ships without themAll five are merged on main and absent from
release/0.12.0:$t(marker freezes the tab, and reopening it freezes it again. The release lockfile pins i18next 26.3.6.libgbm.so.1: cannot open shared object fileon hosts without libgbm.PermissionError) when its process group holds only zombies.Why it matters
Each one is a hang, a launch crash, wrongly failed sub-agents on every macOS host, or a fail-open security gap on 0.12. The release branch takes only fixes of that weight. Every fix must land before the bare-version drop PR and the promotion RC.
Not a goal
/creates a bogus worker tree)What changed (motivation → approach → change)
One squashed commit; the per-pick breakdown is in its message. Picks in order, each with
git cherry-pick -x:7efc7669a46e068b96bbae12pgroup_existse1b9965677ackillpgEPERM7d4da6bb99a53fa3e8024d47cron_turn.pyThe four clean picks are byte-identical to their source diffs.
#27745 conflict,
src/kiro_crew/slack/gateway_runtime/cron_turn.py, both_cron_stream_with_posttoken_resumecalls (_cron_run_sequenceand_cron_run_single). On main the added lines sit next toon_resume=, which #27508 added. The resolution adds exactly the source's two keyword arguments at both sites and leaveson_resume=out:Why it still fails closed:
_cron_stream_with_posttoken_resumeforwards**stream_kwargstostream_and_collect, which on 0.12 already takessession_keyandunattended. Withunattended=Trueit wraps the turn inhooks.unattended_turn, andhook_runtime/governance_gate.pyreturnsBlocked by governance policy: it could not be evaluated for an unattended turnwhen evaluation raises. Those two files,hooks.py, the heartbeat and monitor-wake sites, and the newtest/test_governance_unattended_turns.pymatch the source commit exactly.Backwards compatibility
Compatible: each change matches its main PR. The only tightening is #27745's: an unattended turn whose governance cannot be evaluated is now refused, the behaviour main already has.
Removes nothing: no key, path or record is removed.
Tests
Proof: not provable -- other: release-branch picks of merged fixes; proofs ran on each source PR
test_cron_sequence_member_alias.pyandtest_unattended_slot_guardrails.py, in one process: 890 passed.unattended=Truefrom bothcron_turn.pysites failstest_cron_turn_hands_the_gate_its_session_key_and_marks_itself_unattended.npm ci, thenvitest run src/i18n/nestingRegexBacktrack.test.ts: 4 passed on i18next 26.4.4.packaging.test.jslibgbm test passes. Two NSIS tests in that file fail locally only becausewebsite/electron/node_modulesis not installed in this worktree.Manual verification
N/A - unit coverage from the source PRs applies unchanged; the conflict is covered by the tests above.
Screenshots / video
Why no screenshot: the
website/change is a dependency floor bump and a unit test, with no rendered difference.Related Issues
Refs #22326
Refs #28980
Refs #27615
Refs #27734
Refs #27745
Sibling sweep
Siblings: none -- picks of merged fixes; their own sweeps ran on main.
Pattern harvest
Not generalizable: release-branch picks of fixes already harvested on their source PRs.
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.