Skip to content

fix: backport 0.12.1 batch 45 (9 user-reported fixes) - #32653

Merged
buluoray merged 9 commits into
release/0.12.0from
crew/slack/0.12.1-batch-45
Nov 7, 2026
Merged

buluoray merged 9 commits into
release/0.12.0from
crew/slack/0.12.1-batch-45

Conversation

@buluoray

@buluoray buluoray commented Nov 7, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: Bring nine small user-reported fixes that merged on main after v0.12.0 onto release/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/0.12.0 before the pick; the right-hand column names it.

PR Reported by What a 0.12.0 user sees today Faulty code on release
#30912 prakyg (#26246) A cron script's tool call is sent with a redacted argument, silently. cron_script_context.py:321
#30970 jongaffen (#27810) The security event log's hash chain breaks when its tip read fails. sel.py:305
#26789 rohnkum (#26784) Reverting an artifact keeps the current render kind, not the target version's. artifacts_writes.py:706
#31746 achayapa (Slack) A publish provider name with an underscore is rejected. artifacts_parts/publish_sharing.py:53
#31813 RobStebler (#20451) Gateway restart logs a WARNING traceback when a sibling session holds the cli.json lock. providers/acp.py:318
#31607 gwapoats (#26092) A crashed gateway waits on its core dump before restarting. service/linux_unit.py:388
#31510 kkautenburger (#29365) A member backup inside the agent sandbox fails with a bare errno. member_memory_backup.py:93
#30900 panlm (#29463) A channel sender's access is not re-checked before its turn begins. messaging/dispatch.py:1071
#32082 jongaffen (#27830) On Windows, a dashboard copy from a \\?\ path fails the containment check. dashboard/file_api/workspaces.py:148

Not a goal

What changed (motivation → approach → change)

Each main squash commit was cherry-picked with -x onto release/0.12.0 @ b80bdd7f7d, in merge order on main, one commit each.

# PR main commit release commit
1 #30912 refuse a cron tool call whose argument would be sent redacted 71f6de026e 205386f04c
2 #30970 keep the audit chain linked when its tip read fails cfdec3bbd0 ba4cf93e15
3 #26789 restore the target version's render kind on revert 820ec27645 d57065bb6c
4 #31746 accept an underscore in a publish provider name e51597aa10 751dff8e6f
5 #31813 write the cli.json overlay only off the event loop bd49ea0a87 0a0ccceb3f
6 #31607 stop core dumps from delaying a crashed gateway's restart c40e8b7b7a dabf1598d3
7 #31510 say why a member backup fails inside the agent sandbox 8237603a5f 9aec0780b4
8 #30900 tighten a channel sender access check 05603c8576 4467e609bc
9 #32082 fold the \\?\ prefix in dashboard copy_from containment checks 7a2bf86514 0122dc6bbe

Hand-resolved hunks, docs and tests only:

PR File Resolution
#31813 docs/system-specs/modules/harness-parity.md H6 row Main's one-phrase edit applied to release's longer H6 wording.
#31607 test/test_pod_systemd.py docstring Kept main's docstring, which names the LimitCORE check.
#31510 test/test_member_memory_backup.py imports Added only import errno; release never imported gc.

Every other file has added and removed lines identical to its main commit. #32082 also carries its one-row update to .github/sync-io-in-async-fs-baseline.txt (a line count for workspaces.py).

The three hand-resolved commits, #26789 (non-public author email) and #31746 (non-public author email) are authored with the public identity and name their original author in the message. The other four keep their original author.

builder-security checked: #30900 tightens a channel sender access check and #32082 a file-API containment check, the same changes already merged on main; 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

  • CI does not run the backend suite on PRs into release/*, so the related set was run here: every touched test file plus every test file that imports a touched module, 348 files, -n 4 --dist loadgroup, run on the first head that still carried fix: drop post-quiet-end text held for a steer the turn requeues #32396 (dropping it removes only its own files and tests). 12594 passed, 12 skipped, 8 failed. The same 8 fail on a clean release/0.12.0 @ b80bdd7f7d on this host (real-sandbox tests in test_snapshot_read_only_real_view.py, test_cron_run_scratch.py, test_app_script_cron_sibling_import.py, plus test_auto_improvement_members and test_harness_parity.py::test_is_kiro_cli_is_positive).
  • flake8 passes on the changed Python files.

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

N/A - no frontend file changes.

Related Issues

Backport to the 0.12 line for 0.12.1 of #30912, #30970, #26789, #31746, #31813, #31607, #31510, #30900 and #32082. Source issues: #26246, #27810, #26784, #20451, #26092, #29365, #29463, #27830.

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

  • At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)
  • Existing tests pass and new tests added for new functionality
  • Self-review completed; code follows project style guidelines
  • Documentation updated (if applicable)
  • No secrets, credentials, or internal references in the diff

Contribution License Agreement

Contributions are accepted under the project's Apache-2.0 license; see the Licensing section of CONTRIBUTING.md.

bolichen97 and others added 9 commits November 7, 2026 21:58
…30912)

A script cron's ctx.call_tool() runs the credential scan over its
arguments and sends the redacted copy. When the scan replaced a whole
argument, such as an opaque mail id the bare-secret pass mistakes for a
key, the server got only the placeholder and answered with its own
error ("Id is malformed"), so the script could not see the cause.

The call is now refused inside the audited call, with a ValueError
naming the argument's path and never its value. A credential-shaped
span inside a longer string is still sent with the span redacted, and
kirocrew cron preview runs the same scan. Nothing that was redacted is
sent unredacted.

Closes #26246

(cherry picked from commit 71f6de0)
When the security event log could not open its own file at startup
(another process held it without sharing, or EACCES), the chain tip
read returned "" as if the log were empty. The next record then landed
mid-file with an empty prev_hash, and `kirocrew security verify`
reported the honest log as compromised.

Only an absent or empty log, or one with no parseable record, now
starts the chain from genesis. Any other read failure raises:
construction still succeeds and marks the tip unknown, and the first
append reads it again under its locks. A re-anchor whose read fails
keeps the identity it compared against, so the next append reads the
tip again too. While the read keeps failing the append fails like any
unwritable audit: a critical audit is denied and a best-effort batch
takes the writer's existing failure path.

(cherry picked from commit cfdec3b)
…26789)

* fix(artifacts): restore the target version's render kind on revert

update() records version_kinds so a revert restores the right render-mode,
but on a revert it stamped the CURRENT kind: reverting to a pre-pull widget
snapshot left the artifact as html -- the raw inner fragment the recorded
kind exists to prevent. The revert callers send no kind, so the store is the
only place that can honour it. On event_type=="reverted" with from_version,
restore art.kind from version_kinds[from_version] before stamping the new
version.

Closes #26784

* fix(artifacts): never restore a dashboard or webapp kind on revert

The revert kind restore runs after the dashboard fence and write gate, which saw the current kind. Restoring into or out of dashboard/webapp there would skip them, so those kinds stay as they are.

---------

Co-authored-by: iamwhatever <12429733+iamwhatever@users.noreply.github.com>
(cherry picked from commit 820ec27)

Original author: rohnkum (#26789).
AcpProvider.__init__ wrote the effort and Tool Search overlays itself.
The session allocator builds the provider on the event loop, where a
cli.json lock acquire makes one attempt and never waits, so a sibling
session holding the lock for a moment refused the write and logged a
WARNING traceback at gateway restart.

start() already writes both overlays before every spawn, the first one
included, from a worker thread where the lock wait is real. The
constructor's write was a duplicate of it, so it is removed.

Closes #20451

(cherry picked from commit bd49ea0)

Original author: Bolin Chen (#31813). docs/system-specs/modules/harness-parity.md H6 row hand-resolved: the same phrase edit applied to the release wording.
Every gateway unit this repo writes left RLIMIT_CORE at systemd's
default, which is unlimited wherever core_pattern pipes to
systemd-coredump. A fatal signal then streams the whole multi-GB
gateway into systemd-coredump, the process cannot finish exiting until
that is done, and systemd-coredump stops at its 5-minute runtime cap,
so the dashboard and sessions stay down for about five minutes.

Set LimitCORE=0 in the unit `kirocrew service install` renders (user
and system scope), the pod template unit (and require it, so pod up
re-renders an installed pod unit without it), the EC2 bootstrap unit
and the two documented hand-written units. The faulthandler stack
dump stays the crash record; processes the gateway starts inherit
the limit, and a drop-in restores core files.

Closes #26092

(cherry picked from commit c40e8b7)

Original author: Bolin Chen (#31607). test/test_pod_systemd.py docstring hand-resolved: kept main's docstring, which names the LimitCORE check.
Inside the agent sandbox memory_stores is read-only by design, and
member backups are written inside it, so `kirocrew memory backup` and
the pre-update copy of `kirocrew update` run there failed every member
store with bare EROFS tracebacks. The seal stays. The backup pass logs
one line per member store instead: member stores are read-only in the
sandbox, back up from the dashboard or your own terminal, plus the
errno. It is decided from the refused write's path, the way
`kirocrew config set` already does, and the store-use lock no longer
tries to tighten its directory on a read-only mount, so no traceback
comes first.

(cherry picked from commit 8237603)

Original author: Bolin Chen (#31510). test/test_member_memory_backup.py import hand-resolved: added only import errno; release never imported gc.
A channel sender's access is decided once more before the agent
runs their message.

Closes #29463

(cherry picked from commit 05603c8)
@buluoray
buluoray requested review from a team and KiroCrewCaptain as code owners November 7, 2026 22:11
@buluoray
buluoray requested review from tim-finnigan and removed request for a team November 7, 2026 22:11
@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Nov 7, 2026
@github-actions

github-actions Bot commented Nov 7, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — ⏭️ skipped

Revision 0122dc6bbe100b7284e2341b640849da1685f237 touches no user-facing surface (no changes under website/ or committed screenshots), so the UX review was skipped. Does not block merge.

@github-actions

github-actions Bot commented Nov 7, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — ✅ PASS

Design-level review of 0122dc6bbe100b7284e2341b640849da1685f237 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

Design-Verdict: PASS

Nothing blocks: each of the nine hunks backports a fix already merged on main, and I found no reader outside the patch that breaks.

Notes

  • ## Backwards compatibility has both a Compatible: line that names no reader and a Removes nothing: line. Two items do refuse what used to pass: the cron call_tool now raises ValueError for an argument that was entirely redacted, and the channel turn now refuses at _begin_turn a sender who is no longer admitted. Neither breaks a reader, because the old behaviour sent a placeholder or a turn that should not have run. A Reader: line for each would make the section accurate.
  • The description says "Manual verification: N/A" while carrying LimitCORE=0 (Linux systemd only) and the \\?\ prefix handling (Windows only), and CI does not run the backend suite on release/* PRs. The main PRs' proofs cover the same code. Make sure the 0.12.1 RC soak includes a Linux service restart and a Windows dashboard copy.

[DESIGN-REVIEWED] 0122dc6

Description read

[DESCRIPTION-READ] d0a8453f3474bafae8e1465c84f7b9ee9aeff3edfedd404929a2f3ff11289ae1

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 .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Nov 7, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — ✅ PASS

Premise-level review of 0122dc6bbe100b7284e2341b640849da1685f237 — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

First-Principles-Verdict: PASS

Nothing blocks: this backports nine fixes already merged on main. Each one cites a user-filed issue and names its faulty line on the release branch.

What this change ships

Inventory (10 items) — 10 justified

Intent: give 0.12.x users nine user-reported fixes that main already has, without waiting for 0.13. FIX.

  1. A cron script's tool call is refused, naming the argument's path, when that argument would be sent fully redacted — justified
  2. The security event log's hash chain stays linked when reading its tip fails — justified
  3. Reverting an artifact restores the target version's render kind — justified
  4. A publish provider name with an underscore is accepted — justified
  5. The cli.json overlay is written off the event loop at start, so a lock held by a sibling session no longer causes a WARNING traceback — justified
  6. The gateway's systemd unit, the docs' sample units and the EC2 template set LimitCORE=0, so a crash restarts without waiting on a core dump — justified
  7. An installed pod unit missing LimitCORE= counts as stale and is re-rendered — justified
  8. A member backup inside the agent sandbox fails with a reason instead of a bare errno — justified
  9. A channel sender's access is checked again before its turn begins (messaging and iMessage dispatch) — justified
  10. On Windows, a dashboard copy from a \\?\ path passes the containment check — justified

[FIRST-PRINCIPLES-REVIEWED] 0122dc6

Description read

[DESCRIPTION-READ] 53be4ee00c48c0594158162ea94bd69d818c47856a4ea79a932d9b9457de9d3b

That sha256 names the PR description this verdict read. Recompute it with .github/scripts/pr-description-capture.sh; if it differs, the description changed after this verdict was formed and any finding drawn from the description is unproven.

@github-actions

github-actions Bot commented Nov 7, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — ✅ no blocking findings

GPT 6.1 completed its review of 0122dc6bbe100b7284e2341b640849da1685f237 and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 0122dc6

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 0122dc6bbe100b7284e2341b640849da1685f237 false-positive: <reason, 20 to 500 characters>

@github-actions

github-actions Bot commented Nov 7, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — ✅ no blocking findings

Reviewed 0122dc6bbe100b7284e2341b640849da1685f237 — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 0122dc6

Verdict parsed from the review's SHA-scoped output markers for commit 0122dc6bbe100b7284e2341b640849da1685f237.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 0122dc6bbe100b7284e2341b640849da1685f237 false-positive: <reason, 20 to 500 characters>

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Nov 7, 2026
@buluoray
buluoray force-pushed the crew/slack/0.12.1-batch-45 branch from 9b6303f to 0122dc6 Compare November 7, 2026 22:53
@buluoray buluoray changed the title fix: backport 0.12.1 batch 45 (10 user-reported fixes) fix: backport 0.12.1 batch 45 (9 user-reported fixes) Nov 7, 2026
@buluoray

buluoray commented Nov 7, 2026

Copy link
Copy Markdown
Contributor Author

Goal edited at the maintainer's request: ten fixes -> nine. #32396 is removed from this PR (head 9b6303f -> 0122dc6) because Opus 5.5's finding on dashboard/chat_turn/stream_chunks.py:220 holds: on codex the held reply to a pending steer replays after the tool rows that rendered live meanwhile. main has the same code; #32396 ships later in its own batch with the main fix.

@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention and removed readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Nov 7, 2026
@buluoray
buluoray merged commit 0fb7434 into release/0.12.0 Nov 7, 2026
22 of 35 checks passed
@buluoray
buluoray deleted the crew/slack/0.12.1-batch-45 branch November 7, 2026 23:54
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants