IW-491: phase-pr pushes a commit after reading CI, so its green can describe a stale head #596

Merged

mph merged 23 commits from IW-491 into main 2026-09-21 16:32:13 +00:00

Conversation 0 Commits 23 Files changed 44 +2772 -153

claude commented 2026-09-21 14:52:16 +00:00

Member

Copy link

Summary

The CI gate in phase-pr, phase-merge and ci-gate now judges the commit that is the pull request head.

Defects fixed

Phase 1 — the wait budgets were measured from a head that was gone (false red). The card has a new root field ci_wait_sha, the head that a wait measures. CiGate.poll seeds the wait from it. When the pull request head is not the recorded head, both budgets (run appearance and --timeout) start again. ci_wait_started and ci_wait_sha are written and cleared together. Before this fix, a resumed phase-pr, a phase-merge recovery retry or a ci-gate after a manual push could read "CI did not run" and stop a batch with no retry. (PR #590)

Phase 2 — the Green path committed and pushed the card after the verdict (false green). After a green verdict, phase-pr writes the card to disk. It does not commit or push it. ci_gated_sha is now the pull request head when the command exits. (PR #591)

Testing

• Harness tests in CiGateHarnessTest, PhasePrHarnessTest, PhaseMergeHarnessTest and CiGateCommandHarnessTest reproduce each defect: a recovery push late in a wait, a resumed wait 61 minutes after the stamp (the IW-583 gap), and no commit or push after green.
• Unit tests in CiGateTest, ReviewStateUpdaterTest, ReviewStateValidatorTest and ReviewStateSchemaTest cover the new card field.
• Each phase PR passed CI. The pre-push hook ran the unit suite on this branch.

Follow-ups filed

iw-cli #592, #593, #594, #595; dev-docs #317 (the review-state-schema skill entry for ci_wait_sha). Comments added to #585 and #586.

Release notes

See project-management/issues/IW-491/release-notes.md (Czech).

🤖 Generated with Claude Code

Closes #491

## Summary The CI gate in `phase-pr`, `phase-merge` and `ci-gate` now judges the commit that is the pull request head. ## Defects fixed **Phase 1 — the wait budgets were measured from a head that was gone (false red).** The card has a new root field `ci_wait_sha`, the head that a wait measures. `CiGate.poll` seeds the wait from it. When the pull request head is not the recorded head, both budgets (run appearance and `--timeout`) start again. `ci_wait_started` and `ci_wait_sha` are written and cleared together. Before this fix, a resumed `phase-pr`, a `phase-merge` recovery retry or a `ci-gate` after a manual push could read "CI did not run" and stop a batch with no retry. (PR #590) **Phase 2 — the Green path committed and pushed the card after the verdict (false green).** After a green verdict, `phase-pr` writes the card to disk. It does not commit or push it. `ci_gated_sha` is now the pull request head when the command exits. (PR #591) ## Testing - Harness tests in `CiGateHarnessTest`, `PhasePrHarnessTest`, `PhaseMergeHarnessTest` and `CiGateCommandHarnessTest` reproduce each defect: a recovery push late in a wait, a resumed wait 61 minutes after the stamp (the IW-583 gap), and no commit or push after green. - Unit tests in `CiGateTest`, `ReviewStateUpdaterTest`, `ReviewStateValidatorTest` and `ReviewStateSchemaTest` cover the new card field. - Each phase PR passed CI. The pre-push hook ran the unit suite on this branch. ## Follow-ups filed iw-cli #592, #593, #594, #595; dev-docs #317 (the `review-state-schema` skill entry for `ci_wait_sha`). Comments added to #585 and #586. ## Release notes See `project-management/issues/IW-491/release-notes.md` (Czech). 🤖 Generated with [Claude Code](https://claude.com/claude-code) Closes #491

claude added 22 commits 2026-09-21 14:52:17 +00:00

workflow(dx): Create diagnostic analysis for IW-491 ... 56475c0557

phase-pr pushes a commit after reading CI, so its green can describe a stale head

Artifacts:
- analysis.md

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

workflow(dx): drop the analysis so triage re-runs with both defects ... 7c1051cff4

The first analysis (56475c05) covered defect 1 only. Defect 2 — the
appearance budget measured from a ci_wait_started stamp an earlier head
left behind — lived in an issue comment, and triage reads the issue body.

The body now carries defect 2 and Michal's F1 decision on the CLARIFY, and
the issue points the next analysis at 56475c05 so it reuses the work that
ruled out the format flake.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0179JhvXg4eopsQNmDLUSuim

workflow(dx): Create diagnostic analysis for IW-491 ... 8f2ccc3ea5

phase-pr pushes a commit after reading CI, so its green can describe a stale head

Covers both defects: the Green path pushes the card after the verdict
(fix decided as F1), and the appearance budget and timeout are measured
from a ci_wait_started stamp that an earlier head left behind.

Artifacts:
- analysis.md

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

workflow(dx): record the timeout decision for IW-491 ... b0a8ec7f84

A wait belongs to one head. When the head differs from the one the card
records, the appearance budget and the total --timeout both restart and the
stamp is rewritten, inside one invocation as well as across invocations.

An appearance-only fix would leave the observed failure under a different
word: on IW-583 the stamp was 61 minutes old against a 30-minute timeout, so
the gate would answer "CI timed out" in place of "CI did not run", write
ci_failed either way, and halt the batch either way.

Cost: --timeout means per head, not per wait. e935bd21, the ci_wait_started
schema description and the CiGateTarget doc state the old meaning and change
with the fix. phase-merge's total wall clock becomes bounded by
(--max-retries + 1) x --timeout.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0179JhvXg4eopsQNmDLUSuim

chore(IW-491): update review-state to tasks_ready 866f4e4db9

workflow(dx): Create task index for IW-491 ... c92452c041

phase-pr pushes a commit after reading CI, so its green can describe a stale head

Artifacts:
- tasks.md
- review-state.json (tasks_ready)

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Phase 1: context ready ... e209fb3925

Issue: IW-491
Phase: 1 - The appearance budget and the timeout are measured from a head that is gone

This commit marks the checkpoint for phase 1 context generation.
Use dx-fix IW-491 to return to this point.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

chore(IW-491): update review-state to context_ready c693cb8132

workflow(dx): Generate Phase 1 tasks for IW-491 ... 90b5a41ab4

The appearance budget and the timeout are measured from a head that is gone

Artifacts:
- phase-01-tasks.md

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

chore(IW-491): update review-state to tasks_ready c399cfa7b7

Phase 1: The appearance budget and the timeout are measured from a head that is gone (#590) 464cbc5a26

chore(IW-491): update review-state for phase 01 8b8e34a3f1

chore(IW-491): mark phase 01 complete in tasks.md 54ffed1b0e

Phase 2: context ready ... 4141ca8c18

Issue: IW-491
Phase: 2 - The Green path commits and pushes the card after the verdict

This commit marks the checkpoint for phase 2 context generation.
Use dx-fix IW-491 to return to this point.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

chore(IW-491): update review-state to context_ready ff4ca9bee3

workflow(dx): Generate Phase 2 tasks for IW-491 ... d9e20f5a56

The Green path commits and pushes the card after the verdict

Artifacts:
- phase-02-tasks.md

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

chore(IW-491): update review-state to tasks_ready 11d4d0e8d6

Phase 2: The Green path commits and pushes the card after the verdict (#591) 89d9a61f2a

chore(IW-491): update review-state for phase 02 e18203eb43

chore(IW-491): mark phase 02 complete in tasks.md 335e1e9824

chore(IW-491): file follow-ups on the tracker ... e78cbba4e3

Filed as iw-cli #592-#595 and dev-docs #317; commented on #585 and #586.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

docs(IW-491): add release notes ... d0b06a05a3

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

mph added 1 commit 2026-09-21 14:53:27 +00:00

chore(IW-491): update review-state to all_complete 4b41aaf6b4

mph merged commit d58f788083 into main 2026-09-21 16:32:13 +00:00

mph referenced this pull request from a commit 2026-09-21 16:32:14 +00:00

Merge pull request 'IW-491: phase-pr pushes a commit after reading CI, so its green can describe a stale head' (#596) from IW-491 into main

Sign in to join this conversation.

Reviewers

No reviewers

Labels

Clear labels   

agent-ready

  

bug

Something isn't working

  

contract

Run the tool-contract suite on this PR

  

documentation

Improvements or additions to documentation

  

duplicate

This issue or pull request already exists

  

enhancement

New feature or request

  

good first issue

Good for newcomers

  

help wanted

Extra attention is needed

  

invalid

This doesn't seem right

  

needs-human

  

needs-input

Intake asked the human a question on the issue; waiting for the answer

  

question

Further information is requested

  

wontfix

This will not be worked on

No labels

agent-ready

bug

contract

documentation

duplicate

enhancement

good first issue

help wanted

invalid

needs-human

needs-input

question

wontfix

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

2 participants

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

iterative-works/iw-cli!596

No description provided.