IW-566: phase-pr --no-wait — open the pull request, leave the CI wait to the caller #571

Merged

mph merged 14 commits from IW-566 into main 2026-09-16 09:32:26 +00:00

Conversation 0 Commits 14 Files changed 11 +1925 -18

claude commented 2026-09-16 09:04:01 +00:00

Member

Copy link

Summary

iw phase-pr can now open a pull request and stop, leaving the CI wait to the caller. The switch
is --no-wait.

The wait was previously inseparable from the command: --batch chose what happens after green,
and the timeout flags only sized the wait. That is right when a human is driving. It is wrong for
an unattended phase, where the model's own wall-clock budget is spent inside sleep 30 while the
orchestrator — which already owns a CI wait it never reaches — sits idle. Measured on IW-555:
17.3 minutes of a 90-minute session, and the session was killed with the phase unfinished.

No waiting logic was added. iw ci-gate is already a first-class wait with the same three exit
codes and the same three flags; what this removes is an assumption about who runs it.

Changes

• core/commands/PhasePr.scala
  • --no-wait switch in the declaration, carried through PhasePrOptions and Resolved.
  • --batch --no-wait refused in run, naming both flags, exit 1, before any push or pull
    request — the two do not compose, since --batch merges on the green it was told not to wait
    for.
  • handOffWait: the terminal path. Card at ci_pending with ci_wait_started cleared,
    then commit, print the same JSON, exit 0.
  • reportOpenPullRequest: the commit-and-report tail, previously stated twice.
  • Help text: the flag's own does, the exit 0/1/3 wording, and an example of the unattended pair.
• core/test/PhasePrHarnessTest.scala — 7 tests, one per acceptance criterion (111 added,
  0 removed).

Behaviour that did not change

Without the flag, phase-pr behaves exactly as before. The 53 pre-existing harness tests pass
unmodified — the test file's diff is 111 added / 0 removed, which was the acceptance criterion
for this. When CI is not expected, --no-wait still hands over to awaiting_review, since a
ci_pending card would announce a wait nothing will answer.

Testing

• Core suite 186/186, dashboard 241/241, dashboard itest 240/240.
• No integration test added: --no-wait adds no process boundary, so test/phase-pr.bats stays
  the one-round-trip smoke test.
• Code review: 1 iteration, 6 skills. 1 critical (a scalafmt violation) and 2 warnings, all fixed.
  No reviewer found a correctness defect.

Known red, not caused by this branch: ./iw ./test does not exit 0 here and did not before
either — test/start.bats "the tmux session does not carry iw's internal plumbing" fails
identically on a clean worktree at f3a938bc, which contains none of this code. Tracked as #556.

This does not yet deliver the measured saving

The flag alone moves nothing. The wf-implementation-workflow skill still runs phase-merge
in-session, and phase-merge runs its own CI gate — so the sleep would simply move from one
command to the other. That companion change lives in the kanon plugin repository. Do not report
the 17.3-minute saving as delivered until it ships.

Follow-ups filed

#568 (duplicated ci_pending card shape), #569 (batch/noWait as two guarded Booleans),
#570 (lint gate runs scalafix but not scalafmt).

Release notes

project-management/issues/IW-566/release-notes.md (Czech).

Review

Start at project-management/issues/IW-566/review-packet.md — entry points, scenarios and the
test summary.

Closes #566

🤖 Generated with Claude Code

## Summary `iw phase-pr` can now open a pull request and stop, leaving the CI wait to the caller. The switch is `--no-wait`. The wait was previously inseparable from the command: `--batch` chose what happens *after* green, and the timeout flags only sized the wait. That is right when a human is driving. It is wrong for an unattended phase, where the model's own wall-clock budget is spent inside `sleep 30` while the orchestrator — which already owns a CI wait it never reaches — sits idle. Measured on IW-555: 17.3 minutes of a 90-minute session, and the session was killed with the phase unfinished. No waiting logic was added. `iw ci-gate` is already a first-class wait with the same three exit codes and the same three flags; what this removes is an assumption about who runs it. ## Changes - `core/commands/PhasePr.scala` - `--no-wait` switch in the declaration, carried through `PhasePrOptions` and `Resolved`. - `--batch --no-wait` refused in `run`, naming both flags, exit 1, before any push or pull request — the two do not compose, since `--batch` merges on the green it was told not to wait for. - `handOffWait`: the terminal path. Card at `ci_pending` with `ci_wait_started` **cleared**, then commit, print the same JSON, exit 0. - `reportOpenPullRequest`: the commit-and-report tail, previously stated twice. - Help text: the flag's own `does`, the exit 0/1/3 wording, and an example of the unattended pair. - `core/test/PhasePrHarnessTest.scala` — 7 tests, one per acceptance criterion (111 added, 0 removed). ## Behaviour that did not change Without the flag, `phase-pr` behaves exactly as before. The 53 pre-existing harness tests pass **unmodified** — the test file's diff is 111 added / 0 removed, which was the acceptance criterion for this. When CI is not expected, `--no-wait` still hands over to `awaiting_review`, since a `ci_pending` card would announce a wait nothing will answer. ## Testing - Core suite 186/186, dashboard 241/241, dashboard itest 240/240. - No integration test added: `--no-wait` adds no process boundary, so `test/phase-pr.bats` stays the one-round-trip smoke test. - Code review: 1 iteration, 6 skills. 1 critical (a scalafmt violation) and 2 warnings, all fixed. No reviewer found a correctness defect. **Known red, not caused by this branch:** `./iw ./test` does not exit 0 here and did not before either — `test/start.bats` "the tmux session does not carry iw's internal plumbing" fails identically on a clean worktree at `f3a938bc`, which contains none of this code. Tracked as #556. ## This does not yet deliver the measured saving The flag alone moves nothing. The `wf-implementation-workflow` skill still runs `phase-merge` in-session, and `phase-merge` runs its own CI gate — so the sleep would simply move from one command to the other. That companion change lives in the kanon plugin repository. Do not report the 17.3-minute saving as delivered until it ships. ## Follow-ups filed #568 (duplicated `ci_pending` card shape), #569 (`batch`/`noWait` as two guarded Booleans), #570 (lint gate runs scalafix but not scalafmt). ## Release notes `project-management/issues/IW-566/release-notes.md` (Czech). ## Review Start at `project-management/issues/IW-566/review-packet.md` — entry points, scenarios and the test summary. Closes #566 🤖 Generated with [Claude Code](https://claude.com/claude-code)

claude added 13 commits 2026-09-16 09:04:01 +00:00

workflow(wf): Create analysis for IW-566 ... e403f15350

phase-pr cannot open a pull request without also waiting for CI, so an
unattended phase spends its own budget on sleep

Artifacts:
- analysis.md

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

workflow(wf): Create task index for IW-566 ... 6e0e9aa8c8

phase-pr cannot open a pull request without also waiting for CI, so an
unattended phase spends its own budget on sleep

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

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Phase 1: context ready ... 012bf62024

Issue: IW-566
Phase: 1 - `--no-wait` end to end — options, conflict guard, card state, help

This commit marks the checkpoint for phase 1 context generation.
Use wf-implement IW-566 to return to this point.

🤖 Generated with Claude Code

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

chore(IW-566): update review-state to context_ready 715fc01321

workflow(wf): Generate Phase 1 tasks for IW-566 ... b5fd8f5f5d

`--no-wait` end to end — options, conflict guard, card state, help

Artifacts:
- phase-01-tasks.md

🤖 Generated with Claude Code

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

chore(IW-566): update review-state to tasks_ready 2f3d4520b2

chore(IW-566): update review-state for phase 01 f3a938bc1c

--no-wait end to end — options, conflict guard, card state, help ... c6240a2d8b

- Add a --no-wait switch that opens the pull request and hands the CI wait to a later `iw ci-gate`
- Refuse --batch with --no-wait in run, naming both flags, before any push or pull request
- Leave the card at ci_pending with ci_wait_started cleared, so the next gate does not inherit a previous session's clock
- Keep awaiting_review when the branch carries no workflow file, since there is no wait CI will answer
- Extract reportOpenPullRequest, the commit-and-report tail handOver and handOffWait shared
- Add 7 harness tests; the 53 pre-existing tests stay green unmodified

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

chore(IW-566): update review-state for phase 01 b3b331f227

Merge pull request 'Phase 1: --no-wait end to end — options, conflict guard, card state, help' (#567) from IW-566-phase-01 into IW-566 ... 31383a86f1

Reviewed-on: #567

chore(IW-566): update review-state for phase 01 4205445259

chore(IW-566): mark phase 01 complete in tasks.md 27fa9c399c

workflow(IW-566): review packet, release notes, follow-ups filed ... a526b172e7

Follow-ups filed to the tracker and FOLLOW-UPS.md removed:
- start.bats baseline failure -> comment on #556
- ci_pending card shape duplicated (and handOffWait naming) -> #568
- batch/noWait as two guarded Booleans -> #569
- lint gate runs scalafix but not scalafmt -> #570

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

mph added 1 commit 2026-09-16 09:05:09 +00:00

chore(IW-566): update review-state to all_complete 0c479878a5

mph merged commit 372b27120a into main 2026-09-16 09:32:26 +00:00

mph referenced this pull request from a commit 2026-09-16 09:32:27 +00:00

Merge pull request 'IW-566: phase-pr --no-wait — open the pull request, leave the CI wait to the caller' (#571) from IW-566 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!571

No description provided.