P-61 also reads the linked pull request's own close, reopening and merge, so its run grounds #672
Merged
DO
7d ago+56+56+01234567890123456789-29-29-01234567890123456789
Part of #551 and #672. Asked by the hooks lane: the PR for #672 gives each linked pull request close_events, and its rule a_pull_requests_close_and_reopening_are_timed waits on P-61. So P-61's run must ask GitHub about a pull request's own close and reopening too.
What changes
tools/trial_p61_timeline.sh, which the Maintainer has not run yet:
- Two new steps, before the merge: step 8 closes the pull request unmerged, and step 9 reopens it. There are now eleven steps; the merge is step 10.
- The query also reads the pull request's own timeline for
ClosedEvent,ReopenedEventandMergedEvent, each with its time and actor, along with itsclosedAt. - A read is whole only where both timelines, the issue's and the pull request's, sit on their one page.
- The pass condition, written before the run, adds that each of steps 8 to 10 appears on the pull request's timeline as its own timed event.
tests/test_tools.py::TimelineTrial:
- The query, word for word, and the counts for eleven steps.
- The stand-in gh answers the pull request's timeline.
- A new mode: the pull request's timeline on more than one page is said, and offered again.
Four new mutants, all killed:
- "P-61 asks no close or reopening of the pull request"
- "P-61 reads the pull request's timeline in part"
- "P-61 asks the Maintainer to close nothing"
- "P-61 asks the reopening before the close"
No record and no protected path. Design §6's P-61 row now says ten hand steps, the close and reopening among them. The hooks lane's PR for #672 edits the same row, so whichever merges second resolves it.
Verify
make checks: ok. tests.test_tools.TimelineTrial: OK. Full verify is CI's.
Reviews
| Read | State |
|---|---|
| Claude + security | Nothing blocking. It checked the shell quoting, the query against GitHub's schema, and each mutant by hand. Its two wording points on the header, that both timelines are read whole and that the run records whether a close stands beside the merge, are fixed in d7607b2 |
| Argon | clean |
| Muse | Two minor points: the step order not held by the test, and design §6's row still saying eight steps. Both fixed in 05ec8df |
| Astra | First read (d7607b2): one major, the same step order. Fixed in 05ec8df; re-read at 05ec8df clean |
🤖 Generated with Claude Code