Repository navigation
Conversation
… work as safe Three problems in worktree-audit.sh (cursor#470): - a PR in any non-open state satisfied `$pr != "-"`, so a CLOSED (unmerged) PR made the worktree `safe`; - the computed `remote` column (ahead/no-remote/detached) was never read by the bucket logic, so unpushed commits could be bucketed `safe`; - the header called the script read-only although it runs `git fetch`. Only ancestry of origin/main or a MERGED PR now counts as proof the work exists elsewhere. Worktrees whose HEAD is ahead of / absent from the remote or detached and not on origin/main go to a new `hold-unpushed` bucket, ranked right after hold-wip and hold-open-pr. A CLOSED PR falls through to `review`. The header now states the one write the script performs, and the playbook documents the new bucket in step 1 (steps 4-5 are left untouched to avoid conflicting with cursor#459). Verified on a synthetic repo with a stubbed gh: ahead1+CLOSED and detached off-main -> hold-unpushed; pushed+MERGED, pushed+CLOSED-on-main and detached on origin/main -> safe; pushed+CLOSED with commits not on main -> review. Fixes cursor#470
Deletion advice must distinguish preserved work from local-only commits. Exercise the audit CLI in an isolated repository with closed PRs, squash merges, a behind branch, and a detached head.
A behind branch has no local-only commits. Deleted origin branches and fork branches can still have their exact head preserved by a PR. Compare that head rather than treating every missing remote as unpushed, while holding commits added after a merge.
…#613 main (cursor#613) normalises the PR list to {number, state, head} with a lowercase state and prefers the origin CLI over gh. The audit now carries head_oid through that list (gh provides it, origin's record leaves it empty) so heliowap's PR-head check still works, and the fixture expectations use the lowercase states the script prints.
(cherry picked from commit a20f1aa)
(cherry picked from commit a17c461)
(cherry picked from commit 2310495)
Adds two worktrees to the audit fixture: one with a single untracked file, one with only ignored build output (.gitignore'd build/ and run.log). With cursor#459's change both report wip:N and land in hold-wip; a guard asserts no row ever reports scratch:. Also restores cursor#459's opening paragraph of the cleanup playbook, which the stack merge had dropped, and regenerates the harness tree.
This was referenced Oct 22, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stacks #492 (ours, #470) and #459 (@dajiaohuang, #449) into one mergeable unit and adds the tests #459 did not have. Close this and merge the originals in order (#492 then #459) if that is preferred; this PR exists because the two change the same two files and neither can be reviewed in isolation any more.
What is in here
worktree-audit.shnever buckets unpushed or closed-unmerged work assafe;hold-unpushedbucket; PR head matching byheadRefOid; a bun test harness (worktree-audit.fixture.ts,worktree-audit.test.ts) that builds a real repo with worktrees and a fakegh.-x, authorship preserved):git status --untracked-files=all --ignored=matching, so untracked and ignored paths count aswip:Nand the worktree lands inhold-wip; thescratch:class is gone; the cleanup playbook stops using--force,rm -rfandpruneto get past Git's refusal and requires explicit approval of every path a removal would delete. Rebased onto fix(pstack): worktree-audit never buckets unpushed or closed-unmerged work as safe #492's playbook step 1 and 3 wording; the only hand-resolved hunk is steps 3 to 5 of the playbook, where fix(pstack): protect untracked and ignored worktree files #459's text for 4 and 5 was taken and fix(pstack): worktree-audit never buckets unpushed or closed-unmerged work as safe #492's step 3 kept..gitignoredbuild/andrun.log). Both must reportwip:Nandhold-wip; a guard asserts no row reportsscratch:. Without fix(pstack): protect untracked and ignored worktree files #459 the untracked one reportsscratch:1/reviewand the ignored one reportsclean/review, which is the loss described in pstack: worktree-cleanup treats untracked and ignored files as disposable (--force, rm -rf) #449.Verification
bun test worktree-audit: 12 pass / 0 fail.generate-pstack-harness-trees.mjs --checkup to date. Branch is on today'smain.