You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
fix(server): overlapping status reads share one untracked listing - #34574
Two follow-ups from CodeRabbit's review of #34460, which merged without them:
Overlapping status reads for one worktree each run their own git ls-files --others. The
backoff from fix(server): branch totals stop relisting untracked files after a failed listing #34460 only applies once a listing has failed, so reads that pass the check while a
slow listing is still running (for example a status read restarted by a file change during a
30 s listing timeout) each walk the whole untracked tree again, holding foreground Git slots.
The branch-totals error keeps only the listing error's text. Its exit code, output length and
underlying failure are dropped, against the rule that a wrapping error keeps the error it wraps
as cause.
Change
The untracked listing goes through an Effect Cache keyed by worktree and pathspecs, with a timeToLive of zero. A read that arrives while the same listing is running waits for it instead
of starting another; the result is never kept, so the next read lists again. The cache keeps a
shared listing running until its last waiter leaves, so one interrupted read does not cancel the
others. Each read still prepares its own temporary review index.
The listing's GitCommandError is kept as cause on the branch-totals error, for a fresh failure
and for the remembered one. "Too many untracked files to count." has no cause.
Scope and approval
Follow-up to #34460 (issue #34457), answering its two open CodeRabbit comments. No setting,
contract or UI change. Submitted under the focused obvious-bug exception in CONTRIBUTING.md; no
maintainer approval is claimed.
Verification
New test share one listing between overlapping status reads (GitVcsDriverCore.test.ts): the
spawner holds the first read's listing until the second read's last Git command before its
listing (show-ref of the branch) has returned, then releases it. Both reads report the same
totals.
On main: fails, 2 listings (3 of 3 runs).
With the change: passes, 1 listing (5 of 5 runs).
The cause is not visible through statusDetailsLocal, which turns a totals failure into missing
totals; it shows in the readBranchChangeTotals span. No separate test.
vp test run src/vcs/GitVcsDriverCore.test.ts: all pass except keeps generated worktree folders short without changing long branch names, which also fails on
unchanged main on this Mac (/private/var versus /var). Server typecheck; vp lint and vp fmt on apps/server/src. macOS only; not run on Linux or Windows.
Written with Claude Opus 5.5 in Claude Code; reviewed with gpt-6.1-sol in Codex.
Macroscope's review found this PR approvable — Overlapping status reads now share the same in-flight untracked-file listing, while later reads refresh normally. The change also preserves underlying error details without changing product defaults or persistent repository data.
You can add or adjust custom eligibility rules. Learn more.
No actionable comments were generated in the recent review.
Walkthrough
Overlapping local status reads now share an in-flight untracked-file listing for the same working directory and path arguments. Listing failures retain their GitCommandError cause through retry state and the resulting error.
Changes: Untracked review totals
Layer / File(s)
Summary
Share listings and preserve failure causes apps/server/src/vcs/GitVcsDriverCore.ts, apps/server/src/vcs/GitVcsDriverCore.test.ts
A zero-TTL cache shares concurrent listings with identical working directories and path arguments. Listing failures retain their cause through retry state and the resulting error. A concurrency test checks that two reads return the same totals after only one listing.
Issue 25558: The change shares overlapping untracked listings, but does not address the reported timeout or repeated recomputation on each poll.
Issue 34457: The change shares overlapping failing listings and preserves the GitCommandError cause.
Possibly related PRs
pingdotgg/t3code#28374: Changes the path set passed to the same branch-totals listing path that this change caches.
Suggested reviewers:macroscopeapp, juliusmarminge
Merge Risk: Minimal · up to 42772
The change deduplicates overlapping untracked-file listings and keeps the underlying error cause on failures. No merge-blocking risk was found; the new concurrency test could occasionally be timing-sensitive.
Pre-merge checks: 4
Check name
Status
Explanation
Title check
The title names the main change: overlapping status reads now share one untracked-file listing. It follows conventional commit form and fits the changeset. It omits the error-cause change, which is a …
Description check
The description follows the template. It has Problem, Change, Scope and approval, and Verification sections. It gives a concrete problem statement with reproduction context (a status read restarted du…
Linked Issues check
Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check
Check skipped because no linked issues were found for this pull request.
Full details: Title check
Explanation
The title names the main change: overlapping status reads now share one untracked-file listing. It follows conventional commit form and fits the changeset. It omits the error-cause change, which is a secondary part of the PR and does not need to be in the title.
Full details: Description check
Explanation
The description follows the template. It has Problem, Change, Scope and approval, and Verification sections. It gives a concrete problem statement with reproduction context (a status read restarted during a 30 s listing timeout). It explains the change, cites the follow-up issue and the approval basis, and reports specific test results on macOS with the unrun platforms stated. The template's optional self-attribution line is included. Minor gap: the Problem section does not give a full reproduction in steps, but the scenario is clear enough for a reviewer.
Finishing Touches
Generate unit tests (beta)
Create a new PR
Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Reviewing files that changed from the base of the PR and between 694076d and 427726b.
Files selected for processing: 2
apps/server/src/vcs/GitVcsDriverCore.test.ts
apps/server/src/vcs/GitVcsDriverCore.ts
Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 9 remain after this review.
This branch has not been deployed
No deployments
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
size:M30-99 changed lines (additions + deletions).vouch:unvouchedPR author is not yet trusted in the VOUCHED list.
1 participant
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.
github-actions
Bot
added
the
vouch:unvouched
![macroscopeapp[bot]](https://avatars.githubusercontent.com/in/900172?s=60&v=4){kind=small}
Problem
Two follow-ups from CodeRabbit's review of #34460, which merged without them:
git ls-files --others. Thebackoff from fix(server): branch totals stop relisting untracked files after a failed listing #34460 only applies once a listing has failed, so reads that pass the check while a
slow listing is still running (for example a status read restarted by a file change during a
30 s listing timeout) each walk the whole untracked tree again, holding foreground Git slots.
underlying failure are dropped, against the rule that a wrapping error keeps the error it wraps
as
cause.Change
Cachekeyed by worktree and pathspecs, with atimeToLiveof zero. A read that arrives while the same listing is running waits for it insteadof starting another; the result is never kept, so the next read lists again. The cache keeps a
shared listing running until its last waiter leaves, so one interrupted read does not cancel the
others. Each read still prepares its own temporary review index.
GitCommandErroris kept ascauseon the branch-totals error, for a fresh failureand for the remembered one. "Too many untracked files to count." has no cause.
Scope and approval
Follow-up to #34460 (issue #34457), answering its two open CodeRabbit comments. No setting,
contract or UI change. Submitted under the focused obvious-bug exception in CONTRIBUTING.md; no
maintainer approval is claimed.
Verification
share one listing between overlapping status reads(GitVcsDriverCore.test.ts): thespawner holds the first read's listing until the second read's last Git command before its
listing (
show-refof the branch) has returned, then releases it. Both reads report the sametotals.
statusDetailsLocal, which turns a totals failure into missingtotals; it shows in the
readBranchChangeTotalsspan. No separate test.vp test run src/vcs/GitVcsDriverCore.test.ts: all pass exceptkeeps generated worktree folders short without changing long branch names, which also fails onunchanged main on this Mac (
/private/varversus/var). Server typecheck;vp lintandvp fmtonapps/server/src. macOS only; not run on Linux or Windows.Written with Claude Opus 5.5 in Claude Code; reviewed with gpt-6.1-sol in Codex.