Submodule initialization inherits Git's 30-second total deadline. Slow clones are interrupted mid-checkout, but worktree creation still succeeds. The reported Windows 0.1.5 trace confirms a timeout after 30.9 seconds.
Set timeoutMs: null only for the submodule update, matching the existing cancellable git worktree add path. Cancellation and best-effort failure handling remain unchanged; the warning now describes submodules as potentially incomplete.
Follows #26409 to fix the 30-second submodule-update timeout reported in #26390. #7674 introduced automatic initialization; #23802 later removed the worktree-add deadline. This applies the same policy to submodule checkout.
vp run check:changed: all static checks pass; 204 tests pass, three skipped, ten failures. All ten also reproduce with unmodified upstream code at 8b74a69019 on this Windows/Git 2.32 host.Before — 31.2 s timeout (6575ce3250) | After — 40.9 s complete (6f8633f5de) |
|---|---|
sub03/nested has no files despite clean parent status | All six submodules populated at recorded gitlinks; clean status |
Reproduction fixture and transcripts.
Model and harness: GPT-6.1 Sol via Codex in T3 Code.
Added 1 commit
Verdict: Needs Human Approval
Macroscope's review found this PR requires human approval — Worktree submodule checkouts now wait without the previous 30-second deadline, preserving progress and failure reporting. This changes the default waiting policy: slow clones can finish, while stalled clones require cancellation.
You can add or adjust custom eligibility rules. Learn more.
Submodule updates during worktree creation now run without a timeout. Tests cover delayed recursive and top-level updates, optional progress callbacks, cancellation, and failed updates.
Changes: Submodule worktree updates
| Layer / File(s) | Summary |
|---|---|
Submodule update timeout and worktree validation apps/server/src/vcs/GitVcsDriverCore.ts, apps/server/src/vcs/GitVcsDriverCore.test.ts | Submodule updates now set timeoutMs to null, and the failure warning says submodules may be incomplete. Tests cover updates delayed beyond six simulated minutes, recursive and top-level modes, optional progress reporting, cancellation, and failure reporting while worktree creation succeeds. |
Estimated code review effort: 2 (Simple) | ~12 minutes
Change: Bug fix
Possibly related issues
pingdotgg/t3code#26390: The issue describes the submodule update timeout during worktree creation; this change removes that timeout.Suggested labels: bugfix, size:M
Suggested reviewers: juliusmarminge, macroscopeapp
f8791Slow submodule checkouts can now finish, but a checkout that never exits requires cancellation. This is a bounded concern for the owner rather than an established reason to block merging.
| Check name | Status | Explanation |
|---|---|---|
| Title check | Passed | The title "fix(server): let slow worktree submodule checkouts finish" matches the main change. The PR removes the 30-second Git deadline for the submodule update during worktree creation, so slow chec… |
| Description check | Passed | The description follows the template's main sections: Problem, Change, Scope and approval, Verification, and a model/harness note. The Problem section gives the cause (Git's 30-second total deadline) … |
| Linked Issues check | Passed | Check skipped because no linked issues were found for this pull request. |
| Out of Scope Changes check | Passed | Check skipped because no linked issues were found for this pull request. |
Explanation
The title "fix(server): let slow worktree submodule checkouts finish" matches the main change. The PR removes the 30-second Git deadline for the submodule update during worktree creation, so slow checkouts can complete. It follows the conventional commit format and names the affected component (server) and the behavior being fixed.
Explanation
The description follows the template's main sections: Problem, Change, Scope and approval, Verification, and a model/harness note. The Problem section gives the cause (Git's 30-second total deadline) and the observed trace (30.9 seconds). The Change section explains setting timeoutMs to null for the submodule update and the updated warning. Scope links related issue and PRs (#26390, #26409, #7674, #23802). Verification reports focused tests, regression results, a real-client check, the known failing tests on upstream, and before/after screenshots. The model and harness are named. Minor gaps: the scope section cites related issues but does not clearly state a maintainer approval comment, and the verification section includes an unverified claim that the 10 failures reproduce on upstream without a linked log. Because the template asks for explicit approval evidence in Scope, the author should link the maintainer approval or explain why this small, obvious fix qualifies without it.
apps/server/src/vcs/GitVcsDriverCore.ts-4200-4200: 1
4200-4200: External image Major · Stability & Availability · Heavy liftExternal image Unverified finding Verification ran but could not confirm this finding. It is shown for review, not as a verified issue.
Add a safety limit for a stalled submodule update.
If a submodule remote stops responding without exiting,
timeoutMs: nullleaves the command running indefinitely.createWorktreethen remains pending after it claims the worktree, and the best-effort failure handler cannot report incomplete submodules. Use a generous stall or safety limit that preserves slow checkouts but lets a stuck command reach the existing failure handler. (raw.githubusercontent.com)
eb9f7566-251d-4c77-913e-c58ff73b07e9Reviewing files that changed from the base of the PR and between 8b74a69019e3d0f5be62dc118c526669002fc195 and f87917af447672802eb33060c45e456615dc8e89.
apps/server/src/vcs/GitVcsDriverCore.test.tsapps/server/src/vcs/GitVcsDriverCore.tsIncluded review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 9 remain after this review.
Note
Written by gpt-6.1-sol on behalf of Maria
verified unchanged head f87917af447672802eb33060c45e456615dc8e89 on this linux vm, after reproducing the bug on github/main (7195b69954) in a separate worktree. each checkout ran vp i and its own isolated dev database, with only one dev server running at a time.
used playwright from the repo with /usr/bin/google-chrome. through the web ui i added a real git project, selected codex gpt-6.1-sol, selected new worktree, and submitted a read-only verification turn. the submodule came from a real rate-limited http git remote carrying a 4 mib payload. git and codex were not mocked.
vp run check:changed 8b74a69019e3d0f5be62dc118c526669002fc195 passed through blacksmith-check: format, lint, both knip checks, graphql, three package typechecks, and 217 tests across two files. the pr currently has 31 successful checks. no source changes or pushes were needed.
before, main:
main: real codex turn reports the missing submodule files
after, pr:
pr: real codex turn confirms populated submodule and clean git state
recording: setup remains active past the former deadline, then hands off to the real provider and shows its result:
slow checkout passes 30 seconds and completes before the provider reads the files
cancellation stops setup before the provider starts
short server and git command transcript
native desktop, ios/android, windows/macOS, authenticated external submodule hosts, nested submodules, and runtime top-level/disabled settings were not exercised on this linux web-client pass. an indefinitely silent remote was not left running; explicit cancellation during a real slow clone was verified instead.
