Repository navigation
Conversation
IfkumRfnl
marked this pull request as ready for review
October 15, 2026 08:06
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
Member
|
Taking over this one for batch worktree-git-05. I’ll keep the status behavior covered, verify the subprocess and latency reduction against current main, and publish here without merging. |
This branch has not been deployed
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Status could launch a default-branch Git subprocess even after its configured comparison base resolved, and remote status with an upstream could repeat a successful origin-default read. Candidate lookups now run only when reached, and remote status with an upstream passes its successful origin result into its later comparison. Failed reads still retry; other primary remotes resolve their own default. Candidate order and returned status for the same repository state stay the same, with no persistent cache.
Verified at 26aa1b72c1, integrated with main 687bd55533, on Linux / Node 24.21.0 / Git 2.43.0:
Live server-level proof: invoked the actual base and head status-driver factories against the same disposable Git repository. All 48 alternating pairs returned identical status. Eight pairs per scenario/driver lifetime, warm filesystem cache, 100 base files and 20 modified files. Local status includes Changes totals; remote status disables fetch. Driver construction excluded. Timings describe this shared host; the eliminated Git call is deterministic. The unchanged-command missing-base control shows the timing noise floor, so local timings alone do not establish a speedup.
Raw measurements · Measurement script · Method and paired timings
Remote status without an upstream retains its existing lookup order and duplicate reads. Moving the default read earlier changes the result when its first lookup fails and the next succeeds; that behavior choice is outside this optimization.
No UI behavior changes, so screenshots do not demonstrate this optimization. Clients share the server driver. macOS, Windows, iOS, Android, Electron desktop, remote/relay/tunnel connections, and network latency were not exercised in this pass. Earlier Windows timings were removed because they measured an older revision. Current-head CI passes, including all eight server shards, lint, build/typecheck and the aggregate check. The canceled duplicate fingerprint job was superseded by a passing run on the same head. There are no unresolved review threads. No formal GitHub approval is recorded; both independent review agents found no blockers. Unmerged; auto-merge is disabled.
Original contribution by @IfkumRfnl in #19825. Maintainer takeover: GPT-6 Astra (tests), GPT-6.1 Sol (implementation/integration), Claude Opus 5.5 and GPT-6 Sol (adversarial review), through the Codex and Claude harnesses in T3 Code.