Fix git diff chip refresh flicker with untracked files by solomonneas · Pull Request #9302 · warpdotdev/warp

solomonneas · 2026-04-28T20:24:15Z

Description

The git diff context chip can be updated from GitRepoStatusModel, which includes untracked files, but prompt refreshes could still fall back to the shell generator. That shell fallback runs git diff --shortstat HEAD, which only reports tracked-file changes. When a repo had untracked files, the chip could briefly flip between the structured repo-status count and the tracked-only shell count.

This change keeps externally driven git chips on repo-status metadata when available, clears stale git chip timers when repo status attaches, and still allows initial-value generators for externally driven chips that need them.

Testing

cargo fmt --check
git diff --check
./script/run-clang-format.py -r --extensions 'c,h,cpp,m' ./crates/warpui/src/ ./app/src/
find . -name '*.wgsl' -exec wgslfmt --check {} +
cargo clippy --workspace --exclude warp_completer --all-targets --tests -- -D warnings
cargo clippy -p warp_completer --all-targets --tests -- -D warnings
cargo test -p warp context_chips::current_prompt::tests::test_externally_driven_git_diff_stats_uses_repo_status_without_shell_fallback --features local_fs
cargo test -p warp context_chips::current_prompt::tests::test_git_diff_stats_clears_shell_timer_when_repo_status_attaches_late --features local_fs
cargo test -p warp context_chips::current_prompt::tests::test_git_status_change_updates_chip_value --features local_fs
cargo test -p warp context_chips::current_prompt::tests::test_externally_driven_chip_skips_periodic_timer --features local_fs
cargo nextest run -p warp_completer --features v2
cargo test --doc

Attempted full workspace nextest in a clean headless LXC with CARGO_BUILD_JOBS=1 cargo nextest run --no-fail-fast --workspace --exclude command-signatures-v2; it compiled and ran but failed environment-dependent UI/font tests unrelated to this change.

Server API dependencies

None.

Agent Mode

Warp Agent Mode - This PR was created via Warp's AI Agent Mode

Changelog Entries for Stable

CHANGELOG-BUG-FIX: Fixed the git diff prompt chip flickering between tracked-only and all-file counts when untracked files are present.

cla-bot · 2026-04-28T20:24:18Z

Thank you for your pull request and welcome to our community. We require contributors to sign our Contributor License Agreement, and we don't seem to have the users @solomonneas on file. In order for us to review and merge your code, each contributor must visit https://cla.warp.dev to read and agree to our CLA. Once you have done so, please comment @cla-bot check to trigger another check.

solomonneas · 2026-04-28T20:27:50Z

@cla-bot check

cla-bot · 2026-04-28T20:27:55Z

The cla-bot has been summoned, and re-checked this pull request!

captainsafia · 2026-04-28T22:56:26Z

@solomonneas Once you feel this PR is ready for review, you're welcome to take it out of draft and we'll do a review.

solomonneas · 2026-04-28T23:01:29Z

My bad, I didn’t realize I’d left this as a draft. It should be good to go now.

oz-for-oss · 2026-04-28T23:03:02Z

@solomonneas

I'm starting a first review of this pull request.

I requested changes on this pull request and posted feedback.

Comment /oz-review on this pull request to retrigger a review (up to 3 times on the same pull request).

I requested changes on this pull request and posted feedback.

Comment /oz-review on this pull request to retrigger a review (up to 3 times on the same pull request).

I'm re-reviewing this pull request in response to a review request.

You can view the conversation on Warp.

I completed the review and posted feedback on this pull request.

Comment /oz-review on this pull request to retrigger a review (up to 3 times on the same pull request).

Powered by Oz

oz-for-oss

Overview

This PR changes the current prompt git chip refresh path so externally driven git diff stats use GitRepoStatusModel metadata instead of falling back to shell git diff output, and adds tests for avoiding the tracked-only shell fallback and clearing stale timers.

Concerns

The new repo-status update path can reuse metadata from the previous repository during navigation before async repo detection replaces git_repo_status, which can leave GitDiffStats showing stale counts for the wrong repo.

Verdict

Found: 0 critical, 1 important, 0 suggestions

Request changes

Comment /oz-review on this pull request to retrigger a review (up to 3 times on the same pull request).

Powered by Oz

oz-for-oss · 2026-04-28T23:06:22Z

+        chip_kind: &ContextChipKind,
+        ctx: &AppContext,
+    ) {
+        if let Some(metadata) = self.latest_git_status_metadata(ctx) {


⚠️ [IMPORTANT] This can read metadata from a stale git_repo_status handle: BlockMetadataReceived updates the prompt context before async repo detection swaps the repo-status model, so navigating between repos can populate GitDiffStats from the previous repo until the new metadata arrives. Guard that the status model’s repo_path() still matches the active working directory before applying its metadata, or clear the chip while waiting.

solomonneas · 2026-04-29T02:59:36Z

/oz-review

oz-for-oss

Overview

This PR moves externally driven git prompt chips toward using GitRepoStatusModel metadata and cancels stale shell refreshes to avoid the tracked-only diff fallback flicker.

Concerns

The new repo-status freshness check compares the shell-reported cwd directly against the native/canonical repo path. On Windows Git Bash/MSYS-style sessions, those path formats can differ for the same directory, causing valid metadata to be treated as stale and the git chips to be cleared.
Supplemental security pass found no security-specific issues in the changed code.

Verdict

Found: 0 critical, 1 important, 0 suggestions

Request changes

Comment /oz-review on this pull request to retrigger a review (up to 3 times on the same pull request).

Powered by Oz

oz-for-oss · 2026-04-29T03:03:41Z

+                let status_model = h.as_ref(ctx);
+                let active_working_directory = std::path::Path::new(active_working_directory);
+
+                if !active_working_directory.starts_with(status_model.repo_path()) {


⚠️ [IMPORTANT] active_working_directory is the shell-native cwd, while repo_path() is the canonical/native repo path. For Windows Git Bash/MSYS2 this can be /c/Users/... vs C:\\Users\\..., so this guard treats valid repo metadata as stale and clears the git chips. Convert through the active session's launch data before starts_with, like terminal repo detection does.

The freshness check in latest_git_status_metadata compared the raw shell-reported cwd against the OS-native repo_path. On MSYS2 / Git Bash the shell reports paths like /c/Users/foo/proj while DetectedRepositories stores C:\Users\foo\proj, so starts_with fails and the git chips get cleared even though they are valid for the current repo. Run the cwd through ShellLaunchData::maybe_convert_absolute_path first (handles MSYS2 and WSL via existing helpers, identity for native shells) and fall back to the raw string when no conversion is available. Extract working_directory_belongs_to_repo so the resolution logic is covered by direct unit tests, including the MSYS2-style regression where the converted path is the only one that matches repo_path.

solomonneas · 2026-04-29T14:06:07Z

Pushed 56137b8 — the cwd is now run through ShellLaunchData::maybe_convert_absolute_path before the starts_with(repo_path) ancestor check, with a raw-string fallback for native shells. Extracted the resolution into working_directory_belongs_to_repo so a direct unit test covers the MSYS2 case (where the converted path is the only one that matches repo_path).

/oz-review

oz-for-oss

Overview

This PR updates the current prompt git chips to use GitRepoStatusModel metadata when an externally driven repo-status model is attached, clears stale shell refresh timers, and adds regression coverage for repo-status-driven diff stats and working-directory mismatch handling.

Concerns

No blocking correctness or security concerns found in the changed diff lines.

Verdict

Found: 0 critical, 0 important, 0 suggestions

Approve

Comment /oz-review on this pull request to retrigger a review (up to 3 times on the same pull request).

Powered by Oz

…epo-status # Conflicts: # app/src/context_chips/current_prompt.rs

fix: keep git diff chip on repo status metadata

3eba138

cla-bot Bot added the cla-signed label Apr 28, 2026

solomonneas marked this pull request as ready for review April 28, 2026 23:01

oz-for-oss Bot requested changes Apr 28, 2026

View reviewed changes

fix: guard git prompt chips against stale repo status

1a265d5

oz-for-oss Bot requested changes Apr 29, 2026

View reviewed changes

oz-for-oss Bot reviewed Apr 29, 2026

View reviewed changes

captainsafia added the external-contributor Indicates that a PR has been opened by someone outside the Warp team. label Apr 30, 2026 — with Warp Dev Github Integration

Merge remote-tracking branch 'origin/master' into fix/git-diff-chip-r…

a5173bc

…epo-status # Conflicts: # app/src/context_chips/current_prompt.rs

Conversation

solomonneas commented Apr 28, 2026

Description

Testing

Server API dependencies

Agent Mode

Changelog Entries for Stable

Uh oh!

cla-bot Bot commented Apr 28, 2026

Uh oh!

solomonneas commented Apr 28, 2026

Uh oh!

cla-bot Bot commented Apr 28, 2026

Uh oh!

captainsafia commented Apr 28, 2026

Uh oh!

solomonneas commented Apr 28, 2026

Uh oh!

oz-for-oss Bot commented Apr 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

oz-for-oss Bot left a comment

Choose a reason for hiding this comment

Overview

Concerns

Verdict

Uh oh!

oz-for-oss Bot Apr 28, 2026

Choose a reason for hiding this comment

Uh oh!

solomonneas commented Apr 29, 2026

Uh oh!

oz-for-oss Bot left a comment

Choose a reason for hiding this comment

Overview

Concerns

Verdict

Uh oh!

oz-for-oss Bot Apr 29, 2026

Choose a reason for hiding this comment

Uh oh!

solomonneas commented Apr 29, 2026

Uh oh!

oz-for-oss Bot left a comment

Choose a reason for hiding this comment

Overview

Concerns

Verdict

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

oz-for-oss Bot commented Apr 28, 2026 •

edited

Loading