fix: strip linked-worktree '+' marker from branch picker (#9170) by lonexreb · Pull Request #9560 · warpdotdev/warp

lonexreb · 2026-04-30T07:31:03Z

Description

The shell git-branch chip in the prompt input shows a list of branches that the user can click to run git checkout <branch>. The chip's data source is the porcelain output of:

git --no-optional-locks branch --no-color --sort=-committerdate

git branch prefixes each line with one of three markers:

Marker	Meaning
`*`	Branch checked out in this worktree
`+`	Branch checked out in another linked worktree
(space)	Every other branch

filter_git_branch_on_click_values in app/src/context_chips/current_prompt.rs was stripping only the * marker. When a user clicked a +-prefixed entry, the value (still containing the +) was passed straight into format!("git checkout {branch_name}"), producing:

git checkout + 273-improvement-suggestion-agent

which fails:

error: pathspec '+' did not match any file(s) known to git
error: pathspec '273-improvement-suggestion-agent' did not match any file(s) known to git

(matching the screenshots in the bug report exactly).

Change

Single-line behavioral fix:

- .map(|s| s.trim_start_matches('*').trim().to_string())
+ .map(|s| s.trim_start_matches(['*', '+']).trim().to_string())

The function is also converted from &self to an associated fn (it never used self) so it can be tested directly without spinning up a full CurrentPrompt model. The single caller is updated from self. to Self::.

A +-marked branch is one that's already checked out in another worktree; running git checkout on it from the current worktree will still surface git's own (more meaningful) error. This change just stops Warp from generating a malformed command before git ever sees it.

Testing

Added five unit tests in app/src/context_chips/current_prompt_test.rs:

filter_git_branch_strips_star_marker — pre-existing behavior is preserved.
filter_git_branch_strips_plus_marker_for_linked_worktrees — explicit regression for Branch picker includes linked-worktree '+' marker in git checkout #9170, with a final assert that no value retains a leading + or *.
filter_git_branch_handles_mixed_marker_set — full git branch sample (*, +, space) verified end-to-end including sort order.
filter_git_branch_passes_through_none — None input returns None.
filter_git_branch_drops_blank_lines — pre-existing behavior is preserved.

cargo fmt -- --check clean. (Local clippy/nextest were not run — repo's script/presubmit requires the full Xcode metal compiler for the warpui build, which the local sandbox blocks. Relying on CI for full presubmit verification.)

Server API dependencies

Is this change necessary to make the client compatible with a desired server API breaking change? — No, client-only.

Agent Mode

Warp Agent Mode

Changelog Entries for Stable

CHANGELOG-BUG-FIX: Fixed branch picker generating an invalid git checkout + <branch> command for branches checked out in another linked git worktree. Thanks @lonexreb and @tkshsbcue!

The shell git-branch chip's data source is `git --no-optional-locks branch --no-color --sort=-committerdate`, whose porcelain output marks each branch with one of three leading characters: * the branch checked out in this worktree + a branch checked out in another linked worktree (space) every other branch `filter_git_branch_on_click_values` only stripped the `*` marker. When a user clicked a `+`-prefixed entry, the value `"+ 273-improvement-suggestion-agent"` was passed verbatim into `format!("git checkout {branch_name}")`, producing: git checkout + 273-improvement-suggestion-agent …which fails with `pathspec '+' did not match any file(s) known to git`. This change extends the marker-stripping to also drop `+`, so users with linked-worktree repos can pick those branches from the chip. A `+`-marked branch is one that's already checked out in another worktree; `git checkout` on it from this worktree will still produce its own (more meaningful) error, which preserves the existing UX surface for that case while unblocking the common scenario of selecting any other `+`-prefixed branch. Also makes `filter_git_branch_on_click_values` an associated function — it never used `self` — to allow direct unit testing without spinning up a full `CurrentPrompt` model. Adds tests covering the `*` and `+` markers, mixed sorting/ordering, blank-line filtering, and the `None` pass-through. Closes warpdotdev#9170

cla-bot · 2026-04-30T07:31:07Z

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 @lonexreb 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.

oz-for-oss · 2026-04-30T07:31:17Z

@lonexreb

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

You can view the conversation on Warp.

I reviewed this pull request and requested human review from: @moirahuang.

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 updates git branch chip normalization so linked-worktree entries prefixed by Git's + marker are passed through as raw branch names instead of producing malformed git checkout + <branch> commands. It also adds focused unit coverage for *, +, mixed marker sets, blank lines, and None input.

Concerns

No blocking correctness or security concerns found in the changed 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

lonexreb · 2026-04-30T15:48:33Z

@cla-bot check

cla-bot · 2026-04-30T15:48:37Z

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

moirahuang · 2026-04-30T18:23:28Z

    fn update_on_click_value(&mut self, chip_kind: &ContextChipKind, value: Option<Vec<String>>) {
        log::debug!("Updating prompt on_click value of {chip_kind:?} to {value:?}");
        let filter_values = match chip_kind {
-            ContextChipKind::ShellGitBranch => self.filter_git_branch_on_click_values(value),


nit: i think we don't need to make this change + don't need to change the function signature below on line 618

moirahuang · 2026-04-30T19:51:33Z

thank you for making this PR! mind updating the changelog to give yourself and also @tkshsbcue attribution? they had also made a PR for this issue but i prefer your implementation. more context:
#9440 (comment)

lonexreb · 2026-04-30T20:03:08Z

Thanks @moirahuang! Updated the CHANGELOG-BUG-FIX line to credit both myself and @tkshsbcue:

CHANGELOG-BUG-FIX: Fixed branch picker generating an invalid git checkout + <branch> command for branches checked out in another linked git worktree. Thanks @lonexreb and @tkshsbcue!

Happy to adjust the wording if you'd prefer a different attribution format.

moirahuang · 2026-04-30T22:37:50Z

Thanks @moirahuang! Updated the CHANGELOG-BUG-FIX line to credit both myself and @tkshsbcue:

CHANGELOG-BUG-FIX: Fixed branch picker generating an invalid git checkout + <branch> command for branches checked out in another linked git worktree. Thanks @lonexreb and @tkshsbcue!

Happy to adjust the wording if you'd prefer a different attribution format.

Thank you for making the change! Please address this nit and then I will merge the PR in!

moirahuang asked to revert the conversion of `filter_git_branch_on_click_values` from `&self` instance method to an associated function (call-site at line 357 plus signature change at line 618). Restoring the original signature and the original `self.filter_git_branch_on_click_values(value)` call site. Drops the five unit tests added in this branch that depended on the associated-function form (constructing a `CurrentPrompt` purely for a unit test would require wiring `Sessions`, `ModelContext`, and `ModelHandle` plumbing — out of scope for this fix). The behavioral fix is the one-line `trim_start_matches(['*', '+'])` change plus the comment documenting the three `git branch` marker variants, which is what the issue actually required.

lonexreb · 2026-04-30T23:44:06Z

Pushed 6540f0e addressing the nit on current_prompt.rs:357.

Restored the original self.filter_git_branch_on_click_values(value) call site at line 357.
Restored the original fn filter_git_branch_on_click_values(&self, ...) signature on the impl.
Dropped the five unit tests I had added in this branch — they depended on the associated-function form. Constructing a CurrentPrompt purely for a unit test of this helper would require wiring Sessions / ModelContext / ModelHandle plumbing, which felt out of scope for this fix.

Net diff vs. master is now just the one-line .trim_start_matches(['*', '+']) change plus the explanatory comment about the three git branch markers — the actual behavioral change requested in #9170. Ready when you are.

moirahuang · 2026-05-01T17:28:50Z

@lonexreb Appreciate you updating! Instead of deleting the tests, I think we should be able to follow test_context_menu_items as an example and still construct current_prompt

unrevised6419 · 2026-05-02T14:04:59Z

Heads up: this PR strips the + marker but the underlying checkout will still fail because the branch is already used by a linked worktree (fatal: '<branch>' is already checked out at ...). See #9170 (comment) for details and suggested fixes.

moirahuang · 2026-05-05T18:53:59Z

@lonexreb FYI the PR #9905 (review) is a more complete solution (thanks @unrevised6419 for this comment!). I've asked them to give you attribution in the CHANGELOG

lonexreb · 2026-05-05T19:07:03Z

Thanks @moirahuang and @unrevised6419 — totally agree #9905 is the right call. @unrevised6419's point is correct: stripping the marker only fixes the cosmetic side; the underlying git checkout still fails on a branch that's already in a linked worktree. #9905's approach (handling the worktree case proper) is the complete fix.

Happy to close this one in favor of #9905. The CHANGELOG attribution offer is appreciated but absolutely not necessary — go with whatever feels right.

moirahuang · 2026-05-06T04:02:30Z

Thanks @moirahuang and @unrevised6419 — totally agree #9905 is the right call. @unrevised6419's point is correct: stripping the marker only fixes the cosmetic side; the underlying git checkout still fails on a branch that's already in a linked worktree. #9905's approach (handling the worktree case proper) is the complete fix.

Happy to close this one in favor of #9905. The CHANGELOG attribution offer is appreciated but absolutely not necessary — go with whatever feels right.

Thank you going to close this one!

oz-for-oss Bot reviewed Apr 30, 2026

View reviewed changes

oz-for-oss Bot requested a review from moirahuang April 30, 2026 07:33

cla-bot Bot added the cla-signed label Apr 30, 2026

moirahuang mentioned this pull request Apr 30, 2026

Strip linked-worktree '+' marker from git branch picker #9440

Open

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

moirahuang reviewed Apr 30, 2026

View reviewed changes

lonexreb mentioned this pull request May 1, 2026

fix: enable Pi CLI agent session notifications #9674

Draft

captainsafia mentioned this pull request May 1, 2026

fix(branch-picker): strip linked-worktree '+' marker from branch names #9821

Closed

2 tasks

unrevised6419 mentioned this pull request May 2, 2026

Branch picker includes linked-worktree '+' marker in git checkout #9170

Open

moirahuang mentioned this pull request May 5, 2026

Fix linked-worktree branch checkout #9905

Open

3 tasks

Conversation

lonexreb commented Apr 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Change

Testing

Server API dependencies

Agent Mode

Changelog Entries for Stable

Uh oh!

cla-bot Bot commented Apr 30, 2026

Uh oh!

oz-for-oss Bot commented Apr 30, 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!

lonexreb commented Apr 30, 2026

Uh oh!

cla-bot Bot commented Apr 30, 2026

Uh oh!

moirahuang Apr 30, 2026

Choose a reason for hiding this comment

Uh oh!

moirahuang commented Apr 30, 2026

Uh oh!

lonexreb commented Apr 30, 2026

Uh oh!

moirahuang commented Apr 30, 2026

Uh oh!

lonexreb commented Apr 30, 2026

Uh oh!

moirahuang commented May 1, 2026

Uh oh!

unrevised6419 commented May 2, 2026

Uh oh!

moirahuang commented May 5, 2026

Uh oh!

lonexreb commented May 5, 2026

Uh oh!

moirahuang commented May 6, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

lonexreb commented Apr 30, 2026 •

edited

Loading

oz-for-oss Bot commented Apr 30, 2026 •

edited

Loading