fix(git): skip untracked scan when HK_STASH_UNTRACKED=false by jdx · Pull Request #861 · jdx/hk

jdx · 2026-04-23T20:24:06Z

Summary

When GIT_WORK_TREE points at a large directory (e.g. a YADM dotfile repo where the worktree is $HOME), the unconditional git status --untracked-files=all scan takes tens of seconds and emits hundreds of megabytes of output on every hk invocation.
HK_STASH_UNTRACKED=false previously had no effect on the status scan — it only suppressed stashing. This change honors it in Git::status() (and the internal stash porcelain parse, which already ignores untracked entries) so users of large worktrees can opt out of the scan entirely. Both the libgit2 and shell-git paths are updated.
Fixes #860.

Test plan

mise run test:bats test/bare_repo_env_vars.bats (7 tests × 2 modes)
mise run test:bats test/stash_untracked_with_partial_stash.bats, test/stash_preservation_on_error.bats, test/pre_commit_does_not_stash_staged_only_files.bats, test/git_status_ad_deleted.bats, test/cargo_fmt_does_not_stage_manifest.bats
mise run test:cargo (145 passed)
New regression test: HK_STASH_UNTRACKED=false skips untracked scan in large worktree (#860)

🤖 Generated with Claude Code

Note

Medium Risk
Changes how Git::status() (libgit2 and shell-git paths) and stash detection call git status, which can affect which files hk considers (especially untracked files) and therefore which steps run. The behavior is gated behind HK_STASH_UNTRACKED=false, but it touches core git/file-detection logic.

Overview
Honors HK_STASH_UNTRACKED=false by skipping untracked-file enumeration during status collection (using --untracked-files=no for shell git and disabling untracked options for libgit2) to avoid expensive scans when GIT_WORK_TREE is very large.

Updates the stash porcelain parse to likewise avoid untracked scanning when untracked stashing is disabled, documents the new behavior in docs/environment_variables.md, and adds a Bats regression test ensuring untracked “junk” in a large worktree doesn’t appear in hk check --all output.

^{Reviewed by Cursor Bugbot for commit 4049f37. Bugbot is set up for automated code reviews on this repo. Configure here.}

gemini-code-assist · 2026-04-23T20:24:10Z

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

greptile-apps · 2026-04-23T20:26:18Z

Greptile Summary

This PR makes HK_STASH_UNTRACKED=false actually suppress the underlying git status --untracked-files=all scan, fixing a serious performance regression for large GIT_WORK_TREE directories (e.g. YADM dotfile repos where the worktree is $HOME). Both the libgit2 path (StatusOptions::include_untracked / recurse_untracked_dirs) and the shell-git path (the --untracked-files=… flag) are updated in Git::status(), and the stash porcelain parse in section 3 is updated for consistency. Documentation is updated to document the new side-effect of the flag.

Confidence Score: 5/5

Safe to merge — the logic change is correct in both code paths, preserves existing behaviour when the env var is not set, and is documented.

The only open concern (regression-test assertion quality) was already raised in a prior review thread; no new P0/P1 issues are present. The Rust change is minimal, well-contained, and correctly honours the existing HK_STASH_UNTRACKED semantics in all four affected code sites.

No files require special attention beyond the already-discussed test assertion in test/bare_repo_env_vars.bats.

Important Files Changed

Filename	Overview
src/git.rs	Reads HK_STASH_UNTRACKED once per status() call and threads it through both the libgit2 and shell-git code paths; also propagates it to the stash porcelain parse in section 3 — logic is correct and complete.
docs/environment_variables.md	Adds a clear explanation of the new HK_STASH_UNTRACKED=false scan-skipping behavior, including the YADM / large-GIT_WORK_TREE motivation — accurate and useful.
test/bare_repo_env_vars.bats	Regression test for #860 exercises the code path correctly but the refute_output assertion (already flagged in a prior review thread) may pass even without the fix since untracked filenames are not normally included in hk check --all output.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A["Git::status() called"] --> B["read *env::HK_STASH_UNTRACKED\n→ include_untracked: bool"]
    B --> C{libgit2 repo\navailable?}
    C -- yes --> D["status_options.include_untracked(include_untracked)\nstatus_options.recurse_untracked_dirs(include_untracked)"]
    D --> E["run libgit2 staged scan\nrun libgit2 workdir scan\n(untracked entries absent when false)"]
    C -- no --> F{include_untracked?}
    F -- true --> G["--untracked-files=all"]
    F -- false --> H["--untracked-files=no"]
    G & H --> I["git status --porcelain FLAG -z"]
    E & I --> J["return GitStatus\n(untracked_files empty when false)"]
    J --> K["stash porcelain parse (§3)\nsame FLAG applied"]

_{Reviews (2): Last reviewed commit: "fix(git): skip untracked scan when HK_ST..." | Re-trigger Greptile}

When `GIT_WORK_TREE` points at a very large directory (e.g. a YADM dotfile repo where the worktree is `$HOME`), the unconditional `git status --untracked-files=all` scan takes tens of seconds and produces hundreds of megabytes of output on every hk invocation. The `HK_STASH_UNTRACKED=false` env var previously had no effect on the status scan itself — it only suppressed stashing. Honor it in `Git::status()` (and the internal stash porcelain parse, which already ignores untracked entries) so that users of large worktrees can opt out of the scan entirely. Both the libgit2 and shell-git paths are updated. Closes #860 Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>

@jdx

### 🐛 Bug Fixes - **(git)** skip untracked scan when HK_STASH_UNTRACKED=false by [@jdx](https://github.com/jdx) in [#861](#861) - **(run)** add post-commit and pre-rebase subcommands by [@jdx](https://github.com/jdx) in [#858](#858) ### 📚 Documentation - **(install)** recommend global hooks as primary setup path by [@jdx](https://github.com/jdx) in [#855](#855) - add cross-site announcement banner by [@jdx](https://github.com/jdx) in [#857](#857) - respect banner expires field by [@jdx](https://github.com/jdx) in [#862](#862) ### 🔍 Other Changes - vendor bats test helpers instead of git submodules by [@jdx](https://github.com/jdx) in [#859](#859) ### 📦️ Dependency Updates - bump communique to 1.0.3 by [@jdx](https://github.com/jdx) in [#863](#863) - update anthropics/claude-code-action digest to e58dfa5 by [@renovate[bot]](https://github.com/renovate[bot]) in [#864](#864)  --- > [!NOTE] > **Low Risk** > Low risk release bookkeeping only: version numbers, generated CLI/docs, and lockfile dependency bumps with no runtime logic changes. > > **Overview** > **Release prep for `v1.44.1`.** Updates `Cargo.toml`/`Cargo.lock` (including `libc`) and all version references in generated CLI docs (`docs/cli/*`, `hk.usage.kdl`). > > Refreshes documentation examples to reference the new `v1.44.1` Pkl package URLs and adds the `1.44.1` entry to `CHANGELOG.md`. > > <sup>Reviewed by [Cursor Bugbot](https://cursor.com/bugbot) for commit d7637d4. Bugbot is set up for automated code reviews on this repo. Configure [here](https://www.cursor.com/dashboard/bugbot).</sup>  Co-authored-by: mise-en-dev <[email protected]>

greptile-apps Bot reviewed Apr 23, 2026

View reviewed changes

Comment thread test/bare_repo_env_vars.bats

jdx force-pushed the fix/git-status-skip-untracked-when-disabled branch from 5fe8fdf to 4049f37 Compare April 23, 2026 20:48

jdx enabled auto-merge (squash) April 23, 2026 20:56

jdx merged commit dcc9e64 into main Apr 23, 2026
20 checks passed

jdx deleted the fix/git-status-skip-untracked-when-disabled branch April 23, 2026 20:58

jdx mentioned this pull request Apr 23, 2026

chore: release v1.44.1 #856

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix(git): skip untracked scan when HK_STASH_UNTRACKED=false#861

fix(git): skip untracked scan when HK_STASH_UNTRACKED=false#861
jdx merged 1 commit intomainfrom
fix/git-status-skip-untracked-when-disabled

jdx commented Apr 23, 2026 •

edited by cursor Bot

Loading

Uh oh!

gemini-code-assist Bot commented Apr 23, 2026

Uh oh!

greptile-apps Bot commented Apr 23, 2026 •

edited

Loading

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

jdx commented Apr 23, 2026 • edited by cursor Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test plan

Uh oh!

gemini-code-assist Bot commented Apr 23, 2026

Uh oh!

greptile-apps Bot commented Apr 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Greptile Summary

Confidence Score: 5/5

Important Files Changed

Flowchart

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

jdx commented Apr 23, 2026 •

edited by cursor Bot

Loading

greptile-apps Bot commented Apr 23, 2026 •

edited

Loading