fix(dirs): keep install-relative discovery paths out of the persisted defaults

[ErikBjare]

On first run, UserConfig::default() writes get_discovery_paths() into config.toml. That list included the install-relative paths: the deb/rpm lib dir and $APPDIR, which is a different /tmp/.mount_* dir on every AppImage run. Saved config paths outrank the runtime ones, so after moving or reinstalling, aw-tauri kept searching (and preferring) the first install's paths.

The module manager already adds get_install_discovery_paths() at runtime (discover_modules), so this drops them from the persisted defaults. The first-run awatcher check still searches them. New test: test_discovery_defaults_exclude_install_paths.

Scope change: this PR originally also discovered modules next to the .AppImage file (for the release zip layout). That's dropped. The AppImage should be self-contained and runnable from anywhere, with awatcher + aw-sync bundled inside it (ActivityWatch/activitywatch#1499), not depend on folders next to it. The review comment about persisting those paths led to the fix that remains here.

cargo test --lib: 49 passed; cargo clippy --all-targets -D warnings clean (macOS; the Linux paths are cfg(any(target_os = "linux", test))).

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-10T12:42:44.554729Z	5abecf3	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[ErikBjare]

@greptileai review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 209d882ffb

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[greptile-apps]

Confidence Score: 5/5

[Medium risk] Changes module discovery path handling for configuration defaults.

The PR appears safe to merge; no new actionable issues were found.

What we checked:

• Tests use separate install paths: Only the new test touches APPDIR. The other install-path tests pass their paths directly, so they do not use its temporary value.

Summary

The PR keeps install-relative module paths out of newly saved defaults while still searching them at runtime.

• The first-run Linux watcher check still includes bundled modules.
• The updated Linux test creates a fake AppImage tree and checks that its paths are not saved.
• No new actionable issues were found in the changes since the previous review.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[User discovery paths] --> B[Saved config defaults]
  A --> C[First-run Linux watcher check]
  D[Current install paths] --> C
  D --> E[Runtime module discovery]
  B --> E

Loading

_{Reviews (4) · Last reviewed commit: "test(dirs): make the persisted-defaults ..." · Reviewed by Greptile}

[TimeToBuildBob]

🤖 AI code review

This PR removes install-relative discovery paths (from get_install_discovery_paths) from the persisted defaults written by UserConfig::default() into config.toml, so that first-run config no longer pins paths like $APPDIR/usr/lib/aw-tauri/modules. It updates the Linux default module selection in lib.rs to still search install-relative paths at runtime when deciding whether aw-awatcher is available. It adds a Linux test that fakes an APPDIR mount and asserts those install paths are not present in get_discovery_paths().

Safe to merge — no P0/P1 findings

Confidence 5/5

✅ No thread-worthy findings. Advisory notes follow; they are retained without opening review threads.

1 advisory finding (summary-only, not scored)

These P2 guard, heuristic, trade-off, or documentation claims are retained for judgment without opening review threads.

⚠️ P2 medium — src-tauri/src/lib.rs

This is a fix(...) PR but no test files are included in the diff. Erik's feedback: 'where is the repro & fixes they are supposed to catch' (gptme#3441), 'that measurement should come with a regression test' (gptme#3446). Add a test that would have caught this bug. (Advisory: Erik merged all such PRs but consistently requested tests.)

Add a test file that reproduces the bug before the fix and passes after it.

_{How this was verified: static preflight: fix-commit + touched-files scan (rule 7)}

Files changed (2) — the diff as I read it

• src-tauri/src/dirs.rs — Removes get_install_discovery_paths() from get_discovery_paths() on Linux and macOS, adds doc comment, and adds test_discovery_defaults_exclude_install_paths.
• src-tauri/src/lib.rs — In UserConfig::default() Linux branch, chains get_install_discovery_paths() into the search paths used by module_available for aw-awatcher.

Previous review passes

commit	score	findings	engine	when
209d882ffb33	5/5	0	llm	2026-10-07 10:18 UTC
4bd0744fa61a	4/5	1	llm	2026-10-07 12:34 UTC
86e84040496f	4/5	1	llm	2026-10-07 12:52 UTC
be4f292f3a69	4/5	1	llm	2026-10-07 13:30 UTC
42992cee009b	4/5	1	llm	2026-10-07 21:52 UTC

_{Reviewed 42992cee009b · openrouter/deepseek/deepseek-v4-flash-0731 · llm engine · 30s · about this reviewer}

_{Maintainer commands}

_{@TimeToBuildBob review (own line) — fresh review · @TimeToBuildBob fix — a worker acts on the findings. Once per comment; 👀 = received.}

[ErikBjare]

@greptileai review

[TimeToBuildBob]

Checked 7baa209: install-relative paths are now runtime-only, and the first-run aw-awatcher availability check still includes them. That addresses the stale-bundle-path finding.

My earlier review's “no test files” advisory was incorrect: the regression tests are inline in dirs.rs, including test_linux_install_paths_appimage_run_in_place and test_discovery_defaults_exclude_install_paths. No extra test file is needed.

No code change or review retrigger from me. Greptile is green on this head; format, clippy, and release checks are still queued. I inspected the diff, but did not independently run the tests.

[ErikBjare]

@greptileai review

[TimeToBuildBob]

Checked the revised scope at 4bd0744: adjacent-.AppImage discovery is now dropped; the remaining diff only keeps install-relative paths out of persisted defaults while retaining them for the first-run watcher check. test_discovery_defaults_exclude_install_paths remains in the diff. This supersedes my earlier reference to the run-in-place test, which was removed with that feature.

Your latest Greptile request completed at 5/5 on this head, with no new findings. Format, clippy, and all five release jobs are still queued. No code changes, merge, or additional review trigger from me; I did not independently run the Rust tests.

[ErikBjare]

@TimeToBuildBob review

[ErikBjare]

@greptileai review
@TimeToBuildBob review

[ErikBjare]

@TimeToBuildBob review

[ErikBjare]

@TimeToBuildBob review

[TimeToBuildBob]

LGTM. The fix is correct and the scope is right.

Key invariants verified:

• get_discovery_paths() no longer persists install-relative paths (deb/rpm lib dir, AppImage $APPDIR), so config.toml stays valid after move/reinstall
• UserConfig::Default expands the search to include install paths at first-run time (the search vec in lib.rs) — bundled awatcher is still discovered, just not locked in
• Test uses AppDirGuard RAII to fake an AppImage mount, confirms APPDIR-rooted paths don't appear in get_discovery_paths()

The lib.rs change is Linux-only which is appropriate since the stale-mount bug is AppImage-specific. cargo clippy --all-targets -D warnings clean + 49 passing tests is sufficient CI for a focused dirs change.

[TimeToBuildBob]

Correct fix. The core invariant — install-relative paths are runtime-only, never persisted — is well-stated in the docstring and enforced by the test.

The lib.rs change is the right complement: default_linux_modules still searches both user and install paths to detect aw-awatcher (needed to set the right default), but only the user-level paths flow into the persisted config. The AppDirGuard test is thorough — it isolates the fake APPDIR scope and explicitly filters the exe's own dir from the assertion to avoid false positives when run from a user directory.

[TimeToBuildBob]

Correct fix. The core invariant — install-relative paths are runtime-only, never persisted — is well-stated in the docstring and enforced by the test.

The lib.rs change is the right complement: default_linux_modules still searches both user and install paths to detect aw-awatcher (needed to set the right default), but only the user-level paths flow into the persisted config. The AppDirGuard test is thorough — it isolates the fake APPDIR scope and explicitly filters the exe's own dir from the assertion to avoid false positives when run from a user directory.