[costajohnt]

Fixes #8922.

Locale detection only read LANG, so LC_ALL and LC_MESSAGES were ignored. GNU honors all three, so LC_ALL=fr_FR.UTF-8 sleep --help prints French there but English in uutils.

This applies the POSIX message-locale precedence LC_ALL > LC_MESSAGES > LANG: the first of those that is set and non-empty wins (a set-but-empty value is treated as unset), otherwise it falls back to the existing en-US default. The winning value keeps the same .<encoding> stripping as before (fr_FR.UTF-8 -> fr_FR), so the LANG-only path is unchanged.

The precedence lives in a small pure helper that takes the env lookup as an argument, so it can be unit-tested without mutating process-global env vars (which race under parallel test threads):

fn resolve_locale_string(lookup: impl Fn(&str) -> Option<String>) -> String

detect_system_locale just calls it with std::env::var.

Behavior change

Anyone who has LC_ALL or LC_MESSAGES set to something other than LANG will now get that locale for messages instead of LANG. That is the intended fix for #8922, but noting it since it is user-visible.

Tests

Seven unit tests on the helper: LC_ALL wins, LC_MESSAGES when LC_ALL is unset, LANG still works, empty treated as unset, LC_ALL over LANG, LC_ALL over LC_MESSAGES, and nothing-set fallback.

cargo test, cargo clippy --workspace --all-targets --all-features -- -D warnings, and cargo fmt --check all pass.

[github-actions]

GNU testsuite comparison:

Skip an intermittent issue tests/tail/tail-n0f (fails in this run but passes in the 'main' branch)
Congrats! The gnu test tests/seq/seq-epipe is no longer failing!
Note: The gnu test tests/cp/link-heap is now being skipped but was previously passing.
Note: The gnu test tests/cut/bounded-memory is now being skipped but was previously passing.
Note: The gnu test tests/tail/pipe-f is now being skipped but was previously passing.

[codspeed]

Merging this PR will not alter performance

1 warning, 1 note

⚠️ 114 benchmarks are dominated by system calls

System calls cannot be simulated consistently and are left out of the measure. The reported value understates the real cost. Switch to the Walltime instrument to measure them.

Learn more about system calls

See affected benchmarks

ℹ️ 1 benchmark has no flamegraph

The flamegraph wasn't generated: most of the time was spent in runtime internals, which would leave nothing useful to analyze.

See affected benchmarks

✅ 401 untouched benchmarks
⏩ 70 skipped benchmarks¹

---

_{Comparing costajohnt:fix/8922-lc-all-locale-precedence (8179408) with main (366c18c)²}

Footnotes

1. 70 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

2. No successful run was found on main (6f5138a) during the generation of this report, so 366c18c was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

[github-actions]

Binary size comparison:

Individual binary size comparison VS main (threshold: >=5% AND >=4 KB).

Total size of compared binaries: 147.12 MB (-4.13 MB, -2.73%)

Significant per-binary changes:
  ls     3.28 MB ->    2.11 MB  (-1.17 MB, -35.76%)