fix(ls): force C locale so non-English month names don't break parsing

[GabriWar]

Summary

rtk ls returns (empty) for any non-empty directory when the user's locale prints non-English month abbreviations (e.g. LC_TIME=pt_BR.UTF-8 → abr, mar, out).

LS_DATE_RE in src/cmds/system/ls.rs is English-only and case-sensitive:

r"\s+(Jan|Feb|Mar|Apr|May|Jun|Jul|Aug|Sep|Oct|Nov|Dec)\s+\d{1,2}\s+(?:\d{4}|\d{2}:\d{2})\s+"

When ls -la prints a Portuguese/French/etc. month, the regex doesn't match, parse_ls_line returns None for every line, and compact_ls falls through to the (empty) sentinel at line 194 — silently hiding every file and directory.

Repro

$ echo $LC_TIME
pt_BR.UTF-8

$ /usr/bin/ls -la ~/.local/bin/ | head -3
total 330912
drwxr-xr-x 1 gabriel gabriel      2592 abr 18 18:59 .
-rwxr-xr-x 1 gabriel gabriel        99 jan 13 16:25 agent

$ rtk -vvv ls ~/.local/bin/
Chars: 9773 → 8 (100% reduction)
(empty)

A 100% reduction on a directory with 140+ entries is a strong hint the filter is eating real data.

Fix

Set LC_ALL=C on the spawned ls process, forcing English month names regardless of the user's locale. Smallest possible change, localized to where the locale-dependent parsing happens.

cmd.env("LC_ALL", "C");

This keeps the existing regex/parser contract intact (it already assumes a stable English format) and doesn't touch any other command path.

After

$ rtk ls ~/.local/bin/ | head -3
agent  99B
aider -> /home/gabriel/.local/share/uv/tools/aider-chat/bin/aider  56B
alphashape  185B

Scope check: grep -rn "Jan|Feb|Mar" src/ confirms ls.rs is the only runtime path with locale-dependent date parsing. tree / find / other proxies don't hit this.

Test plan

• Manual repro under LC_TIME=pt_BR.UTF-8 before fix — reproduces (empty)
• Manual verification after fix — rtk ls lists the directory correctly
• cargo build --release clean
• Existing ls tests unaffected (they feed the parser directly and are locale-independent)

[CLAassistant]

All committers have signed the CLA.

[pszymkowiak]

[w] wshm · Automated triage by AI

📊 Automated PR Analysis

🐛 Type	bug-fix
🟢 Risk	low

Summary

Fixes a bug where rtk ls returns (empty) for non-empty directories when the user's locale uses non-English month abbreviations. The fix sets LC_ALL=C on the spawned ls process so month names always match the English-only regex.

Review Checklist

• Tests present
• Breaking change
• Docs updated

---

Analyzed automatically by wshm · This is an automated analysis, not a human review.

[Copilot]

Pull request overview

Fixes rtk ls incorrectly returning (empty) under non-English locales by forcing the spawned ls process into a locale that matches the English-only LS_DATE_RE parsing contract.

Changes:

• Set a fixed locale on the ls subprocess to ensure month abbreviations match the parser’s regex.
• Add inline documentation explaining the locale/parsing interaction and failure mode.

[Copilot]

Using LC_ALL=C forces all locale categories for the spawned ls process. That can change filename rendering/escaping and collation (and may cause non-ASCII filenames to be shown differently), even though this bug only depends on month names from LC_TIME. Prefer setting LC_TIME=C (or LC_TIME=POSIX) so date parsing is stabilized without affecting character encoding/display behavior.

Suggested change

	// Force C locale so month names match LS_DATE_RE, which is English-only.
	// Without this, non-English locales (e.g. LC_TIME=pt_BR.UTF-8 → "abr"/"mar")
	// cause every line to fail the regex, and compact_ls returns "(empty)"
	// even for non-empty directories.
	cmd.env("LC_ALL", "C");
	// Force only the time locale to C so month names match LS_DATE_RE,
	// which is English-only. Without this, non-English locales
	// (e.g. LC_TIME=pt_BR.UTF-8 → "abr"/"mar") cause every line to fail
	// the regex, and compact_ls returns "(empty)" even for non-empty
	// directories, while preserving other locale-sensitive output behavior.
	cmd.env("LC_TIME", "C");