Repository navigation
Tags
Tags
v0.3.1 On the 94-tool development set, tools judged outright wrong fell from 27 to 7. Four defect families repaired and ratcheted at zero: single-dash long options (132 tools, 8,784 flags), bare-block extent, repeated-character flags, and bundled shorts. Fleet effect +498 flags across 45 tools, with six tools going from zero flags to a full table. New instruments: per-family detector calibration against human verdicts, sweep-diff, a must_not_contain_flags contract field so an invention defect can be pinned, and a frozen 2,299-tool sampling queue. No fleet accuracy figure is claimed; spec.md's [M-20] stays unfilled until a fresh draw of unseen tools is reviewed.
v0.2.2 — tab-aligned tables, and 89.23% -> 94.18% Three general parser fixes: - A tab counts as a column gap. `mokutil --help` was 38 flags / 0 described with the descriptions plainly present; now 38 at 100%. Also recovers 11 real commands for mysqladmin/mariadb-admin. - A second column of option *spellings* is not a description. `awk --help` would otherwise have reported 28 flags, 100% described, every description a lie. - A positional documented as the first row of an options table no longer discards the table. `kill` goes 0 -> 6 flags at 100%. Records [M-14]: what a man-page tier would recover, measured rather than assumed.
v0.2.1 — the empty argument that actually reset a machine, and `-h` t… …hat would have Fixes `<tool> -- ""`, measured terminating every process in a PID namespace via `pkill -- ""`, and emitted at every tool on PATH. Refused now at the exec chokepoint for all tools rather than hidden behind a thirteen-name list. Removes the clap CompleteEnv probe, which produced that argv and never identified a real clap tool. Restricts process-signalling and machine-state tools to exactly `--help` instead of banning them: `-h` was measured *attempting the real operation* on halt/poweroff/reboot/shutdown. Eleven tools gain real flag lists; no-tier 14 -> 2, described 89.19% -> 89.23%.
PreviousNext