v2.0.3

Bugfixes

1. We were trying to open a cache file when using -c --no-cache, breaking some CI workflows. This is now handled more gracefully.

2. When comparing mod times, we used the precision the underlying file system gave us. Some formatters, such as dos2unix, manipulate the mod time, but not to the same precision. This was causing false positives when using --no-cache --fail-on-change. We now truncate the mod time to second-level precision before comparing, as POSIX specifies that mod time should be EPOCH (second) precision. This brings us back in line with v1 behaviour.

Summary Improvements

We have improved the summary statistics presented at the end of each run, clarifying what has happened and removing unnecessary precision in the elapsed time.

❯ nix fmt 
traversed 116 files
emitted 116 files for processing
formatted 67 files (0 changed) in 696ms

Changelog

• a642145: feat: display elapsed time to nearest millisecond (@brianmcgee)
• f266fc8: feat: if we can't open the cache, fallback to no cache (@brianmcgee)
• da7b015: feat: improve summary stats (@brianmcgee)
• 33a7bba: fix: --no-cache (@brianmcgee)
• 85ce0a2: fix: use second precision when comparing file mod times (@brianmcgee)

Nixpkgs PR Tracker

https://nixpk.gs/pr-tracker.html?pr=325465