GitoxideLabs/gitoxide gix-filter-v0.36.1

gix-filter v0.36.1

on GitHub

Performance

• only read the index object when there is a CRLF to convert
  In auto-text mode, eol::convert_to_git() read the file's object from the index
  to see if it has CRLF, even when the worktree content had no CRLF to convert.
  That check can only prevent a conversion, so without a CRLF the result is
  always the same. Reading the object means a pack lookup and decompression for
  every file. With core.autocrlf=input, that was 38% of the user CPU of
  gix status on a repository whose index stat was out of date.

  Git also reads the object in that case, but it rarely gets there for unchanged
  files as its status refreshes the index.

Bug Fixes

• run filter drivers from the worktree root

  Status now passes worktree-relative paths to filters. When invoked outside
  that worktree, a required clean filter reading sibling configuration through
  %f inherited the caller CWD and failed to find .clean-map.

  Set the CWD in the shared driver launcher so both single-file and process
  filters run from the configured worktree, matching Git.

  Make gix_command::Context require absolute repository paths and have
  Repository::command_context() resolve GIT_DIR and GIT_WORK_TREE
  against Repository::current_dir(). Using the live process CWD in the
  filter launcher would lose the base captured when a relative repository
  was opened. Anchoring paths at context creation keeps them valid after
  either the caller or the child changes CWD, without redundant resolution
  or environment overrides in the filter launcher.

  Extend the status regression to read .clean-map and compare against
  Git using git diff --exit-code from a subdirectory. Cover a filter
  invoking Git from outside its worktree with an absolute context, and
  verify that command contexts for normal and bare repositories retain the
  captured CWD after the process changes directory. Update command tests
  to use absolute paths and verify unchanged environment forwarding.

Commit Statistics

• 7 commits contributed to the release over the course of 14 calendar days.
• 14 days passed between releases.
• 2 commits were understood as conventional.
• 0 issues like '(#ID)' were seen in commit messages

Commit Details