GitoxideLabs/gitoxide gix-discover-v0.57.0

gix-discover v0.57.0

on GitHub

Bug Fixes

• derive trust from all detected repository paths

  There are some concerns over performance, but it seems the work it does is
  required after all. There are possible optimisations left on the table,
  as it might do duplicate IO to check ownership.
  More tests are read-only now as well, it really does like its tempdirs and
  direct Git invocations.

  Checking ownership of only the discovery candidate or resolved git directory
  can grant full trust to a linked repository with foreign-owned paths. Add
  gix_discover::repository::trust() to compute the minimum trust of the
  original candidate, resolved git directory, common directory referenced by
  commondir, and existing worktree. Use it during upward discovery and direct
  and environment-based gix opens, before selecting trust-dependent options
  or reading repository configuration.

  Never upgrade an untrusted candidate. Skip a duplicate git-directory check
  only when its spelling exactly matches the candidate after joining with the
  current directory; filesystem aliases can still cause repeated checks. Skip
  a missing worktree so a private git directory remains usable after checkout
  removal, but propagate missing git or common-directory errors.

  Preserve TrustPolicy::Assume and explicit opening trust overrides, and
  document the distinction between ownership-derived trust, required thresholds,
  and assumed trust. Add deterministic ownership regressions through private
  callbacks, including linked paths, duplicate checks, and missing directories.
  Replace the ancestor-discovery test's dependency on the source checkout with
  read-only nested repository fixtures.

  Git reference: setup.c::ensure_valid_ownership at
  d38352cd43ab9745686d697872408bc3249a153f checks gitfile, worktree and gitdir.
  The common directory is also checked here as a configuration source.

Commit Statistics

• 9 commits contributed to the release over the course of 13 calendar days.
• 13 days passed between releases.
• 1 commit was understood as conventional.
• 0 issues like '(#ID)' were seen in commit messages

Commit Details