GitoxideLabs/gitoxide gix-object-v0.48.0

gix-object v0.48.0

on GitHub

New Features

• add commit::ExtraHeaders::find_pos(), and expose "gpgsig" header name.
  That way it it's easier to manipulate existing extra-header fields.
• add CommitRef|TagRef|BlobRef::into_owned() as shortcut.
  Otherwise one would have to know it converts into Commit via From.

Bug Fixes

• Support empty tags with or without trailing NL
  When representing an annotated tag with an empty commit message, we used
  to only support a tag ending with two newlines (one after the tagger
  line + one after the empty commit message).

  This was due to a misconception that annotated tags ending with a single
  NL shouldn't exist in the wild since there isn't an obvious way to
  create them from with git.
  This misconception shows up in the discussion thread for
  issue 603

  It turns out that both encodings of empty annotated tags appear in the
  wild.

  We must be able to parse either and roundtrip for either.

  Before PR 604, we
  used to special case the empty tag msg case and not add a NL.
  To be able to represent b"", b"\n", b"\n\n"..., we special case
  any message that is a pure sequence of b'\n' and actually parse the
  b'\n' into the tag's message.

  This allows us to calculate the correct size that matches the number of
  bytes that git would produce, as well as round-trip to and from the
  commit encoding.

  The existing tests (in particular round_trip for empty.txt and
  empty_missing_nl.txt) convince me that the logic is sound.
  Also, the size being 139 bytes for empty_missing_nl.txt and 140
  bytes for empty.txt matches the output of cat file | wc -l for each
  file.

Other

• Repository::commit() now explains how to create a commit without ref updates.

Changed (BREAKING)

• drop migration shims for fallible hashing
  Since the APIs were already adjusted and all callers migrated, we
  only need to drop the migration shims.
• make gix_object::compute_hash fallible
  compute_stream_hash is already fallible, so we don’t want to keep
  the try_* prefix on the fallible API.
• adjust error return types to handle collision detection
  This does mean a lot of churn across the tree, but the change is
  usually just an adjustment to variants of an existing error type,
  so I expect that most downstream users will require little to no
  adaption for this change.
• use separate error type for I/O hashing operations
  Prepare for hashing becoming fallible.
• adjust hash verification return types for the common interface
  This mostly just affects return types – using
  git_hash::verify::Error instead of bespoke duplicated versions
  thereof, and occasionally returning an ObjectId instead of ()
  for convenience.

Commit Statistics

• 33 commits contributed to the release.
• 9 commits were understood as conventional.
• 0 issues like '(#ID)' were seen in commit messages

Thanks Clippy

Clippy helped 1 time to make code idiomatic.

Commit Details