build(deps): bump gix from 0.68.0 to 0.85.0

[dependabot]

Bumps gix from 0.68.0 to 0.85.0.

Release notes

Sourced from gix's releases.

gix-odb v0.84.0

Refactor (BREAKING)

• add object_hash to gix-odb::Store::at_opts() and remove it from options This aligns at and at_opts, putting mandatory information as separate parameters.

New Features (BREAKING)

• require object_hash in Store::at Split Store::at_opts for callers that want to pass Options.
• (re-)introduce at(path, hash_kind) This way, a hash must be chosen, so that SHA-1 isn't hardcoded anymore.

Changed (BREAKING)

• fall back to the raw text when unquoting fails. Both callers of gix_quote::ansi_c::undo() turned a failure into an error of their own, while Git keeps the raw, still-quoted token as the value:

  • parse_attr_line() in attr.c falls through to its unquoted branch
  • Git's alternates parsing in odb.c spells the unterminated case out in a comment of its own

  For attributes this also covers invalid escapes, which were previously a hard error for the whole line. Git keeps "\!x" as the pattern, where the backslash goes on to escape the ! for the matcher rather than negating the pattern.

  Both fall back before checking for the [attr] macro prefix, as Git does, so a line like "[attr]x stays a pattern on either side.

  The Unquote variants are unreachable now and have been removed.

Test

• use gix_testtools::object_hash in a few tests

• explicitly pass Sha1 for static fixtures

• exercise alternates parsing without touching the filesystem. The previous test reached content() through alternate::resolve(), which meant creating a directory whose name begins with the quote that is never closed. That is not a legal filename on Windows, where it failed with InvalidFilename.

  Calling content() directly tests the same parsing without a filesystem, and adds a companion case showing that intact quoting still decodes its escapes.

New Features

• expose alternate-file parsing via alternate::parse() This is mostly for completeness and more low-level access.

Bug Fixes

... (truncated)

Commits

• 6428edc Release gix-date v0.15.5, gix-hashtable v0.15.2, gix-object v0.62.0, gix-attr...
• 1d301e2 report for June 2026
• da6b267 Merge pull request #2642 from 10ne1/dev/aratiu/sha256-transport
• 3927d70 Add journey-tests for SHA-256 clones and ref-lists
• 9929ece fix!: avoid duplicate network connection when adopting a remote-controlled ha...
• ed998d1 review
• f7d6608 Merge pull request #2664 from ameyypawar/docs/2421-index-write-tree-cache
• 325d2a1 map fetch pack/index checksums to their SHA-256 values
• a507ee0 Merge pull request #2661 from GitoxideLabs/negative-refspec-parsing
• 470e689 derive update_refs expected ids from the fixture (hash-aware)
• Additional commits viewable in compare view

[sidracha]

Version update fails compilation. gix 0.85.0 returns
Result<gix::ObjectId, gix::gix_hash::hasher::Error>

gix 0.68.0 returns
gix::hash::ObjectId

[nasamuffin]

What's your recommendation in that case? Should we not update gix? Should we refactor to adopt this update?

[sidracha]

Refactor to adopt this update. I think its low priority for now, I will write a refactor PR for this after the backend db store functionality is merged in