gix: negotiate and adopt sha256 object format by 10ne1 · Pull Request #2642 · GitoxideLabs/gitoxide

Adrian Ratiu (10ne1) · 2026-06-11T17:40:58Z

What

This teaches the v2 protocol fetch client to negotiate object-format with sha256 remotes and to adopt the remote object format when cloning. Before this, gix assumed sha1 during protocol negotiation and hit an unimplemented panic when a clone target turned out to use sha256.

This addresses the gix-protocol line item of Batch 3 in #281:

Accept and preserve SHA256 object-format negotiation end to end.

Testing

These tests were suggested by Sebastian Thiel (@Byron), so I also added their equivalent to the justfile:

GIX_TEST_FIXTURE_HASH=sha256 cargo nextest run -p gix --features async-network-client --no-fail-fast
GIX_TEST_FIXTURE_HASH=sha256 cargo nextest run -p gix --features blocking-network-client --no-fail-fast

AI disclosure
This PR was written with the help of Claude Code, but was not vibe-coded. I'm not a fan of vibe-coding. I used AI to better understand the codebase myself (asked it a lot of questions), review both my code and the code generated by AI, used it especially to detect bugs and corner-cases and suggest fixes, however I very carefully reviewed & edited each code/comment lines, the commit messages and so on, until I was satisfied with the result.

Byron's Tasks

initial refackiew
see if Attempt can be removed, as it does affect callers apparently (*progress).
journey tests maybe?

Copilot

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

This PR adds SHA-256 object-format awareness to fetch/clone flows (notably protocol v2), including automatically adopting a remote’s object format during a fresh clone, and extends the test suite/fixtures to cover SHA-256 scenarios.

Changes:

Echo object-format in protocol v2 commands and parse server-advertised object-format into the fetch ref-map.
Teach clone::fetch_only to adopt a remote’s object hash (SHA-256) during initial clone by reinitializing/reopening the empty repo and retrying.
Add fixtures and tests (including nextest matrix updates) to validate SHA-256 behavior across protocol/refmap/clone.

Reviewed changes

Copilot reviewed 12 out of 16 changed files in this pull request and generated 7 comments.

Show a summary per file

File	Description
justfile	Runs `gix-protocol` tests with additional `sha256` feature combinations.
gix/tests/gix/util.rs	Extends SHA1↔SHA256 fixture hash mapping for tests.
gix/tests/gix/clone.rs	Stabilizes shallow ID comparisons; adds SHA-256 clone adoption test.
gix/tests/fixtures/make_sha256_remote.sh	Adds a deterministic SHA-256 remote fixture generator.
gix/src/remote/connection/fetch/update_refs/tests.rs	Makes tests hash-kind agnostic; fixes NULL handling for SHA-256.
gix/src/clone/fetch/util.rs	Adds repo reinitialization helper to adopt remote object format.
gix/src/clone/fetch/mod.rs	Implements adopt-and-retry on object-format mismatch during `fetch_only`.
gix-worktree-stream/tests/fixtures/generated-archives/.gitignore	Ignores a new SHA-256 generated fixture archive.
gix-protocol/tests/protocol/fetch/v2.rs	Updates v2 request expectations; adds SHA-256 v2 test.
gix-protocol/tests/protocol/fetch/ref_map.rs	Adds unit tests for `object-format` capability parsing.
gix-protocol/src/fetch/refmap/init.rs	Parses `object-format` into `gix_hash::Kind` and uses hash-specific null IDs.
gix-protocol/src/command.rs	Echoes server `object-format` in all v2 commands; validates it as an allowed feature.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot

Pull request overview

Copilot reviewed 12 out of 16 changed files in this pull request and generated 6 comments.

Adrian Ratiu (10ne1) · 2026-06-18T17:17:58Z

I reworded the commit message subject prefixes to follow the Gitoxide conventions (added fix: to a few commits). No code changes.

RefMap::from_refs only recognized sha1, so it refused to talk to sha256 remotes even with the sha256 feature on. Parse the server's object-format instead, accepting any hash the build supports. When the server omits the capability, default to Sha1 if available, otherwise report UnknownObjectFormat. This also replaces a hardcoded `Kind::Sha1` null placeholder, which did not compile in sha256-only builds, with the resolved hash. Tests cover the sha1, sha256, unknown, and missing cases. Co-Authored-By: Claude Opus 4.8 <[email protected]> Signed-off-by: Adrian Ratiu <[email protected]>

A v2 client must report its object format to the server, otherwise the server assumes SHA-1 and aborts the request with a "mismatched object format" error when talking to a SHA-256 repository. This affects both ls-refs and fetch. A persistent connection retains the object format negotiated during ls-refs, but a stateless transport like HTTP sends each command as its own request, so every v2 command must echo the object-format the server advertised. So add the echo to `Command::default_features` so it applies to every v2 command and permit object-format during v2 feature validation. Also add v2 wire-format sha256 tests for ls-refs and ref-in-want. Co-Authored-By: Claude Opus 4.8 <[email protected]> Signed-off-by: Adrian Ratiu <[email protected]>

A clone into a freshly initialized repository hit an `unimplemented!` panic when the remote used sha256, since the local repository defaults to sha1. Reconfigure the still-empty local repository to the remote's object format and retry the fetch, matching git's behavior of inheriting the remote's hash on clone. Without the sha256 feature gix_hash::Kind has a single variant, so the local and remote hashes can never differ; the mismatch check is compiled out entirely in that case. Co-Authored-By: Claude Fable 5 <[email protected]> Signed-off-by: Adrian Ratiu <[email protected]>

The clone fetch loop adopts the remote's object format and retries when it differs from the freshly initialized local repository. Adoption makes the next iteration's hashes match, but if the reopened repository somehow still disagrees the loop would re-handshake forever. Track whether we already retried and fail with IncompatibleObjectHash on a second mismatch instead. The error variant is introduced here, as this is its only user. Also silence clippy::never_loop on default (sha1) features, where the sha256-gated `continue` is absent and the loop always returns first pass. Co-Authored-By: Claude Fable 5 <[email protected]> Signed-off-by: Adrian Ratiu <[email protected]>

Add a dedicated test for the clone-time object-format adoption. Basically a fixture remote created with --object-format=sha256 independent of GIX_TEST_FIXTURE_HASH gets cloned into a fresh SHA-1-default repository which must adopt the remote format. The test asserts the adopted format both in memory and as persisted on disk, and that the remote is written exactly once despite the adoption retry. Co-Authored-By: Claude Opus 4.8 <[email protected]> Signed-off-by: Adrian Ratiu <[email protected]>

Derive expected object ids from the fixture repo instead of hardcoding SHA-1 hex, so the update_refs tests pass under both SHA-1 and SHA-256 fixtures (GIX_TEST_FIXTURE_HASH). Co-Authored-By: Claude Opus 4.7 <[email protected]> Signed-off-by: Adrian Ratiu <[email protected]>

Copilot

Pull request overview

Copilot reviewed 12 out of 16 changed files in this pull request and generated 2 comments.

The fetch_pack tests assert on pack data and index checksums, which are hash-algorithm dependent. Add the SHA-256 equivalents so the tests pass under GIX_TEST_FIXTURE_HASH=sha256. Co-Authored-By: Claude Opus 4.8 <[email protected]> Signed-off-by: Adrian Ratiu <[email protected]> compare shallow boundary commits order-independently The on-disk order of shallow boundary commits is hash-dependent. This means that inside a hash alorithm the commit order is deterministic and it reshuffless across hashes. E.g. a commit which sorts 1st under sha1 might sort 3rd under sha256. This is the reason why the clone shallow/unshallow tests fail under sha256 when comparing against the sha1-ordered expectations, due to the hex_to_id() calls preserving the sha1-ordering even when converting the hashes to sha256. Thus we sort the expected side before comparing. Note: only the expected side needs sorting, as the actual side already comes sorted by gix_shallow::read. Co-Authored-By: Claude Opus 4.8 <[email protected]> Signed-off-by: Adrian Ratiu <[email protected]> add generated sha256 archive for symlink-prefix-reuse fixture Adds the SHA-256 variant of the make_symlink_prefix_reuse_advisory fixture archive, matching its committed SHA-1 sibling. Contains no host-absolute paths, so it is safe to track. Co-Authored-By: Claude Opus 4.7 <[email protected]> Signed-off-by: Adrian Ratiu <[email protected]> do not track the sha256 basic fixture archive This mirrors the existing basic.tar file exclusion. Co-Authored-By: Claude Opus 4.8 <[email protected]> Signed-off-by: Adrian Ratiu <[email protected]>

Sebastian Thiel (Byron) · 2026-06-21T16:05:41Z

Thanks!

I couldn't believe my eyes when I couldn't find these lines in the justfile, I genuinely thought I accidentally removed them.

GIX_TEST_FIXTURE_HASH=sha256 cargo nextest run -p gix --features async-network-client --no-fail-fast
GIX_TEST_FIXTURE_HASH=sha256 cargo nextest run -p gix --features blocking-network-client --no-fail-fast

After all, they are needed to run the changes introduced to gix.

Adrian Ratiu (10ne1) · 2026-06-21T17:18:29Z

I couldn't believe my eyes when I couldn't find these lines in the justfile, I genuinely thought I accidentally removed them.
GIX_TEST_FIXTURE_HASH=sha256 cargo nextest run -p gix --features async-network-client --no-fail-fast
GIX_TEST_FIXTURE_HASH=sha256 cargo nextest run -p gix --features blocking-network-client --no-fail-fast
After all, they are needed to run the changes introduced to gix.

Sebastian Thiel (@Byron) Yes, I added them in this commit (right at the end): 9067bcb

Perhaps I should have explicitly added them in a separate commit. 🙂

Edit: They seem to be present in your latest push: 0ed353f

Sebastian Thiel (Byron) · 2026-06-21T22:32:38Z

Adrian Ratiu (@10ne1) 9067bcb just adds them for gix-protocol, gix was missing. The review commit (which somehow has an extra file curtesy of my editor) adds these test-runs.

In any case, I'd expect this to PR to merge soon.

Copilot

Pull request overview

Copilot reviewed 12 out of 21 changed files in this pull request and generated 1 comment.

Copilot

Pull request overview

Copilot reviewed 12 out of 21 changed files in this pull request and generated 1 comment.

The major change is to actually run the `gix` tests for the affected code. Co-authored-by: Sebastian Thiel <[email protected]>

Copilot

Pull request overview

Copilot reviewed 23 out of 32 changed files in this pull request and generated 6 comments.

…olled hash kind. Connections can now be created from detached remotes, internally, which makes them independent of the underlying repository. This is breaking as it comes with a cleanup of `Connection` lifetimes, which will break anyone who stores them in a struct. So hopefully nobody is affected. Co-authored-by: Sebastian Thiel <[email protected]> address auto-review

Co-authored-by: Sebastian Thiel <[email protected]>

Sebastian Thiel (Byron)

Alright, that's it 🎉!

The Attempt workaround was removed in favor of the refactoring that was needed to be able to change/reopen the repository in place. In future, that's even more important as it will also have to deal with reftable.
Due to not using Attempt, it will be avoiding to redo the entire fetch operation.
And as a side-effect, a long-standing Connection lifetime bug was also fixed.

Journey tests were added as well, proving that SHA-256 transport and protocol parts now work. I also validated that manually, awing at the SHA-256 references provided by the remote.

Will merge soon.

Copilot

Pull request overview

Copilot reviewed 24 out of 33 changed files in this pull request and generated 2 comments.

Copilot AI review requested due to automatic review settings June 11, 2026 17:40

Copilot AI reviewed Jun 11, 2026

View reviewed changes

Adrian Ratiu (10ne1) force-pushed the dev/aratiu/sha256-transport branch from 51c5f0e to 73c1cdf Compare June 12, 2026 16:46

Copilot AI review requested due to automatic review settings June 12, 2026 17:07

Adrian Ratiu (10ne1) force-pushed the dev/aratiu/sha256-transport branch from 73c1cdf to 8cb8290 Compare June 12, 2026 17:07

Copilot AI reviewed Jun 12, 2026

View reviewed changes

Adrian Ratiu (10ne1) mentioned this pull request Jun 16, 2026

gix-transport: add SHA-256 object-format negotiation test coverage #2647

Closed

Adrian Ratiu (10ne1) force-pushed the dev/aratiu/sha256-transport branch from 8cb8290 to 9758eff Compare June 18, 2026 17:16

Adrian Ratiu (10ne1) and others added 6 commits June 21, 2026 19:22

Sebastian Thiel (Byron) force-pushed the dev/aratiu/sha256-transport branch from 9758eff to b439eaa Compare June 21, 2026 11:27

Copilot AI review requested due to automatic review settings June 21, 2026 11:27

Copilot started reviewing on behalf of Sebastian Thiel (Byron) June 21, 2026 11:27 View session

Copilot AI reviewed Jun 21, 2026

View reviewed changes

Comment thread gix-protocol/tests/protocol/fetch/v2.rs

Comment thread gix/src/clone/fetch/util.rs Outdated

Sebastian Thiel (Byron) force-pushed the dev/aratiu/sha256-transport branch from b439eaa to 0ed353f Compare June 21, 2026 16:03

Sebastian Thiel (Byron) force-pushed the dev/aratiu/sha256-transport branch from 0ed353f to e9445c7 Compare June 21, 2026 23:26

Copilot AI review requested due to automatic review settings June 21, 2026 23:26

Sebastian Thiel (Byron) self-assigned this Jun 21, 2026

Copilot started reviewing on behalf of Sebastian Thiel (Byron) June 21, 2026 23:27 View session

Sebastian Thiel (Byron) force-pushed the dev/aratiu/sha256-transport branch from e9445c7 to 3262d99 Compare June 21, 2026 23:31

Copilot AI reviewed Jun 21, 2026

View reviewed changes

Comment thread justfile Outdated

Copilot AI review requested due to automatic review settings June 21, 2026 23:34

Sebastian Thiel (Byron) force-pushed the dev/aratiu/sha256-transport branch from 3262d99 to ea0f627 Compare June 21, 2026 23:34

Copilot started reviewing on behalf of Sebastian Thiel (Byron) June 21, 2026 23:34 View session

Copilot AI reviewed Jun 21, 2026

View reviewed changes

Comment thread gix/src/clone/fetch/util.rs

review

ed998d1

The major change is to actually run the `gix` tests for the affected code. Co-authored-by: Sebastian Thiel <[email protected]>

Sebastian Thiel (Byron) force-pushed the dev/aratiu/sha256-transport branch 2 times, most recently from b866ed3 to c6789d3 Compare June 22, 2026 06:49

Copilot AI review requested due to automatic review settings June 22, 2026 06:49

Copilot started reviewing on behalf of Sebastian Thiel (Byron) June 22, 2026 06:49 View session

Copilot AI reviewed Jun 22, 2026

View reviewed changes

Comment thread gix/src/clone/fetch/mod.rs

Comment thread gix/src/clone/fetch/mod.rs Outdated

Comment thread justfile

Comment thread tests/fixtures/generated-archives/.gitignore

Comment thread gix/tests/fixtures/generated-archives/.gitignore

Comment thread gix/src/clone/fetch/util.rs

Sebastian Thiel (Byron) force-pushed the dev/aratiu/sha256-transport branch from c6789d3 to 2bf6a59 Compare June 22, 2026 07:22

Sebastian Thiel (Byron) and others added 2 commits June 22, 2026 15:42

Add journey-tests for SHA-256 clones and ref-lists

3927d70

Co-authored-by: Sebastian Thiel <[email protected]>

Copilot AI review requested due to automatic review settings June 22, 2026 07:42

Sebastian Thiel (Byron) force-pushed the dev/aratiu/sha256-transport branch from 2bf6a59 to 3927d70 Compare June 22, 2026 07:42

Copilot started reviewing on behalf of Sebastian Thiel (Byron) June 22, 2026 07:43 View session

Sebastian Thiel (Byron) approved these changes Jun 22, 2026

View reviewed changes

Sebastian Thiel (Byron) enabled auto-merge June 22, 2026 07:45

Copilot AI reviewed Jun 22, 2026

View reviewed changes

Comment thread gix-protocol/tests/protocol/fetch/ref_map.rs

Comment thread gix/src/remote/connection/mod.rs

Sebastian Thiel (Byron) merged commit da6b267 into GitoxideLabs:main Jun 22, 2026
32 checks passed

Adrian Ratiu (10ne1) deleted the dev/aratiu/sha256-transport branch June 25, 2026 13:15

Uh oh!

Uh oh!

Conversation

Adrian Ratiu (10ne1) commented Jun 11, 2026 • edited by Byron Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Byron's Tasks

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Adrian Ratiu (10ne1) commented Jun 18, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Sebastian Thiel (Byron) commented Jun 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Adrian Ratiu (10ne1) commented Jun 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Sebastian Thiel (Byron) commented Jun 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Sebastian Thiel (Byron) left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Adrian Ratiu (10ne1) commented Jun 11, 2026 •

edited by Byron

Loading

Sebastian Thiel (Byron) commented Jun 21, 2026 •

edited

Loading

Adrian Ratiu (10ne1) commented Jun 21, 2026 •

edited

Loading

Sebastian Thiel (Byron) commented Jun 21, 2026 •

edited

Loading