test: add tests for symlinks in gix-blame

[Christoph Rüßler (cruessler)]

This PR adds dedicated tests for symlinks in gix-blame. None of the tests required any code changes to make them pass. This is based on a TODO mentioned in this comment.

I think the TODO comment can potentially be removed. We might also want to change it to highlight the fact that we force set_resource to treat symlinks as blobs by unconditionally passing EntryKind::Blob. As far as I can tell at least, set_resource does not check on its own whether an object is a symlink, but trusts its caller.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 7047944e4d

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "Codex (@codex) review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "Codex (@codex) address that feedback".

[Sebastian Thiel (Byron)]

Thanks a lot! It's good to see that this works, but also still a little bit surprising.

So in the interest of time, I asked Codex, so maybe there is something you can additionally poke at. In the worst case, it's more coverage. That's also why I left the comment for now.

---

Is there anything that you can imagine that won't work with @function.rs (767:767)?

TreeDiffChange discards both entry modes, then blob_changes() labels both resources as regular blobs. Git applies user diff drivers only to regular files, not symlinks. Consequently, gix may treat the link target as binary or converted data and assign blame differently.

The best additional regression test would be:

• Mark one changing symlink path as -diff.
• Create it pointing to one target.
• Change its stored target.
• Compare blame with Git.

File↔symlink transitions under such attributes should also be tested because each side requires a different mode. The proper eventual fix is to preserve source/destination modes in TreeDiffChange and pass their actual EntryKind into blob_changes().

[Christoph Rüßler (cruessler)]

Thanks for the context! I’ve put it onto my list of things that I want to have a look at. 😀