Hey @lrm29 - curious if this is ready for review? Is there anything that I can help with?

Sorry, yes, I'm using this patch and think I would have heard of problems by now... I'm wary of not having test coverage for this in libgit2 itself (would be nice to be able to mock the server to be able to test the logic, rather than having to add a new test server).

It's clearly unusual to receive empty responses. It seems to be quite particular to a test Git server I've got running in a docker container, and tests against that pass with this patch.

There's more description of the issue in #7013.

Thanks! Wanted to check in because I've been ignoring it since it was marked as draft but it also looks correct and complete. I'll pull it in to a v1.9.1.

Hi, when will this fix be reflected in pygit2?

@ethomson, would be good to get this merged in if it passes review, seems to be more common than I originally thought.

seems to be more common than I originally thought

Specifically, it affects all redirects from gitlab, both self-hosted as well as on gitlab.com.

I can confirm that this PR solves the problem.

EDIT: Specifically gitlab 18 is changing the behaviour to return an empty body, earlier versions had a little piece of HTML as response body

Doesn't seem to reproduce anymore on python 3.12 + pygit2 1.18.2 + gitlab cloud. Is this fix already merged into pygit2?

@nadavhaik gitlab reverted their change and now it provides a non-empty redirect response again

@nadavhaik gitlab reverted their change and now it provides a non-empty redirect response again

That's good to know. It's still just one service though: it doesn't seem beyond the realms of possibility that other services might return empty redirect responses in the future, so it still seems like a worthwhile fix given the impact it would have on projects using libgit2 with such a service.