GeezerVison

@JelleZijlstra Seems like the same thing is happening here. Funny enough, the lint job succeeded and the build job failed here, versus the opposite happened on #2228 despite being re-run at the same time, so seems pretty random. If you restart the failing workflows, seems like there's a chance it succeeds.

It worked :)

But the deploy job failed: https://github.com/python/peps/actions/runs/1685665720. Let's just give GH some time to fix things and maybe rerun it later.

Looks like it has been resolved. Could you re-run?

And, the re-run seems to have succeeded!

Thanks for catching this Cam/all -- docutils can be annoyingly fickle sometimes.

As a post-mortem, what we should have done initially is made all target children of the references section children of the document.

Deferring the transform is probably the smallest change to fix it though, and we don't rely on any behaviour transforming the footer links, so all good. Thanks again!

A

@AA-Turner Yup, great point—decreasing the priority to an arbitrarily low value felt kinda hacky, and I thought about just moving the link targets somewhere else, but given it didn't appear to be necessary, would be a much bigger change and I didn't want to mess that up, plus I wanted to get the important fix out as quickly as practical, so I ended up going with it (I also initially thought it only affected the legacy build system, since I didn't realize I was looking at a stale rendered file, so I wasn't as concerned with implementing a hacky solution).