@haampie @alalazo
Well, I was going to update the checksums in #35662 or #35648 (now
reverted), but whatever.

So, does this mean that we should never use URL patches from gitlab?

And in the future, I would prefer you not to submit, approve and merge
patches to hpctoolkit all while I (as maintainer) am asleep.

In this case, the better fix is to delete the 411 patch and replace it
with a conflict as I did in #35662. But it's short and I'll fix it in
#35662.

And in the future, I would prefer you not to submit, approve and merge patches to hpctoolkit all while I (as maintainer) am asleep.

@mwkrentel Apologies, that seemed a trivial fix for something that was reported broken since last Wednesday - so I went ahead and merged as requested by @haampie I'll keep that in mind that you prefer to review all changes to hpctoolkit in the future. Also, let me know if you have other preferred solutions, I'll try to prioritize reviewing them.

The other thing to note is that, since hpctoolkit is in pipelines, failure to verify the checksum of patches means failure to pass pipelines tests.

@alalazo I mainly wanted a chance to discuss updating the checksums
vs copying the patch. Does this mean that we should never use a URL
patch from gitlab?

I mainly wanted a chance to discuss updating the checksums vs copying the patch

The chance is still there, and we can revert and go back to a URL if we find a way to have a stable shasum. This was mainly to avoid failures in pipelines due to a changing shasum. I didn't check yet if Gitlab has something similar to ?full_index=1 for GitHub

When it comes to code and commits, I'm like John Roberts (Chief
Justice, US Supreme Court). I prefer to make the smallest change that
gets the job done, unless there's a compelling reason to do otherwise.

In this case, if it were me, I would have just updated the sha256 sums
as the smaller change. But now that they're switched to actual
patches, I prefer to leave it as is. I'm going to delete one of them
in a couple days anyway.

I'm sure GitLab will at some point produce a slightly different diff
with one whitespace or line of context more ore less, etc. It'd be
better if we as a general rule would not allow any dynamically
generated content in source and patch urls.