Need help with fixing the `clang-format` CI?

[hartwork]

Hi!

I noticed that the clang-format CI has been red for a long time and continues to make new pull requests red. I was wondering what you know about the current state, if it's just a version mis-match or if mass-reapplication would be needed etc. I would like to understand if anything makes the fix a hard problem and if maybe I can help some way: I consider continuing to enforce auto-formating the right path. What do you think?

Best, Sebastian

[ColinH]

Thank you for the kind offer, the last time I tried to consistently apply clang-format the problem was that it simply mis-interpreted and therefore wrongly formatted some code pieces, IIRC it was mixing up the two different meanings of &&. And I did not want to spend a lot of time hunting down all the exceptions, there are already a lot of them given that grammars are formatted completely differently for better readability. For me, personally, readability - that might require some exceptions - is more important than strict conformance, at least for this project that I'm doing in my free time. You can give it a try, but I don't see it as a high priority, and I'm not promising that I will use the results. I'm currently working on some more interesting things like a proper expression parser...

[hartwork]

CI seems to be using clang-format 20.x.x (as a binary from DoozyX/clang-format-lint-action@bcb4eb2#diff-5c175e2009ba54cbc9513ad1db48fcc235a6bcedc55913cfa65333c979d5d333). clang-format-22.1.8 is arguably the latest we could use. When I do git ls-files \*.{cpp,hpp} | xargs -r clang-format -i with 22.1.8 twice, the second run produces the same output for me — same with 20.1.8 — so it /superficially) seems stable enough to be a judge in CI. From a quick look at git grep '&&' things look sane, but it was just a quick look. I know that clang-format is not perfect, but I'm happy enough with it (e.g. at https://github.com/libexpat/libexpat/blob/master/.github/workflows/clang-format.yml). I have created PR #393 now.