Proposal: Add a pre-commit hook configuration

[Scheremo]

Hi everyone,

I’ve noticed that some of the .td (TableGen) files in the repository contain trailing whitespace and minor formatting inconsistencies. To help avoid these and others in the future, and to make contributor workflows a bit smoother, I’d like to propose adding a pre-commit configuration to the repo.

A pre-commit setup would:

• Automatically strip trailing whitespace and fix end-of-file newlines
• Enforce consistent formatting by using the same fixed clang-format version used in CI
• Make it easier for contributors to catch common issues locally before committing

This would not affect build behavior and can be gradually adopted.
If people are open to the idea, I can open a PR with a minimal .pre-commit-config.yaml and documentation on how to enable it.

WDYT?

[seldridge]

I'd like to avoid pulling in new dependencies (even if they are purely up to the user to use/instal) that don't align with upstream. We haven't continued the discussion about CIRCT going upstream in several years. However, I don't want to add any potential divergence with upstream.

Upstream does provide some hooks here that seem to be both (1) what is used by upstream CI and (2) can be installed and used as git hooks by users: https://github.com/llvm/llvm-project/tree/9ab16778c6e37e3d34e28b990bdb8839e5961805/llvm/utils/git Can we re-use these as they exist (given that we have LLVM source submoduled already)? Barring that, I'd say that we should align with this rough structure for any hooks.

Would this work?

Going further, it would be could align with what upstream is doing here, even if they are our own scripts. Even further would be to then use these for CI. (GitHub Actions would then call these scripts, e.g.)

[Scheremo]

I think we can start from the hooks available upstream, they seem to largely fulfill the same purpose - I'll give it a go!