This sounds reasonable. The CI failure looks to be due to the generated files not being updated. Could you run the gen program and update this PR?

Another reason a lot of things are bunched in the general module is that Linux headers often pull in other Linux headers, so it's difficult to get clearly scoped modules that just have headers for one purpose and not also all their transitive header dependencies. Because if every module has its dependencies, we'd end up with a lot of duplication. I imagine one way forward is to make more use of bindgen's ability to filter out symbols by things like prefix names. That won't always work (as in the AT_ case here), but could help in many cases.

This sounds reasonable. The CI failure looks to be due to the generated files not being updated. Could you run the gen program and update this PR?

I did try but I must be doing something wrong or there is some inconsitencies with the generation because I tried multiple times and I always get the same output. I am not sure why the ci is generating something different.

Another reason a lot of things are bunched in the general module is that Linux headers often pull in other Linux headers, so it's difficult to get clearly scoped modules that just have headers for one purpose and not also all their transitive header dependencies. Because if every module has its dependencies, we'd end up with a lot of duplication. I imagine one way forward is to make more use of bindgen's ability to filter out symbols by things like prefix names. That won't always work (as in the AT_ case here), but could help in many cases.

Contributing a few times to this project made me realize we could make much better use of bindgen features and refactor the codegen a lot. I was planning to work on something of that nature but I am still unsure on what the best approach is.

I rebased to main and regenerated the bindings. Hopefully it should work now. It also seems I have caught some stray changes from a previous pr that wasn't correctly updated

The bindgen output seems to be sensitive to the version of clang installed. The CI is using clang-22.

On further consideration, I think this makes enough sense that I'm just going to merge it and regenerate the files separately.

This is now released in linux-raw-sys 0.11.0.