c2hs fails to produce the correct alignment about c2hs

[deech]

I started a small repo to capture and run alignment and sizeof test cases, feel free to fork it and add more. Once the feature is complete we can port them back to this repo.

[deech]

This is happening because currently c2hs ignores all aligned (and packed) GNU specific attributes. In this particular case 1 is accidentally correct because 384 is a multiple of 64 and so the field will naturally align at a 64 bit boundary. I'm not sure when I'll have time to add support for them.

[Kleidukos]

@deech If you have an idea of how to do it maybe I can take a shot at it?

[deech]

Yes absolutely, the addition would start where we calculate struct and struct field alignments and sizes and affect calculations downstream of that. The GNU specific packed and aligned attributes are passed along in a CAttr, you can read more about them in the GCC manual. I can even walk you through the code that currently exists if you like. Brace yourself, this is a relatively involved change. 😄

The other option is we simply don't support alignment attributes in c2hs and actually error in their presence although I would suspect this would break a bunch of code in the wild.

[deech]

Yet another option is to generate a C file that includes the header and print the values we need using the builtin sizeof, alignof and offsetof. We do this currently to get the size of a bool. Honestly if we can reliably generate the file this might be the best approach.

[deech]

@Kleidukos Are you still interested in working on this issue? I'm happy to help you with it.

[Kleidukos]

@deech I would love some help on that one. :)

[deech]

Cool! Send an email to the last maintainer address listed on the Hackage page and we can set up a time to jump on a call and we can talk through options.

[Kleidukos]

@kozross ☝️

[deech]

Bump

[kozross]

@deech @Kleidukos and I are looking at the example and working out a way forward: thanks for that!

[deech]

@Kleidukos @kozross Are you still planning on working this issue? Is there anything I can do to help?

[Kleidukos]

@deech We had to pause a bit this last month but I'd like to work on it again. :) We'll ping you it ever is blocked by something big. In the meantime if someone submits a PR with a fix, don't wait for us!

[Kleidukos]

@deech Coming back to say that it is going to be quite hard for me to bring this to completion. If you ever feel like fixing it, do feel free. :) Thank you for your support in any case!