Boolean struct members are incorrectly marshalled about c2hs HOT 3 CLOSED

Hi,
I unfortunately was not notified of any issues or PRs until today when someone directly @'ed me. In any case, we should at least be able to test the size of a bool locally and generate either a CInt or CChar. The only concern is that will break codebases that now depend on a bool marshalling to CInt.

I agree that using CBool is the cleanest solution but I'm not sure how many people are still using c2hs on GHC's older than 2017.

Any thoughts?

from c2hs.

Hi deech,
Oh no, I'm glad to hear there was only a notifications issue but sorry you have to deal with a backlog.

I think the right approach for C2HS is to maintain backwards compatibility as much as possible. After looking at the available C types in Foreign.C.Types again, using CUChar (rather than CChar as I originally suggested) for a one-byte representation of C's bool and selecting which C type to use based on the bool size test seems like a good plan.

There should be no functional difference between CUChar and CBool as both are newtypes wrapping a Word8 and we would still use toBool to convert to Haskell's Bool. A quick peek at the CBool definitions for various classes like Eq seems to show they are derived from Word8. I'm not entirely comfortable reading that code quickly, so it wouldn't hurt to double check.

from c2hs.

Hi @deech, I ran into this problem again. Were you ok with the solution I proposed earlier? I'd like to fix it instead of writing workarounds in my own code.

from c2hs.