error on stdio (Sierra)

[RonVanderMeyden]

After an upgrade to OSX Sierra, I see the following ...

c2hs: C header contains errors: /usr/include/stdio.h:133: (column 19) [ERROR] >>> Syntax error ! The symbol `_close' does not fit here.

It seems others have hit this issue as well, as yet unresolved as far as I can see. See https://github.com/tweag/HaskellR/issues/275

[schrammc]

Is there a known workaround for this?

[ian-ross]

Not that I know. Errors like this are usually down to C syntax features unsupported by the language-c package (which is what C2HS uses to parse header files). The best way to get movement on issues like these is probably to make a small test case and submit it as an issue to the language-c maintainers: https://github.com/visq/language-c/issues/

[kolmodin]

@kryoxide The only workaround I know is to define these variables to empty strings. Not sure if it's guaranteed to work since you may end up with invalid C code, but it worked in my project. See https://github.com/grpc/grpc-haskell/issues/6.

[sebeaumont]

This is a show stopper for me on OSX any proper fix in the pipe?

[kolmodin]

@sebeaumont Not that I know. If somebody with a Mac could follow @ian-ross suggestion in https://github.com/haskell/c2hs/issues/176#issuecomment-270351933 that would be a good first step.

[deech]

At least one issue I'm seeing off the bat is the language-c doesn't know how to parse Apple's block syntax, (^ ...), but they are always between __BLOCKS__ so I'm looking into seeing if these can be ignored by c2hs by passing a flag.

[sebeaumont]

Yes I've found a few more things - I'm thinking c-language only supports c90 - since I've full control over headers I'm using to I/f library so I'm taking a defensive approach and working around these things now. Once I've worked out the marshalling I will probably go back to using vanilla Haskell FFI and doing it by hand. In the meantime I'll enforce some c90 tests on my headers and take it from there.

[deech]

Ok I've found the two issues, and have been able to work around them. They are:

• Apple's block syntax (^ ...)
• Use of the _Nonnull keyword in headers, eg.

  void    *bsearch(const void *__key, const void *__base, size_t __nel,
          size_t __width, int (* _Nonnull __compar)(const void *, const void *));

I successfully built my C2HS project by adding the following to the .cabal file:

    if os(darwin)
       cc-options: -U__BLOCKS__ -D_Nonnull=

[ian-ross]

@deech That preprocessor undefine to deal with the blocks thing should already be there (https://github.com/haskell/c2hs/blob/master/src/Main.hs#L580). That was one of the few OS X specific things I managed to find an easy workaround for...

[deech]

Should be fixed in 0.28.2.

[sebeaumont]

Thanks very much. Currently I've sidestepped these issues so as to proceed with 0.28.2 I can go back and undo my workarounds.