Richard, If you create a straight C header that's able to call into the C++ library then what you're trying to do will work. This is the standard (though frustrating) approach to using C++ libraries within Haskell. I can send you an example cabal file for building the Haskell code against the C headers and archives if you'd like. There isn't much more to it than what you're already doing though. The key is that the headers have to be pure C and you have to force cabal to build the C code with g++ (since the implementing .c files will necessarily make C++ calls). Thanks, Arjun On Mon, Jan 26, 2015 at 11:29 AM, Donn Cave wrote:

Quoth Richard Lewis , ...

...

There's a specific question here which goes something like: how do I get hsc2hs to compile code that uses STL headers?

For me, that would be the first thing to fix. Use #ifdef __cplusplus, or just reorganize, so the hsc2hs-generated files are just C. You wrote your wrapper library for the C++ part, and if the callers need to be C++ it kind of defeats the purpose.

From there, I would have guessed "extra-libraries: stdc++" would do it, (assuming you have also arranged to get your wrapper library in there), but I'm no cabal expert. I use C++ more or less like you're doing, so there's hope, but I haven't gotten around to the cabal part of the exercise.

Donn _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

Unfortunately, Cabal doesn't support C++ sources directly. So if you plan to use make or something to build the .C files outside of cabal, you're right. But if you plan to throw the .C files into the cabal file, you'll run into an issue with cabal calling gcc instead of g++ on the C++ sources. gcc might do the right thing for you, but my experience is that the result is thorny. The approach I've managed to get working is to have cabal call g++ in place of gcc, which works for both C and C++ sources. Thanks, Arjun On Mon, Jan 26, 2015 at 5:37 PM, Donn Cave wrote:

...

Richard, If you create a straight C header that's able to call into the C++

Quoth Arjun Comar , library

...

then what you're trying to do will work. This is the standard (though frustrating) approach to using C++ libraries within Haskell. I can send you an example cabal file for building the Haskell code against the C headers and archives if you'd like. There isn't much more to it than what you're already doing though. The key is that the headers have to be pure C and you have to force cabal to build the C code with g++ (since the implementing .c files will necessarily make C++ calls).

This is probably an absurdly elementary question, but at this last point, "... force cabal to build the C code with g++ ...", wouldn't this happen automatically, since they'll really be .C, not .c, files?

E.g., in

Blegger.hsc (#include "blegger.h") Blegger_stub.c (C code created by hsc2hs) blegger.h (C include file) blegger.C (extern "C" { #include "blegger.h" })

... _stub.c shouldn't need g++, and .C should get it naturally?

Donn _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

Alexey, It's unfortunately been about a year since I've worked on the project that needed FFI support so I don't recall the precise issue any longer. From talking to people in #haskell, I discovered that it would be necessary to call g++ and not gcc because gcc is unable to locate some headers and libraries that inevitably get included by C++ projects. Thanks, Arjun On Mon, Jan 26, 2015 at 5:49 PM, Alexey Shmalko wrote:

Arjun,

I'm not an expert in writing Haskell binding but pretty much experienced in C/low-level stuff.

You shouldn't force compilation of .c files with g++. Instead, you should keep your wrapping functions as extern "C" and place their implementation in .cpp file (which of course should be compiled with g++). Don't forget to link to C++ runtime library (-lstdc++).

There is also a wiki page[1] that describes how to call C++ from Haskell.

Best regards, Alexey

[1]: https://wiki.haskell.org/CPlusPlusFromHaskell

2015-01-26 23:27 GMT+02:00 Arjun Comar :

...

Richard, If you create a straight C header that's able to call into the C++ library then what you're trying to do will work. This is the standard (though frustrating) approach to using C++ libraries within Haskell. I can send you an example cabal file for building the Haskell code against the C headers and archives if you'd like. There isn't much more to it than what you're already doing though. The key is that the headers have to be pure C and you have to force cabal to build the C code with g++ (since the implementing .c files will necessarily make C++ calls).

Thanks, Arjun

On Mon, Jan 26, 2015 at 11:29 AM, Donn Cave wrote:

...

Quoth Richard Lewis , ...

...

There's a specific question here which goes something like: how do I get hsc2hs to compile code that uses STL headers?

For me, that would be the first thing to fix. Use #ifdef __cplusplus, or just reorganize, so the hsc2hs-generated files are just C. You wrote your wrapper library for the C++ part, and if the callers need to be C++ it kind of defeats the purpose.

From there, I would have guessed "extra-libraries: stdc++" would do it, (assuming you have also arranged to get your wrapper library in there), but I'm no cabal expert. I use C++ more or less like you're doing, so there's hope, but I haven't gotten around to the cabal part of the exercise.

Donn _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

Donn, Arjun, Alexey, Ian, Thanks, all, for your suggestions. At Mon, 26 Jan 2015 08:29:24 -0800, Donn Cave wrote:

Quoth Richard Lewis , ...

...

There's a specific question here which goes something like: how do I get hsc2hs to compile code that uses STL headers?

For me, that would be the first thing to fix. Use #ifdef __cplusplus, or just reorganize, so the hsc2hs-generated files are just C. You wrote your wrapper library for the C++ part, and if the callers need to be C++ it kind of defeats the purpose.

Yes, so it turns out the problem was that my wrapper .h was including one of the headers of the third party library which, in turn, was including vector. But I've managed to arrange it so that I don't need to include that third party header anymore so gcc can compile my translation unit without problems.

From there, I would have guessed "extra-libraries: stdc++" would do it, (assuming you have also arranged to get your wrapper library in there), but I'm no cabal expert. I use C++ more or less like you're doing, so there's hope, but I haven't gotten around to the cabal part of the exercise.

The challenge with Cabal now is to get it to link my .a library with the libHSMyLibrary.a that GHC creates. It seems, from what I've read [1], that there is no Cabal hook into the linker so it may involve quite a bit of work in the Setup.hs file. I was just following up a suggestion [2] that (I think), if the .a file Cabal has created for my wrapper static library is in the same directory as, um, not sure what--output libHSMyLibrary.a? gcc working directory?--and if its name is of the form "libxxx.a", then it may be possible to add -lmylib to gcc's options, and gcc will pick it up. Anyway, I altered my Setup.hs to do this, putting libmylib.a in dist/build and dynamically adding /path/to/dist/build to the "extra-libraries" option, but nm still reports that the resulting libHSMyLibrary.a does not contain libmylib.a (or would it be mylib.o?). This might be a discussion for a new thread, however. The immediate problem of how to compile the C wrapper is now fixed. Richard [1] http://stackoverflow.com/questions/18543228/statically-linking-a-c-library-w... [2] https://www.haskell.org/pipermail/haskell-cafe/2006-June/016022.html