Sadly that version doesn't compile in unix ... at least not without compiling *all* of ctypes (which is a lot for just libffi)

Is that make file issues, or the technical aspects of the code? If its just makefile stuff, I am sure our expert scons person will sort it out with a scons script. The windows copy can be built rather easily on its own. Of course, using two different copies of libffi for the different platforms is just asking for trouble... Thanks Neil

On 6/7/06, Tom Shackell wrote:

No I believe it's just makefile issues, it seems the ctypes people stripped the standard makefile out of libffi ...

Then it's no problem at all, I've already sconsified cpphs so I'll just sconify ctypes as well :-) Cheers guys, Andrew

Oh the other problem with the ctypes version of libffi is that it has no AMD64 port, whereas the redhat one does. Perhaps a merging of the two would be best?

And perhaps the best people to do this are the ctypes people :) Does someone want to prod their mailing list. I noticed that GHC is thinking of moving to libffi, so we should figure out the best ffilib for Haskell compilers and share our knowledge with them Thanks Neil

On 6/8/06, Malcolm Wallace wrote:

"Andrew Wilkinson" wrote:

...

I've already sconsified cpphs ...

Is it sensible to push a patch for this into the cpphs repository? Or does it really only apply to building yhc?

The scons script I've written doesn't actually build the cpphs binary, it just grabs the source files which are then linked into the yhc binary. I don't think you can have a scons script that can be used both as a standalone script and called as part of another build process. If you wanted me to properly sconsify it then I could, but I'd still need the special YHC build script for it. Andrew