[MLton] Windows port of MLton using the Microsoft tools (e.g. without MinGW)

[Nicolas Bertolotti]

I'm sorry I can't tell you much more as I don't remember exactly what I did.

When you build the package using MinGW, you can see that some assembly files
are built. The assembly files are produced from some templates that are
processed using 'm4' which generates files that can only be compiled using
'gas'. The autoconf scripts also generate some symbolic links and compile
the files that are pointed by those links.

When I got the results from the MinGW compilation, I processed all this
using a pragmatic approach, that is refer to the original source file when
the compiled file was a symbolic link and try to find the corresponding
generic C file (when there was one) when the file was an assembly file.

It's definitely not difficult to do but I'm afraid I can't provide an
algorithm that would make it possible to automate the process.

> -----Original Message-----
> From: Peter Tanski [mailto:p.tanski at gmail.com]
> Sent: Thursday, July 26, 2007 6:04 PM
> To: Nicolas Bertolotti
> Cc: 'skaller'; mlton at mlton.org
> Subject: Re: [MLton] Windows port of MLton using the Microsoft tools (e.g.
> without MinGW)
> 
> On Jul 26, 2007, at 10:43 AM, Nicolas Bertolotti wrote:
> > In order to do it fast, I first built the library for MinGW using the
> > automake system.
> >
> > Then, I used the build log in order to write a Makefile that does
> > the same
> > thing but uses cl.exe instead of gcc.
> >
> > Finally, I had to fix some instances of the "long long" type and a few
> > things in the config.h file in order to use the generic C version
> > of some
> > functions rather than the assembly code version.
> >
> > You'll find a patch for gmp-4.2.1 in attachement.
> 
> Nice work!  Thanks!  The generic files you are using are the ones
> listed after the .c$(OBJ): rule in the Makefile?  What were the
> problems with the assembler files?
> 
> Cheers,
> Pete
>