[Openmcl-devel] h-to-ffi giving kooky results

[Gary Byers]

When I run darwin-headers/libc/C/populate.sh on a 10.3.9, I get a single
warning (the scalb() function in /usr/include/math.h has a prototype that
conflicts with a GCC 4.0 compiler intrinsic.)  I'm using the FFIGEN package
from

<ftp://clozure.com/pub/testing/ffigen-bin-darwinppc-gcc-4.0.0-2005-05-17-11-45-56.tar.gz>


The older (MUCH older) version of FFIGEN that's on the server is based
on GCC 2.95, and it expects that the installed "gcc" driver is GCC
2.95's (and "gcc -E" invokes gcc 2.95's preprocessor).  If the
installed "gcc" driver belongs to GCC 3.x or later (likely, over the
last few years), I believe that that can lead to errors like those
that you're getting.  (I'm actually surprised that it would work at
all.)  There's a footnote in section 9.7 of the documentation that
explains this issue; as of Tiger (or XCode 2.x), I don't believe that
the suggested workaround would work.


On Tue, 6 Sep 2005, Hamilton Link wrote:

> OK, for a very very simple file with just #defines in it, h-to-ffi.sh is 
> working.
>
> With anything that includes stdio or basically anything of substance, I get a 
> mountain of:
>
> /usr/include/ppc/endian.h:0: warning: unrecognized text at end of #line
> /usr/include/ppc/endian.h:22: warning: unrecognized text at end of #line
> /usr/include/ppc/endian.h:56: warning: unrecognized text at end of #line
>
> for file after file after file, and line after line after line.
>
> I couldn't get past this for gandalf, and as it turns out when I run 
> darwin-headers/libc/C/populate.sh I get the same thing.
>
> This is a preprocessor warning of some sort that means... what?  That I'm 
> using the wrong h-to-ffi.sh?  Or the wrong compiler? I googled this and saw 
> other people with basically the same problem, but those guys were mostly 
> trying to recompile linux and gcc and things like that, generally with other 
> incompatible versions of things.
>
> I have the gandalf bundle built, no problemo, and can open it using UFFI, but 
> I don't have reader macros working yet because of this.  Anyone with clever 
> ideas out there?  I'm out in deep water with this one, I'm afraid.
>
> thanks,
> hamilton
>
> _______________________________________________
> Openmcl-devel mailing list
> Openmcl-devel at clozure.com
> http://clozure.com/mailman/listinfo/openmcl-devel
>
>