From: "Mark E." <snowball3 AT bigfoot DOT com>
To: GCC Patches <gcc-patches AT gcc DOT gnu DOT org>, djgpp-workers AT delorie DOT com
Date: Tue, 16 May 2000 00:34:26 -0400
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
Subject: Re: DJGPP patch for float_format 
Message-ID: <39209792.27559.3E5FC5@localhost>
References: <200005152236 DOT RAA05926 AT hp2 DOT xraylith DOT wisc DOT edu>
X-mailer: Pegasus Mail for Win32 (v3.12c)
Reply-To: djgpp-workers AT delorie DOT com


>   In message <200005152236 DOT RAA05926 AT hp2 DOT xraylith DOT wisc DOT edu>you write:
>   > > Such as?  Do we need to extend the things provided by gcc's float.h?  Or 
>   > is
>   > > it in our best interest to give up and use the system one?
>   > 
>   > The extra stuff has to do with dealing with the FPU (exceptions codes,
>   > control routines, and a bunch of other stuff). We could probably modify 
>   > the system headers to include these with math.h as well with preprocessor 
>   > guards, but that'll create an incompatibility with headers supplied by 
>   > Microsoft in it's Platform SDK and that I'd like to avoid. 
> Would the #include_next trick work, or is it inappropriate?
> 

These very same extras are in DJGPP's float.h and for reasons that go way back I'm 
sure. I would think adding this trick will work for Mingw32, since the C runtime comes 
with the compiler. But this won't work with DJGPP (at least not right away), since the 
libc release is on an independant schedule. So we'd need another solution. Perhaps 
adding magic to float.h to include a file that adds in the system-specific stuff. Or 
perhaps including the extra stuff in the generated float.h itself?