This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: 3.4 PATCH: Fix libffi MIPS O32/N64 ABI handling
On Wed, Oct 08, 2003 at 08:09:10PM +0200, Rainer Orth wrote:
> Daniel Jacobowitz writes:
>
> > GNU/Linux, on the other hand, expects _MIPS_SIM_ABI32. It'd be a
> > little hard to change at this date, since the kernel and C library have
> > it all over the place.
>
> Do you know where the get the definition of that symbol? Only from GCC's
> predefines, or from <sgidefs.h> or some GNU/Linux equivalent thereof.
Yes, GNU/Linux gets this from <sgidefs.h>, in both glibc and the
kernel.
> It seems strange that MIPS GNU/Linux depends on GCC predefining
> _MIPS_SIM_ABI32 for the O32 ABI, but can easily live without the
> corresponding definitions for N32/N64 (where _MIPS_SIM is only defined in
> terms of _ABIN32/_ABI64).
>
> If GCC really must continue to predefine _MIPS_SIM_ABI32 for O32, I'm
> certainly fine to leave that definition for linux (just a one-line change
> to config/mips/linux.h), but I think it's better for the GCC tree to be
> internally consistent and use the _ABI* values throughout for all three
> ABIs.
>
> > Not sure if you were proposing that, just commenting.
>
> Have you checked the actual patch?
>
> http://gcc.gnu.org/ml/gcc-patches/2003-10/msg00607.html
I replied before you posted the patch. See my other question:
Would this have always have the same value as _MIPS_SIM_ABI32, i.e. do
all systems use the same number? It appears OK for mips-linux.
drow@nevyn:/opt/src/glibc/libc/sysdeps% grep -r MIPS_SIM mips
unix/sysv/linux/mips | grep ABI32 | wc
55 298 3584
On the other hand I believe they are all a recent development and could
be changed. But I don't know what else is using this value.
And Alex's concern about the origins, of course.
--
Daniel Jacobowitz
MontaVista Software Debian GNU/Linux Developer