[PATCH] PPC enable Altivec for setjmp/longjmp part 1

[Ulrich Drepper]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Steve Munroe wrote:

> --- libc23-cvstip-20040129/sysdeps/powerpc/Versions	2002-09-05 04:54:59.000000000 -0500
> +++ libc23/sysdeps/powerpc/Versions	2004-01-29 13:12:15.000000000 -0600
> [...]
> +ld {
> +  GLIBC_2.3.4 {
> +     _longjmp; __sigsetjmp; _setjmp;
> +     longjmp; setjmp;
> +  }

Why would ld.so need to export these interfaces?  Look at the ld.map
file in your build directory, only interfaces related to dynamic linking
are exported.



> +/* 
> + * ALTIVECPIM.pdf specifies for SVR4 ABI and EABI on 32 bit vrsave must 

Drop this reference.  Filenames have no place in the headers.  Fix the
comment layout, we don't use * at the beginning of the lines.  And fix
the typos.


> +/* Save the current program position in ENV and return 0.  */
> +int
> +__novmx__libc_sigsetjmp (jmp_buf env, int savemask)
> +{
> +  /* Save the signal mask if requested.  */
> +  __novmx__sigjmp_save (env, savemask);
> +
> +  __set_errno (ENOSYS);
> +  /* No way to signal failure.	*/
> +  return 0;
> +}

What is the meaning of the ENOSYS?  The function worked, there is no
reason to set errno.  This is the compatibility code and it must behave
exactly as the code which is used now.



- -- 
➧ Ulrich Drepper ➧ Red Hat, Inc. ➧ 444 Castro St ➧ Mountain View, CA ❖
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)

iD8DBQFAJa6D2ijCOnn/RHQRAiOmAKCiXlxLOrEYmnpjsDSi1d8pR4pXGwCeOSX+
2tlB6gkYQKUbgb8+KucVqSs=
=ozIO
-----END PGP SIGNATURE-----