3.4 PATCH: Consistently use _ABIO32 for _MIPS_SIM

[cgd@broadcom.com]

At Mon, 20 Oct 2003 18:22:23 +0000 (UTC), "Rainer Orth" wrote:
> Eric Christopher writes:
> 
> > I'm worried about current users. Anything use it in, say, the linux
> > kernel after including sgidefs?
> 
> There's no change here: with <sgidefs.h> included, _MIPS_SIM_ABI32 and
> _ABIO32 are identical (1), so _MIPS_SIM has the same value before and after
> my patch.
> 
> There's only a change when someone used _MIPS_SIM *without <sgidefs.h>* and
> relied on the fact that _MIPS_SIM was 0 (a value not matching any defined
> _MIPS_SIM_ABI*/_ABI* value) in this case.  As Alexandre agreed, this use
> was broken from the beginning and I don't think we should support this.

I happen to agree with this sentiment (old gcc behaviour was broken
and unintentional, and shouldn't be supported).

Unfortunately, it just so happens that a bunch of files in glibc do
exactly this (use _MIPS_SIM without <sgidefs.h>).  So, it looks like
glibc builds (for mips-linux, mips64-linux) are probably broken by
this change.

(I didn't check the latest greatest glibc sources yet to make sure,
but i did check the ChangeLogs and it doesn't look like there have
been any changes w.r.t. uses of <sgidefs.h> since the version i
checked and found broken.  I'm not going to have time to look into/fix
this, unfortunately.)



chris