This is the mail archive of the
gdb-patches@sources.redhat.com
mailing list for the GDB project.
Re: [RFA] ppc-linux-nat.c AltiVec regs ptrace
Daniel Jacobowitz writes:
> On Wed, Feb 20, 2002 at 04:06:56PM -0500, Elena Zannoni wrote:
> > Right will fix. [I had an ongoing bet :-)]
>
> What, whether Andrew would get to you before I did? :)
> <duck and run>
No it was actually Kevin, I was thinking of.
<duck and run myself>
>
> > > > +int have_ptrace_getvrregs
> > > > +#ifdef HAVE_PTRACE_GETFPXREGS
> > > > + = 1;
> > > > +#else
> > > > + = 0;
> > > > +#endif
> > > > +
> > >
> > > Huh? You defined GETVRREGS unconditionally above. GETFPXREGS has no
> > > place in this file, does it? Or do the headers define GETFPXREGS?
> > > You also continue this confusion all the way down the patch.
> > >
> >
> > The glibc headers define GETFPXREGS, and that's what we test for in
> > the configury. But we are not dealing with floating point registers
> > here, so I used the 'correct' name where I could. It would be more
> > confusing to talk about FPX regs while instead there are none.
> > I explained this in the comments.
> >
> > I guess I can do the following if it helps.
> > #ifdef HAVE_PTRACE_GETFPXREGS
> > #define HAVE_PTRACE_GETVRREGS
> >
> > Whatever I end up using it's partially going to be a lie. I would
> > prefer using the VRREGS nomenclature where relevant, though.
>
> I'm confused.
Yeah, you are not the only one.
>
> On i386, glibc defines PTRACE_GETFPXREGS. On PowerPC, in current FSF
> glibc, sys/ptrace.h does not define anything along these lines at all.
OK, I have downloaded glibc 2.2.5, and sysdeps/unix/sysv/linux/sys/ptrace.h
defines PTRACE_GETFPXREGS.
Then on my system, I have /usr/include/sys/ptrace.h which also defines it.
But I think I have an older version of glibc installed.
What I am not understanding is where the installed file comes from, is
it the same as sysdeps/unix/sysv/linux/sys/ptrace.h?
> The kernel <asm/ptrace.h> define GETVRREGS (not that we should be
> including that header, of course). [<sys/ptrace.h> is an
> architecture-specific header, which may not have been apparent.]
>
Right. I didn't rely on it.
> If there are outstanding patches to glibc, which defines
> PTRACE_GETFPXREGS on PowerPC, then they are still mutable. They should
> be updated to a reasonable value.
>
I think that rather than oustanding patches we may have older versions.
I see that in glibc2.2.5 the file
sysdeps/unix/sysv/linux/powerpc/sys/ptrace.h
doesn't use the values 18 and 19.
If I determine that the version of glibc I have used is obsolete, then
I can clean that up. Let me have a look.
Elena
> --
> Daniel Jacobowitz Carnegie Mellon University
> MontaVista Software Debian GNU/Linux Developer