[Bug bootstrap/82045] [8 regression] SPARC bootstrap broken: ICE in emit_library_call_value_1, at calls.c:4565

["rsandifo at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82045

--- Comment #4 from rsandifo at gcc dot gnu.org <rsandifo at gcc dot gnu.org> ---
(In reply to ro@CeBiTec.Uni-Bielefeld.DE from comment #3)
> > --- Comment #1 from rsandifo at gcc dot gnu.org <rsandifo at gcc dot
> > gnu.org> ---
> > Could you attach the .i file?  Using a cross compiler, I was able to build the
> 
> Sure, done.

Thanks.  This also works with the cross compiler, even though it
does seem to be going through the same code path as the one shown
in the backtrace.

> > parts of libgcc that don't depend on system headers, and _multc3.o built OK for
> > me.  But that probably isn't indicative of what happens on real Solaris.
> 
> Natively, I can easily reproduce the ICE with
> 
> $ cc1 -fpreprocessed libgcc2.i -quiet -mcpu=v9 -o libgcc2.s
> 
> Please note that I'm using a 32-bit compiler; maybe the problem doesn't
> occur on a 64-bit host?

That's certainly a possibility, although I don't *think* we changed
anything in terms of variable sizes.

This is a bit of a long shot, but maybe still worth trying.
Could you see what happens if you change:

#if GCC_VERSION >= 4000
#define ALWAYS_INLINE inline __attribute__ ((always_inline))
#else
#define ALWAYS_INLINE inline
#endif

so that the plain "inline" version is always chosen?