[Bug target/87702] [6/7/8/9 Regression] Segfault in glibc if compiled with -march=amdfam10 -O2 (x86)

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87702

--- Comment #6 from Uroš Bizjak <ubizjak at gmail dot com> ---
(In reply to Uroš Bizjak from comment #4)
> Program received signal SIGSEGV, Segmentation fault.
> _IO_vfscanf_internal (s=0xffffc068, format=0x80ae357
> "%llu%llu%u%u%u%u%u%u%llu%llu%u%u%u%u%u%u", 
>     argptr=0xffffc140
> "\354\273\v\bܻ\v\b\374\273\v\b\004\274\v\b,
> \274\v\b(\274\v\b\020\274\v\b\f\274\v\b\364\273\v\b\344\273\v\b", errp=0x0)
> at vfscanf.c:2447
> warning: Source file is more recent than executable.
> 2447                    *ARG (float *) = negative ? -d : d;
> (gdb) list
> 2442              else
> 2443                {
> 2444                  float d = __strtof_internal
> 2445                    (char_buffer_start (&charbuf), &tw, flags & GROUP);
> 2446                  if (!(flags & SUPPRESS) && tw != char_buffer_start
> (&charbuf))
> 2447                    *ARG (float *) = negative ? -d : d;
> 2448                }
> 2449
> 2450              if (__glibc_unlikely (tw == char_buffer_start (&charbuf)))
> 2451                conv_error ();

Also here, please provide preprocessed source of vfscanf.c from the build and
exact compile flags that will generate the failing assembly.