[MEncoder-users] scaling segfaulting...

[John Doe]

From: Carl Eugen Hoyos <cehoyos at ag.or.at>

> To be able to reproduce your issue, your command line, complete,  uncut output 
>of
> MPlayer/MEncoder and the output of "bt" (in gdb) are still  missing.

It was there, except the gdb prompts, here it is again:

Program received signal SIGSEGV, Segmentation fault.
0x0854a312 in hScale_MMX2 (dst=Cannot access memory at address 0xfffffc48
) at swscale_template.c:2095
2095            __asm__ volatile(
(gdb) bt
Cannot access memory at address 0xfffffc44
(gdb) disass $pc-32 $pc+32
Dump of assembler code from 0x854a2f2 to 0x854a332:
0x0854a2f2 <hScale_MMX2+353>:   dec    %ebp
0x0854a2f3 <hScale_MMX2+354>:   sbb    $0xf,%al
0x0854a2f5 <hScale_MMX2+356>:   out    %eax,(%dx)
0x0854a2f6 <hScale_MMX2+357>:   call   *-0x77(%ebp)
0x0854a2f9 <hScale_MMX2+360>:   lds    0xb6(%ebp),%ecx
0x0854a2ff <hScale_MMX2+366>:   add    %cl,(%edi)
0x0854a301 <hScale_MMX2+368>:   mov    $0x4,%bh
0x0854a303 <hScale_MMX2+370>:   sub    (%edi),%cl
0x0854a305 <hScale_MMX2+372>:   mov    $0x5c,%bh
0x0854a307 <hScale_MMX2+374>:   sub    (%edx),%al
0x0854a309 <hScale_MMX2+376>:   movq   (%ecx,%ebp,8),%mm1
0x0854a30d <hScale_MMX2+380>:   movq   0x10(%ecx,%ebp,8),%mm3
0x0854a312 <hScale_MMX2+385>:   movd   (%esi,%eax,1),%mm0
0x0854a316 <hScale_MMX2+389>:   movd   (%esi,%ebx,1),%mm2
0x0854a31a <hScale_MMX2+393>:   punpcklbw %mm7,%mm0
0x0854a31d <hScale_MMX2+396>:   punpcklbw %mm7,%mm2
0x0854a320 <hScale_MMX2+399>:   pmaddwd %mm1,%mm0
0x0854a323 <hScale_MMX2+402>:   pmaddwd %mm2,%mm3
0x0854a326 <hScale_MMX2+405>:   movq   0x8(%ecx,%ebp,8),%mm1
0x0854a32b <hScale_MMX2+410>:   movq   0x18(%ecx,%ebp,8),%mm5
0x0854a330 <hScale_MMX2+415>:   movd   0x4(%esi,%eax,1),%mm4
End of assembler dump.
(gdb) info all-registers
see previous post
(gdb)

The command line is still the same as my previous post.
(gdb) r bunny.ogv -o bunny.avi -oac mp3lame -ovc lavc -vf scale=480:268
The output is the same as my previous post.

Dunno if the fact that it says MMX2 and my /proc/cpuinfo only says mmx is the 
problem.
Dunno if it is another lib that is the problem.
Anyway, I don't have too much free time right now and if it works for you, 
something must be wrong on my side.
I will try again later with a new revision and new libs versions.

Thx for your time,
JD