[MPlayer-users] BUG: mencoder of pre5 crashes everytime

[Martin Simmons]

>>>>> On Mon, 26 Jul 2004 13:31:47 +0200, Stefan Sassenberg <stefan.sassenberg at gmx.de> said:

  Stefan> (gdb) disass $pc-32 $pc+32
  Stefan> Dump of assembler code from 0x82e8fdf to 0x82e901f:
  Stefan> 0x082e8fdf <sad16_y2_mmx2+79>:  add    %cl,(%edi)
  Stefan> 0x082e8fe1 <sad16_y2_mmx2+81>:  outsl  %ds:(%esi),(%dx)
  Stefan> 0x082e8fe2 <sad16_y2_mmx2+82>:  add    $0x7,%al
  Stefan> 0x082e8fe4 <sad16_y2_mmx2+84>:  movq   (%edx,%eax,1),%mm2
  Stefan> 0x082e8fe8 <sad16_y2_mmx2+88>:  pavgb  %mm2,%mm0
  Stefan> 0x082e8feb <sad16_y2_mmx2+91>:  movq   (%ecx,%eax,1),%mm2
  Stefan> 0x082e8fef <sad16_y2_mmx2+95>:  psadbw %mm2,%mm0
  Stefan> 0x082e8ff2 <sad16_y2_mmx2+98>:  add    %esi,%eax
  Stefan> 0x082e8ff4 <sad16_y2_mmx2+100>: movq   (%edi,%eax,1),%mm1
  Stefan> 0x082e8ff8 <sad16_y2_mmx2+104>: movq   (%edx,%eax,1),%mm3
  Stefan> 0x082e8ffc <sad16_y2_mmx2+108>: pavgb  %mm1,%mm3
  Stefan> 0x082e8fff <sad16_y2_mmx2+111>: movq   (%ecx,%eax,1),%mm1
  Stefan> 0x082e9003 <sad16_y2_mmx2+115>: psadbw %mm1,%mm3
  Stefan> 0x082e9006 <sad16_y2_mmx2+118>: paddw  %mm3,%mm0
  Stefan> 0x082e9009 <sad16_y2_mmx2+121>: paddw  %mm0,%mm6
  Stefan> 0x082e900c <sad16_y2_mmx2+124>: add    %esi,%eax
  Stefan> 0x082e900e <sad16_y2_mmx2+126>: js     0x82e8fe0 <sad16_y2_mmx2+80>
  Stefan> 0x082e9010 <sad16_y2_mmx2+128>: mov    0x10(%ebp),%edi
  Stefan> 0x082e9013 <sad16_y2_mmx2+131>: mov    0xffffffe8(%ebp),%edx
  Stefan> 0x082e9016 <sad16_y2_mmx2+134>: mov    %ebx,%eax
  Stefan> 0x082e9018 <sad16_y2_mmx2+136>: neg    %eax
  Stefan> 0x082e901a <sad16_y2_mmx2+138>: lea    0x8(%edi,%ebx,1),%ecx
  Stefan> 0x082e901e <sad16_y2_mmx2+142>: mov    0xc(%ebp),%edi
  Stefan> End of assembler dump.
  Stefan> (gdb) info all-registers
  Stefan> eax            0xbfffdcc1       -1073750847
  Stefan> ecx            0x415a5150       1096438096
  Stefan> edx            0x41535140       1095979328
  Stefan> ebx            0x2f00   12032
  Stefan> esp            0xbfffdcc1       0xbfffdcc1
  Stefan> ebp            0xbfffdcd8       0xbfffdcd8
  Stefan> esi            0x2f0    752
  Stefan> edi            0x41534e50       1095978576
  Stefan> eip            0x82e8fff        0x82e8fff

This register dump suggests to me that something nasty is happening (esp is
slightly wrong and eax is set to the same wrong value, which is nothing like
it should be).  It could be worthwhile doing some more runs and posting the
debugging info to see if there is a pattern.

Also, if sad16_y2_mmx2 (or some similarly named function) appears in
subsequent crashes too, then the output of the gdb command

disass sad16_y2_mmx2

(or whatever the function is) might be useful.

__Martin