[MPlayer-dev-eng] workaround for SEGFAULTs during rtsp download

[Jindrich Makovicka]

Willem de Bruijn wrote:
>>>Can you use electricfence?  You might also look at the use the memory
>>>malloced
>>>before the one that crashes, and see if it overwrites the end of the
>>>array. Maybe some code assumes sizeof(int)==sizeof(void*), which isn't
>>>true on x86-64, is it?
>>
>>ok, I will. But it'll have to wait until monday as my only 64bit machine is
>>at work. First I'll check out electric fence. I completely forgot about
>>that.
> 
> 
> I've tried to get electric fence to work, but I need some help here. The 
> latest version from Bruce Perens doesn't crash during self-tests, so I expect 
> that it's okay.
> 
> However, when backtracing mplayer linked against -lefence I get the following 
> output:
> 
> #607 0x2f65726168732f72 in ?? ()
> #608 0x64782f666d786574 in ?? ()
> #609 0x7273752f202f6976 in ?? ()
> #610 0x632f65726168732f in ?? ()
> #611 0x4f43006769666e6f in ?? ()
> 
> this goes on for some 600 lines before stopping. The same works fine without 
> -lefence as I have built mplayer with -ggdb. I'm hoping that this is some 
> electric fence newbie thing with which you have had prior experience. It 
> seems there's no debugging information available in the electric fence 
> library or so.

I'd suggest also using valgrind, it usually gives better results than 
efence. You just have to disable postprocessing, as it contains some 
offending code (creating temporary variables in asm code) which valgrind 
doesn't like.

-- 
Jindrich Makovicka