your mail

[trev . saunders]

Hi,

> On Mon, Jan 18, 2010 at 08:16:10AM EST, trev.saunders at gmail.com wrote:
> > Hi,
> > 
> > In trying to fix seg faults in sd_espeak I decided that rewritting was prob
> ably the best option.  Seeattached my first attempt.  Please note this is ver
> y experimental, and needswork.  However I believe 
> 
> A couple of questions:
> 
> 1. Did you check to see where the segfaults were occurring with gdb?

So far what I know is the following.

1. dmesg says it's in libc.
2. the only time I see dmesg saying it took place in sdaudio was on comming out 
of s2disk.
3. a while back (couple months) I was seeing seg faults in sdaudio during 
normal operation.
4. I am currently waiting for sd_espeak to seg fault while atttached to gdb.


> 2. What audi ooutput method were you using at the time of the segfault?

I am using alsa, I've been meaning to try libao, but haven't done so yet.

> 
> Yes sd_espeak is known to segfault, but the segfault generally occurrs within
>  libsdaudio, the audio output library, inside the alsa output code. THis is a
> n issue that nobody has yet successfully tracked down, although I suspect its
>  a threading/data access related bug.
> 

from the conditions I mentioned above I believe that the segfaults coming out 
of s2disk are related to  the order  in which things happen coming out of 
hiernate, and do have to do with sdaudio.
It may be that the call to libc that causes the seg fault is in sdaudio.  I 
suspect you are right in thinking this is a problem with threading and data 
access in sdaudio and sd_espeak. The threading situation in sd_espeak looks far 
to complicated to me but I will listen to an explianation of why the is this if 
there is one.

Trev