[mythtv-users] Timestretch causes 'illegal instruction' exit in 0.24

John Pilkington J.Pilk at tesco.net
Thu Dec 9 08:31:51 UTC 2010


On 09/12/10 00:44, Jean-Yves Avenard wrote:
> Hi
>
> On 9 December 2010 00:04, John Pilkington<J.Pilk at tesco.net>  wrote:
>> I just checked the f12 core2duo machine with Myth 0.24; that does
>> timestretch properly, as seemed likely from the lack of reports about it.
>>   Main motivation for putting 0.24 on the P4 was to see it in action (and try
>> out the CentOS builds) before switching 'production' from 0.23-fixes.  Dual
>> booting the production box tends to restrict the opportunities for
>> experiment, but I think I'm close to switching point now.
>
> Can you provide a backtrace so we can identify which instruction is invalid ?

I've no experience of this, so it won't be quick.  stuarta posted this 
four months ago:-

http://svn.mythtv.org/trac/attachment/ticket/8740/illegal-instruction-gdb.txt
>
> It seems that in the SSE2 code used in the soundtouch library (used
> for "stretching" the sound) there are a few SSE3 instructions
> incorrectly placed there which would make your processor crash as they
> aren't supported.
>
> The other possibility is that you recompile without SSE support (it's
> a configure option)

... and again, I've let Axel do the compiling so far and am not sure how 
much of the system would need to be respun.  Maybe that would be a 
reasonable option for me.  I don't know how many other people are likely 
to run into the problem as 0.24 takeup grows.

Thanks.
>
> Jean-Yves


More information about the mythtv-users mailing list