[vlc-devel] commit: Extensions: new script capability "input-listener" ( Jean-Philippe André )

[Jean-Philippe André]

Wednesday 03 February 2010, 21:01:19,
Laurent Aimar wrote:
> On Wed, Feb 03, 2010, git version control wrote:
> > +            p_ext->p_sys->p_input = p_input ? vlc_object_hold( p_input )
> > +                                            : p_input;
> 
>  Just to nitpick, using p_input ? hold() : NULL will be more
> readable.

Right, this was a typo. (I won't commit just for this, but it will be "fixed" 
:-) )

>On Wed, Feb 03, 2010, git version control wrote:
>> +    vlc_mutex_lock( &p_ext->p_sys->command_lock );
>>      vlc_cond_signal( &p_ext->p_sys->wait );
>> +    vlc_mutex_unlock( &p_ext->p_sys->command_lock );
>
>lock(); signal(); unlock() usually means there is a problem, unless
>of course the wake up condition is protected by another lock and is
>already set before these lines.

Oh well, I'm confusing myself then. The condition depends on the command_lock.
The issue I had was that the thread sometimes never woke up. After these 
lines, the calling thread will wait for the signaled thread to end (join).

Actually, as I'm tracking a race condition, I don't know how to find the right 
fix. But since that commit, the race didn't appear :-/

Best regards,

-- 
Jean-Philippe André (jpeg)