Re: [xine-devel] java?, xine and visual_x11_2

[Matthias R. <mri...@in...>]

On 23.01.2008, at 22:30, I wrote:

>>> I've had debug printfs showing, that indeed, there is some xine
>>> code which
>>> locks the
>>> mutex twice. but everything is working.
>>
>> That's wrong, definitely.
>
> Interesting. Then, I'm gonna provide you (the list) with a trace of
> this.


Hi Matthias, :)

Ok. Here's the trace of video_out_xv.c...

open_plugin_2:
-- LOCK
-- xv_store_port_attribute  
    -- LOCK/UNLOCK
-- xv_check_capability
    -- xv_set_property
       -- LOCK/UNLOCK
-- xv_update_XV_DOUBLE_BUFFER
    -- LOCK/UNLOCK
-- UNLOCK

open_plugin_2 is calling LOCK (resp XLockDisplay) and then calls  
various other functions which will again call LOCK.
I don't know if this is a problem at all as XLockDisplay is supposed  
to be called several times by the same thread.
But, it certainly would not hurt, if video_out_xv would call  
XLockDisplay (or the user supplied function) only once per
x11 call.

I will use a simple reference count in the LOCK macro to see, if my  
problems with deadlocks with the java binding go away.
If this should be fixed, I can try to be the nit-picker and shuffle  
the LOCKs around a bit.

cheers
matthias