On 04/01/2011 07:09 PM, Reinhard Tartler wrote:
>> +typedef struct AVPacketQueue {
>> + AVPacketList *first_pkt, *last_pkt;
>> + int nb_packets;
>> + int size;
>> + int abort_request;
>> + pthread_mutex_t mutex;
>> + pthread_cond_t cond;
>> +} AVPacketQueue;
>
> err, does really each element of the queue need it's own mutex and
> condition variable? AFAIU your code, there is no race on single elements
> of the queue, but you are syncorinzing get() and puts() on elements of
> the queue.
I basically copied the ffplay one with the idea to factor it out once
the thing works.
> I wonder if this AVPacketQueue wouldn't be better placed somewhere in
> libavformat/
It is there mostly because I wanted to test it, sadly I got wrong the
packet duplication.
> TBH, I'm actually a bit surprised that we don't have a syncronized list
> implementation already.
I know and I want one (the decklink avdevice will use it)
>
>> #if CONFIG_AVFILTER
>>
>> static int configure_filters(AVInputStream *ist, AVOutputStream *ost)
>> @@ -447,6 +560,15 @@ static int decode_interrupt_cb(void)
>> return q_pressed || (q_pressed = read_key() == 'q');
>> }
>>
>> +// kill the thread and free the queue
>> +static void stop_output_queue(AVOutputQueue *out)
>> +{
>> + //pthread_cancel(out->th);
>
> ??
This code shouldn't be committed, it is just for comment and ideas.
> I didn't test your patch, but you are changing the semantics of the
> loop. Shouldn't a 'new_pkt = *pkt' remain at the head of the loop?
Uhm... that is part of the first experiment to fix the breakage from
this patch ^^;
lu
--
Luca Barbato
Gentoo/linux
http://dev.gentoo.org/~lu_zero
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel
