From: Thibaut M. <tma...@no...> - 2002-05-06 21:37:07
|
On Monday 06 May 2002 21:11, Miguel Freitas wrote: > On Mon, 2002-05-06 at 15:56, Thibaut Mattern wrote: > > > I think i got your point, since BUF_CONTROL_END will close decoders= and > > > therefore audio_out it will probably slow things down. I have no id= ea > > > of how to fix these, since demux would have no way to know that you > > > will want to do a seek in near future... any suggestion? > > > > Yes, you got the point. > > My suggestion : > > 1) use a mutex to protect a/v loop sync from seeking. > > What do you mean by a/v loop sync? The situation you describe when video_loop receive a BUF_CONTROL_END and = wait=20 that audio_loop receive one. In this case fifo_buffer_clear() should not=20 delete a BUF_CONTROL_END. > > 2) add a new parameter to fifo_buffer_clear(), I think about somethin= g > > like passing BUF_MAJOR_MASK for normal clear and an other mask for a > > seek. I don't know if it can work with discontinuities. > > This is dangerous. As i said before you risk leaving a odd number of > discontinuties/start/stop buffers and xine will hang forever. Ok, if it's too dangerous I will try to find an other way. Let's see what happens if the demuxer wait before sending end buffers. > regards, > Miguel Thanks for your advice. Thibaut |