From: Felipe C. <fel...@gm...> - 2009-02-14 17:13:36
|
On Thu, Feb 5, 2009 at 11:56 AM, Mark Nauwelaerts <ma...@sk...> wrote: > Felipe Contreras wrote: >> >> On Sat, Jan 31, 2009 at 1:05 PM, Felipe Contreras >> <fel...@gm...> wrote: >>> >>> On Fri, Jan 30, 2009 at 1:08 PM, Mark Nauwelaerts <ma...@sk...> >>> wrote: >>>> >>>> * serialize EOS handling to prevent EOS event from "speaking before >>>> turn" >>> >>> Since omx is asynchronous at the time we receive the EOS event from >>> gst, there are buffers that are still being processed, we need to wait >>> for them before the EOS event is propagated to the next element, >>> otherwise the buffers will be dropped. >>> >>> It looks like your patch will serialize the EOS event, so if for some >>> reason omx signals an EOS, gst will receive it too. But there's no way >>> that would happen for filter elements since gst-omx is the one that >>> flags omx buffers with the EOS flag *after* we receive the gst EOS >>> event. >> >> After further discussion now I understand what you mean and what this >> patch really does. Could you please re-send the patch with more >> explanation about the issue and the fix, also remove the internal >> Maemo bug id. >> >> Other than that it's good. Thanks! > > Patch attached as requested. I forgot to mention that the patch was pushed. Thanks. -- Felipe Contreras |