[FFmpeg-devel] [PATCH] handle EAGAIN case in ffmpeg.c

[Ramiro Polla]

Hi,

Benoit Fouet wrote:
> Hi,
>
> here is a patch to $subj
>
> what i'm trying to do is having v4l2 working
>
> this raises a question to me: how can we wait for the driver to give us
> a frame. I think we should wait somewhere, depending on the framerate we
> want.
>   

To try and finish frame rate emulation [1], I've analyzed all grab 
devices except bktr and audio_beos.
v4l, v4l2, dv1394, vfw32, oss will block themselves until there's a 
valid frame.
vfw32 has some issues on Windows 98 (as in returning invalid frame if 
called too frequently).
There really is no need for frame rate emulation in those devices unless 
you want to drop frames, so it'd be best to just remove their frame rate 
emulation code.

x11grab can give a valid frame at any time, so it should require rate 
emulation, but IMHO it should be done in the user app, not the demuxer.
gdi32 gives invalid frames if called too frequently, so it must use rate 
emulation.

The best thing IMO is to make realtime grabbing in separate threads. It 
requires many changes in ffmpeg though. It's been suggested to make 
libavgrab [2], and I'm working on this too.

Anyways, this is a hasty report. I have many exams this week, so I'll 
only look further on the weekend. I really should go back to studying now =)

[...]
> -        if (av_read_frame(is, &pkt) < 0) {
> +        if ((ret=av_read_frame(is, &pkt)) < 0) {
> +            if (ret != AVERROR(EAGAIN))
>   

Won't this busy-wait?

Ramiro Polla
[1] http://article.gmane.org/gmane.comp.video.ffmpeg.devel/55462
[2] http://article.gmane.org/gmane.comp.video.ffmpeg.devel/56399