[FFmpeg-devel] [PATCH] Warn about PAFF & Spatial

[Serguei Miridonov]

On Wednesday 16 July 2008, Paul Kendall wrote:
> On Thursday 17 July 2008 09:31:58 Michael Niedermayer wrote:
> > On Wed, Jul 16, 2008 at 01:57:43PM -0700, Serguei Miridonov wrote:
> > > On Wednesday 16 July 2008, Serguei Miridonov wrote:
> > > > On Wednesday 16 July 2008, Michael Niedermayer wrote:
> > > > > On Wed, Jul 16, 2008 at 08:21:02AM -0700, Serguei Miridonov 
wrote:
> > > > > > On Wednesday 16 July 2008, Michael Niedermayer wrote:
> > > > > > > On Tue, Jul 15, 2008 at 11:36:35PM -0700, Serguei
> > > > > > > Miridonov
> > > >
> > > > wrote:
> > > > > > > > While the issue 460 is not resolved, just to avoid
> > > > > > > > log flood:
> > > > > > > >
> > > > > > > > See attachment..
> > > > > >
> > > > > > [...]
> > > > > >
> > > > > > > non constant static,
> > > > > >
> > > > > > Sorry, I don't know what you mean...
> > > > >
> > > > > I mean if i run
> > > > > mplayer *.h264
> > > > > then the warning will be printed just once, it should be
> > > > > once per file also if i decode 2 h264 files in parallel i
> > > > > want to see 2 warnings if it applies to both files
> > > >
> > > > Well, for me, and may be for many other users, it would be
> > > > enough to see a warning only once: just to know that ffmpeg
> > > > does not support that mode. And this can be implemented very
> > > > easy by this patch.
> > > >
> > > > Another approach is to make it as you suggest. But this, as I
> > > > understand, will require to add a variable in file or stream
> > > > context structure. If you know how to do that, do ahead. I
> > > > just don't feel myself very familiar with ffmpeg and
> > > > therefore don't want to add a complexity which, I think, is
> > > > not necessary.
> > > >
> > > > May be the best approach is to make the log subsystem to
> > > > report some kinds of errors/warnings once using a flag like
> > > > AV_LOG_FLAG_ONCE. But this also requires additional
> > > > complexity in data structures...
> > >
> > > So, Michael, what do you suggest? Do you think that log flood
> > > with warning after every frame is better than to show the
> > > message just once?
> >
> > I think nasty hacks are worse than a few log messages.
> > If the log flood bothers you, find a clean solution or fix the
> > decoder so the combination is fully supported or in case its
> > alraedy fully supported and the message is wrong provide some
> > evidence for that. Or change it to warning instead of error ...
> > [...]
>
> Or perhaps, adding the 'once' variable to H264Context is better,
> and obviously choosing a better name?

Sorry, I won't do anything. At least at this time because I don't 
understand quite well Michael's paradigm.

As Michael wrote:

> > I think nasty hacks are worse than a few log messages.

Not a few, unfortunately. They are sent practically after every frame.

> > find a clean solution

What is a clean solution? I don't consider sending error message every 
frame as a cleaner solution. This is why I have suggested a patch.

> > fix the decoder so the combination is fully supported

Sorry, I think there are some people here who know both H.264 and 
ffmpeg infrastructure, so they can do that much faster than me. I 
understand, they have other things to do, so I, and, perhaps other 
ffmpeg/mplayer users, will wait.

Paul Kendall wrote:

> Or perhaps, adding the 'once' variable to H264Context is better,
> and obviously choosing a better name?

I don't think that this would be a cleaner solution from Michael's 
point of view. How many 'once'-like variables in H264Context is 
necessary to add? One for every warning or error which may 
potentially flood the log? I'm sorry but since I could neither find 
nor derive criteria of cleanness from recent postings, I leave it as 
is: in my personal svn repository, so that future updates will 
include my patch automatically for my own needs.

Best regards,
Serguei.