[FFmpeg-devel] New asf demuxer

Carl Eugen Hoyos cehoyos at ag.or.at
Sun Jun 28 17:02:50 CEST 2015

On Sunday 28 June 2015 04:51:53 pm wm4 wrote:
> On Sun, 28 Jun 2015 16:46:12 +0200
> Carl Eugen Hoyos <cehoyos at ag.or.at> wrote:
> > On Sunday 28 June 2015 04:42:30 pm Nicolas George wrote:
> > > Le decadi 10 messidor, an CCXXIII, Carl Eugen Hoyos a écrit :
> > > > But please allow me to repeat my question: Why do
> > > > think the new demuxer should be used? What sample
> > > > does it fix?
> > >
> > > Why do you think it should NOT be used?
> >
> > The main reason imo is that it is (apparently, I may
> > of course be wrong) not intended to fix anything but
> > just to remove a demuxer that has seen many (man-)
> > years of testing.
> The new demuxer was written based on the official ASF spec, 

I am not sure what you are trying to say here...

> and was tested against a number of real world samples.

Please share the samples, I would also like to test!

> Also, the code being old is really meaningless here. The old 
> demuxer didn't handle all files well either and it's probably 
> full of bugs. 

Are you thinking of a specific issue?
As said, it seems to me that the new demuxer does not fix any 
issue that the old demuxer has.

> > From a very quick look, the new code seems mostly
> > unreviewed
> This is wrong.

From a very cursory look, I saw things like:

int a = 0;
int b = 1;
a = 0;


//Set dts
pts = ...

I of course do not claim that these issues are important but 
I believe that code that is intended to replace existing code 
should have a slightly higher standard.

> The code has been in review for several months before it 
> was finally merged.

I of course cannot really comment on this claim but this is 
not what I felt happened on avconv-devel.

> > which makes me think that it can't really
> > be tested.
> This makes literally no sense at all. None.

The reviews appear to have been very cursory, no sample was 
named that got fixed: Why do you think that any testing 
was done?

> > Do you disagree? Do you know of samples that work
> > better with the new demuxer?
> It is known that it fixed some samples.

So you apparently did more testing than me. That's great, 
please share "some samples"!

(And no, a link to trac is not a wild claim.)

Carl Eugen

More information about the ffmpeg-devel mailing list