[vlc-devel] Re: re : At last, the problem! Now, what to do?

[John Michael Zorko]

Gildas,

>> This is a problem with input_Seek(), my ff / rewind code
>> exacerbates it because that's how it does its' thing.  I
>> want to fix this, but i'm not sure how yet.  Any ideas?
>> Should seeks be constrained by input_ManageClockRef(), as
>> reads are (by the demux module)?
>
> A seek operation should normally reset the input's clock so new data 
> is properly time-stamped. This should be enough to prevent the problem 
> you are describing so there has to be a bug somewhere which let's 
> improperly time-stamped samples pass through.

... but what about when the seeks happen often?  The ff / rewind code I 
have (and put on a website for all of you to see) allows such things as 
"play 10 frames, then go back 20, play 10 frames, then go back 20 ..."  
So, the seek always resets the clock, but if, as in this example, 10 
MP3 frames happens before the delay that input_ManageClockRef() does, 
then it seems to me that this could be a problem, i.e. we're resetting 
the clock all of the time, and never letting the clock expire 
naturally, so we in essence have no clock anymore.

> Seeks shouldn't have to be constrained by input_ManageClockRef(). This 
> function is only used to regulate the reading pace of the demux (well, 
> it is also used to detect PCR gaps and adjust the input's clock in 
> these cases but this operation is done by the next demux loop after 
> the seek).
>
> You could maybe try to put a printf in 
> src/input/input_clock.c:ClockNewRef() to check when the clock is 
> reinitialized, as well as print the time-stamps of the samples the 
> demux puts in the decoder fifo.

I'll look in ClockNewRef() ... I remember printing the timestamps 
before, i'll put that code back in.

Regards,

John

Falling You - exploring the beauty of voice and sound
New EP, "hope thrown down," available now at
http://www.mp3.com/fallingyou





-- 
This is the vlc-devel mailing-list, see http://www.videolan.org/vlc/
To unsubscribe, please read http://developers.videolan.org/lists.html
If you are in trouble, please contact <postmaster at videolan.org>