[FFmpeg-devel] [PATCH] Add libavsequencer.

[Ronald S. Bultje]

Hi,

On Wed, Aug 25, 2010 at 12:15 PM, Sebastian Vater
<cdgs.basty at googlemail.com> wrote:
> Ronald S. Bultje a ?crit :
>> On Wed, Aug 25, 2010 at 8:36 AM, Sebastian Vater
>> <cdgs.basty at googlemail.com> wrote:
>>> My plan to implement the tests is, that I write, while playback &
>>> mixing, to a text file containing the internal dump of the player / BSS
>>> / mixer structures. This has to be done twice, one for original
>>> TuComposer, and one for AVSequencer.
>>
>> Why not just compare the PCM output? If they are identical for a bunch
>> of files, likely the mixer works the same as before. This is how all
>> FFmpeg regtests works (http://fate.ffmpeg.org/).
>
> My plan here was to take all IFF-TCM1 files I have for the test, whose
> number is around 1000, let's say for an average duration of 3 mins as
> 32-bit 44,1kHz stereo samples, we would have a total duration around
> 3000 minutes, which is around 50 hours of total playback time. Storing
> such a huge amount requires lots of HDD space (1211 MB for 60 minutes).
>
> We either should use then a lower sample rate (4kHz output should be
> enough for a fate test), or we use mp3 64kbit/s as output codec.

You're getting good at this. :-). I don't think you need 3 minutes,
that samplerate is indeed high also.

> Another drawback would be, if a test fails, we don't have any
> informations why and what broke, but in that case I could still run a
> specialized test.

The purpose of FATE is not to tell us what you broke. It is to tell us
you broke it. That's what machines are good at, especially at such
large quantities.

We, humans, will then find out what we broke, that's what we're good at.

Ronald