[Ffmpeg-devel] make test fails

Panagiotis Issaris takis.issaris
Tue Oct 3 17:25:37 CEST 2006


Hi,

On Tue, Oct 03, 2006 at 11:11:20AM -0400, Dana Hudes wrote:
[...]
> As to how they work, i'm not sure  I understand your remark. they are 
> run by "make test" (the usual such tag). there is also "make fulltest" 
> (also a common naming practice). If you mean you don't understand what 
> the regression tests do, that detail I am still sorting out myself.
> The question remains whether the tests are broken or the code is broken.
> I can work to come up with a build that doesn't include the code that 
> breaks the test but it requires more knowledge of the project than I 
> have at this time to understand whether code or test is broken.
> Documentation of the tests would be very helpful.
I do not know if this is the kind of information you were looking for,
but anyway...

The regression tests work as follows (AFAIK):
There are several applications (audiogen.c, rotozoom.c, videogen.c) generating
video streams and an audio stream.  These generated streams are encoded using
several codecs with various flags. The checksum (md5sum) of the resulting
encoded files is compared to the (reference) checksum which is stored in SVN for
that specific test.  The encoded file is also decoded again, and the checksum of
resulting decoded file is also compared agains a stored (reference) checksum.

So, if the testsuite succceeds it means that the (tested) encoders and decoders
give the exact same results (bit for for) as on the codec authors/reference
system.

Hope this helps a bit.

With friendly regards,
Takis




More information about the ffmpeg-devel mailing list