[FFmpeg-devel] [PATCH] fate: add tests for some video source filters

Nicolas George george at nsup.org
Fri Jul 21 13:28:13 EEST 2017

Le tridi 3 thermidor, an CCXXV, Michael Niedermayer a écrit :
> this patch breaks the selftests on mips or maybe more precissely
> it adds a test which fails

Yes, I understand that. What I ask is who is responsible for the
failure: the test or the code being tested?

If somebody adds a test on a code that uses floats, it will fail on
x86_32: the code is correct, the failure is the test's fault.

If somebody adds a test on a code that has bogus SSE optimizations, it
will fail on machines with SSE: the test is correct, the failure is the
test's fault.

What I am asking is: which one of these is it today?

> the format used seems the native one that has different endianness
> and thus different checksums
> md5sum out.png mips/out.png
> 012e1906fe84ee6f757ef3f1473cc26d  out.png
> 012e1906fe84ee6f757ef3f1473cc26d  mips/out.png

md5sum tells us the files are identical, but it does not tell us if the
output is correct. To see if the output is correct, I would suggest to
convert it using a tool entirely different from FFmpeg (ImageMagick?)
losslessly into a common format and compare.

Plus, of course, visually checking the frames.

The real question here is: on MIPS, is this filter's output correct?


  Nicolas George
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20170721/4a9c89dc/attachment.sig>

More information about the ffmpeg-devel mailing list