[FFmpeg-devel] [PATCH 2/2] fate: add tests for pan audio filter

[Tobias Rapp]

On 15.02.2018 09:39, Nicolas George wrote:
> Tobias Rapp (2018-02-15):
>> Signed-off-by: Tobias Rapp <t.rapp at noa-archive.com>
>> ---
>>   tests/fate/filter-audio.mak       | 30 ++++++++++++++++++++++++++++++
>>   tests/ref/fate/filter-pan-mono1   | 26 ++++++++++++++++++++++++++
>>   tests/ref/fate/filter-pan-mono2   | 26 ++++++++++++++++++++++++++
>>   tests/ref/fate/filter-pan-stereo1 | 26 ++++++++++++++++++++++++++
>>   tests/ref/fate/filter-pan-stereo2 | 26 ++++++++++++++++++++++++++
>>   tests/ref/fate/filter-pan-stereo3 | 26 ++++++++++++++++++++++++++
>>   tests/ref/fate/filter-pan-stereo4 | 26 ++++++++++++++++++++++++++
>>   7 files changed, 186 insertions(+)
>>   create mode 100644 tests/ref/fate/filter-pan-mono1
>>   create mode 100644 tests/ref/fate/filter-pan-mono2
>>   create mode 100644 tests/ref/fate/filter-pan-stereo1
>>   create mode 100644 tests/ref/fate/filter-pan-stereo2
>>   create mode 100644 tests/ref/fate/filter-pan-stereo3
>>   create mode 100644 tests/ref/fate/filter-pan-stereo4
> 
> I think the change in lswr is sane, but I do not know enough of that
> code to be definite. Thanks for tracking down that ticket.
> 
> As for this patch, thanks for implementing it. It looks reasonable. But
> I am not entirely sure lsws is bit-exact in this kind of situation.

I have run the Fate tests on Linux 32/64bit and Mips/Qemu, but possibly 
the float channel coefficients could be fragile (especially of the 
stereo3 test). Do you have some idea on how to make them more reliable? 
Or shall we give it a try and remove problematic tests when they arise?

Regards,
Tobias