[FFmpeg-devel] PowerPC regression tests

[Kostya]

On Sat, Feb 14, 2009 at 04:33:07AM +0100, Diego Biurrun wrote:
> I've done a bunch of bisecting and come to the following conclusions:
> 
> The trouble starts with r15305, the revision that switches the regression
> tests to swscale.  codectest and libavtest and seektest work fine in
> r15304.
> 
> In r15305 codectest gives the following diff:
> 
> diff -u -w "/tmp/ffmpeg"/tests/ffmpeg.regression.ref   tests/data/vsynth.regression
> --- /tmp/ffmpeg/tests/ffmpeg.regression.ref     2009-02-14 04:03:25.000000000 +0100
> +++ tests/data/vsynth.regression        2009-02-14 04:10:10.000000000 +0100
> @@ -150,9 +150,9 @@
>  3520692 ./tests/data/a-snow53.avi
>  799d3db687f6cdd7a837ec156efc171f *./tests/data/snowll.vsynth.out.yuv
>  stddev:    0.00 PSNR:999.99 bytes:  7603200/  7603200
> -9bfab037b2779eef4b95773788bbe369 *./tests/data/a-dv.dv
> +5321fc00d99defa9d2fcc9657a6a5921 *./tests/data/a-dv.dv
>  7200000 ./tests/data/a-dv.dv
> -eed5e51bbb178e0a54853f1b9878b841 *./tests/data/dv.vsynth.out.yuv
> +0dc46930dcd133f24f3a479df4c1c8fb *./tests/data/dv.vsynth.out.yuv
>  stddev:    6.90 PSNR: 31.34 bytes:  7603200/  7603200
>  d601c05c2c2bc7153ba687dadb2ef51c *./tests/data/a-dv411.dv
>  7200000 ./tests/data/a-dv411.dv
> make: *** [codectest] Error 1
> 
> It's the same diff that occurs in HEAD.  libavtest and seektest also fail.
> I have attached the diff for both, libavtest diff is the same as in HEAD,
> I have not compared seektest, but I expect it to be the same.
> 
> Diego

Looks like there is a problem with swscale. When I hack swscale.c to not include
AltiVec and compile swScale_C instead regtests pass.