[FFmpeg-devel] fate & zlib 1.2.6

[Clément Bœsch]

On Fri, Feb 03, 2012 at 08:38:29PM +0100, Clément Bœsch wrote:
> On Fri, Feb 03, 2012 at 08:20:29PM +0100, Reimar Döffinger wrote:
> > On Fri, Feb 03, 2012 at 08:13:24PM +0100, Clément Bœsch wrote:
> > > Upgrading from zlib 1.2.5 to 1.2.6 breaks fate-vsynth1-flashsv2. I tried
> > > to reduce compression level to 6 like it was done recently with the zmbv
> > > test, but it didn't help (still a mismatch between the two versions).
> > 
> > What exactly changes? zmbv disables checksum, too, so this might be an
> > option.
> 
> size and md5: http://pastie.org/3310367
> 
> -compression_level 0 seems enough to fix the issue (size and md5 match),
> but a higher value leads to a failure (of both size and md5).
> 
> [...]
> 
> -- 
> Clément B.

> From 024193492f43aecf3b9ccb570dc9cd0596fea949 Mon Sep 17 00:00:00 2001
> From: =?UTF-8?q?Cl=C3=A9ment=20B=C5=93sch?= <ubitux at gmail.com>
> Date: Fri, 3 Feb 2012 20:28:32 +0100
> Subject: [PATCH] Use compression level 0 for flashsv2 fate test.
> 
> Mismatch between version 1.2.5 and 1.2.6 if compression level > 0.
> ---
>  tests/codec-regression.sh  |    4 ++--
>  tests/ref/vsynth1/flashsv2 |    8 ++++----
>  2 files changed, 6 insertions(+), 6 deletions(-)
> 
> diff --git a/tests/codec-regression.sh b/tests/codec-regression.sh
> index b8d27cb..e0a22f4 100755
> --- a/tests/codec-regression.sh
> +++ b/tests/codec-regression.sh
> @@ -294,8 +294,8 @@ do_video_decoding "" "-pix_fmt yuv420p -sws_flags area+accurate_rnd+bitexact"
>  fi
>  
>  if [ -n "$do_flashsv2" ] ; then
> -do_video_encoding flashsv2.flv "-an -vcodec flashsv2 -sws_flags neighbor+full_chroma_int+accurate_rnd+bitexact -strict experimental"
> -do_video_encoding flashsv2I.flv "-an -vcodec flashsv2 -sws_flags neighbor+full_chroma_int+accurate_rnd+bitexact -strict experimental -g 1"
> +do_video_encoding flashsv2.flv "-an -vcodec flashsv2 -sws_flags neighbor+full_chroma_int+accurate_rnd+bitexact -strict experimental -compression_level 0"
> +do_video_encoding flashsv2I.flv "-an -vcodec flashsv2 -sws_flags neighbor+full_chroma_int+accurate_rnd+bitexact -strict experimental -compression_level 0 -g 1"
>  do_video_decoding "" "-pix_fmt yuv420p -sws_flags area+accurate_rnd+bitexact"
>  fi
>  
> diff --git a/tests/ref/vsynth1/flashsv2 b/tests/ref/vsynth1/flashsv2
> index 6c43bda..63d49f9 100644
> --- a/tests/ref/vsynth1/flashsv2
> +++ b/tests/ref/vsynth1/flashsv2
> @@ -1,6 +1,6 @@
> -ec13d68cbce263f589fe79bcd2e98364 *./tests/data/vsynth1/flashsv2.flv
> -8406137 ./tests/data/vsynth1/flashsv2.flv
> -ed28c0021768629b2675e72bf1632426 *./tests/data/vsynth1/flashsv2I.flv
> -8755723 ./tests/data/vsynth1/flashsv2I.flv
> +8d105780bb79915bc36f96f7a754f289 *./tests/data/vsynth1/flashsv2.flv
> +9371095 ./tests/data/vsynth1/flashsv2.flv
> +b7bdeb5643ce8c1faefca42c316357cd *./tests/data/vsynth1/flashsv2I.flv
> +9371095 ./tests/data/vsynth1/flashsv2I.flv
>  efa88d09115a2e947eff00ee435ba3f3 *./tests/data/flashsv2.vsynth1.out.yuv
>  stddev:    3.47 PSNR: 37.31 MAXDIFF:   49 bytes:  7603200/  7603200
> -- 
> 1.7.9
> 

Also note that actually, the second test with intra only doesn't need the
hack, so I may remove it from the diff chunk.

-- 
Clément B.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 490 bytes
Desc: not available
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20120203/c4260cbf/attachment.asc>