[FFmpeg-devel] [PATCH] fate: add decimate test.

[Nicolas George]

Le quintidi 5 brumaire, an CCXXIV, Michael Niedermayer a écrit :
> seems failing as 32bit on 64bit x86 linux (works on 64bit)
> 
> --- ffmpeg/tests/ref/fate/filter-decimate-simple       2015-10-25 23:19:29.618103426 +0100
> +++ tests/data/fate/filter-decimate-simple      2015-10-26 00:02:08.838157343 +0100
> @@ -2,17 +2,17 @@
>  0,          1,          1,        1,   115200, 0x5e0f54f3
>  0,          2,          2,        1,   115200, 0xd5c25c93
>  0,          3,          3,        1,   115200, 0x372c655c
> -0,          5,          5,        1,   115200, 0xedbaed7f
> +0,          4,          4,        1,   115200, 0xedbaed7f
>  0,          6,          6,        1,   115200, 0x8b2a40a6
>  0,          7,          7,        1,   115200, 0x4e0547fe
> -0,          9,          9,        1,   115200, 0xca5a3761
> +0,          8,          8,        1,   115200, 0xca5a3761
>  0,         10,         10,        1,   115200, 0xe7562c6f
>  0,         11,         11,        1,   115200, 0x709f3960
> -0,         13,         13,        1,   115200, 0x2f76d8f2
> +0,         12,         12,        1,   115200, 0x2f76d8f2
>  0,         14,         14,        1,   115200, 0x72045ce1
>  0,         15,         15,        1,   115200, 0xa6b36e8e
> -0,         17,         17,        1,   115200, 0x275a0a3a
> +0,         16,         16,        1,   115200, 0x275a0a3a
>  0,         18,         18,        1,   115200, 0x87f04127
>  0,         19,         19,        1,   115200, 0x442d4f33
> -0,         21,         21,        1,   115200, 0x8e892c86
> +0,         20,         20,        1,   115200, 0x8e892c86
>  0,         22,         22,        1,   115200, 0xb214258f
> Test filter-decimate-simple failed. Look at tests/data/fate/filter-decimate-simple.err for details.
> make: *** [fate-filter-decimate-simple] Error 1

Thanks for catching this. It selects the same frame, but gives it a
different timestamp. I just noticed the timestamp computations are made
with doubles, and does not take the input frame timestamps into account. I
would say that this filter is broken in several ways, there is no need for
testing it before it is fixed. Patch dropped. 

Regards,

-- 
  Nicolas George
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20151029/c46c5bc0/attachment.sig>