[Ffmpeg-devel] [PATCH] ffserver regression test status

[Aurelien Jacobs]

On Sun, 26 Feb 2006 00:40:13 +0100
Diego Biurrun <diego at biurrun.de> wrote:

> On Sat, Feb 25, 2006 at 12:26:48AM +0100, Jindrich Makovicka wrote:
> > Aurelien Jacobs wrote:
> > > On Wed, 22 Feb 2006 02:45:37 +0100
> > > Michael Niedermayer <michaelni at gmx.at> wrote:
> > > 
> > >>On Wed, Feb 22, 2006 at 12:13:41AM +0100, Aurelien Jacobs wrote:
> > >>
> > >>>I want to summarize the current status of the ffserver regression tests.
> > >>>There's currently several problems.
> > >>>(some informations are the results of some test made by Luca Abeni
> > >>>about one month ago).
> > >>>
> > >>> 1) It's currently all broken since 2005-11-13 03:32:12 UTC
> > >>>    Since then, there was lots of changes which would have needed
> > >>>    to update ffserver.regression.ref, so IMHO the only solution
> > >>>    to this would be to update the md5sums with the current values.
> > >>>    Anyone against it ? Can I do it ?
> > >>
> > >>do whatever you like with ffserver, i wont touch it until the regression
> > >>tests at least pass most of the time on 2 different x86 linux debian systems
> > > 
> > > Ok, now the ffserver regression tests are almost working.
> > 
> > The only issue is that they are still completely bogus, unless the goal
> > of ffserver is to produce no stream, only the header. If this is the
> > case, I vote for the simplification proposed recently :)
> > 
> > The problem is that the last stream definition in test.conf doesn't work
> > - there is no MJPEG encoder. To fix it, one has to change the default
> > codec in mpjpeg to LJPEG, and set Strict -1 for the last stream to make
> > it accept the colorspace. After that, the regression test starts to
> > produce something.
> 
> > > Ok, now the ffserver regression tests are almost working.
> > 
> > The only issue is that they are still completely bogus, unless the
> > goal of ffserver is to produce no stream, only the header. If this is
> > the case, I vote for the simplification proposed recently :)
> > 
> > The problem is that the last stream definition in test.conf doesn't
> > work - there is no MJPEG encoder. To fix it, one has to change the
> > default codec in mpjpeg to LJPEG, and set Strict -1 for the last
> > stream to make it accept the colorspace. After that, the regression
> > test starts to produce something.
> 
> Still not perfect in any case, though.  On PPC I get without your patch:
> 
> 127.0.0.1 - - [Sun Feb 26 00:33:51 2006] "GET
> /test.mjpg?date=19700101T000000Z HTTP/1.0" 200 107
> --- data/ffserver.regression    2006-02-26 00:33:51.000000000 +0100
> +++ /home/diego/src/ffmpeg/tests/ffserver.regression.ref
> 2006-02-24 16:57:59.000000000 +0100
> @@ -2,7 +2,7 @@
>  ef4887809ddeb782ed66bf34cb506dcf  ff-test_l.avi
>  d41d8cd98f00b204e9800998ecf8427e  ff-test_h.mpg
>  d41d8cd98f00b204e9800998ecf8427e  ff-test_l.mpg
> -0c4803b8a7e8e7beba1b0fb3979cc5bd  ff-test.swf
> +0f1795f0496f37aeea6272d272140efa  ff-test.swf
>  f842864a654593ea9ee9d7a1bd024627  ff-test_h.asf
>  5c12fa55e6b7fbade1b142ecf430a679  ff-test_l.asf
>  71bfe6784e025f6eea9f92113d8c7dda  ff-test_h.rm
> 
> Server regression test: Error.
> make: *** [test-server] Error 1
> 
> 
> and with your patch:
> 
> 
> 127.0.0.1 - - [Sun Feb 26 00:34:58 2006] "GET
> /test_small.jpg?date=19700101T000000Z HTTP/1.0" 200 109783
> /home/diego/src/ffmpeg/tests/server-regression.sh: line 41: 20031
> Terminated              ../ffserver -d -f data/test.conf 2>/dev/null
> --- data/ffserver.regression    2006-02-26 00:34:58.000000000 +0100
> +++ /home/diego/src/ffmpeg/tests/ffserver.regression.ref
> 2006-02-24 16:57:59.000000000 +0100
> @@ -1,12 +1,12 @@
> -bc8536e736a8ac7e6bf413c8e045ae71  ff-test_h.avi
> -7c815102170cf9c34e582252e928dc89  ff-test_l.avi
> +03867bfc9f7dd494cbe2b71aa0ce6537  ff-test_h.avi
> +ef4887809ddeb782ed66bf34cb506dcf  ff-test_l.avi
>  d41d8cd98f00b204e9800998ecf8427e  ff-test_h.mpg
>  d41d8cd98f00b204e9800998ecf8427e  ff-test_l.mpg
> -45e62ee9700e6b9967d8c184bd599391  ff-test.swf
> -866dcc41cd4d1b208be94706527d0774  ff-test_h.asf
> -ff697fdbd9a16aa90168c703b1a425a0  ff-test_l.asf
> -821889bfde2ce05cc66b0ab4a49e11b9  ff-test_h.rm
> -699cae45889e887ce150805405203587  ff-test_l.rm
> -d33513448925faabeed883dc8cc2e2ab  ff-test.jpg
> -963e3e7c2f67e4b4c7545d7e6cb0d485  ff-test_small.jpg
> -526e52492813910331eec4c9d3a79140  ff-test.mjpg
> +0f1795f0496f37aeea6272d272140efa  ff-test.swf
> +f842864a654593ea9ee9d7a1bd024627  ff-test_h.asf
> +5c12fa55e6b7fbade1b142ecf430a679  ff-test_l.asf
> +71bfe6784e025f6eea9f92113d8c7dda  ff-test_h.rm
> +d5c1cc12240a4169abea3fb0fc7f4db3  ff-test_l.rm
> +d41d8cd98f00b204e9800998ecf8427e  ff-test.jpg
> +d41d8cd98f00b204e9800998ecf8427e  ff-test_small.jpg
> +6ce0af834ee0abf4753df419f67648c5  ff-test.mjpg

x86 gives different results for all the above tests except test.mjpg
which gives the same result as ppc...
Now the big question. Why the hell does adding Strict -1 in the
test.mjpg section affect all the other tests ??

Aurel