Re-enable VP9 resize test.

Re-enable VP9 resize test.

TBR=stefan@webrtc.org
BUG=webrtc:5097
Committed: https://crrev.com/09b38f3ca055100f38ddc3280a21febb4bac9797
Cr-Commit-Position: refs/heads/master@{#10409}

[marpan, 2015-10-21 17:36:37]

Description was changed from

==========
Re-enable VP9 resize test.

TBR=stefan@webrtc.org
BUG=
==========

to

==========
Re-enable VP9 resize test.

TBR=stefan@webrtc.org
BUG=
==========

[marpan, 2015-10-21 17:39:07]

Description was changed from

==========
Re-enable VP9 resize test.

TBR=stefan@webrtc.org
BUG=
==========

to

==========
Re-enable VP9 resize test.

TBR=stefan@webrtc.org
BUG=webrtc:5097
==========

[stefan-webrtc, 2015-10-22 09:44:14]

https://codereview.webrtc.org/1409143005/diff/100001/webrtc/modules/video_cod...
File webrtc/modules/video_coding/codecs/test/videoprocessor_integrationtest.cc
(right):

https://codereview.webrtc.org/1409143005/diff/100001/webrtc/modules/video_cod...
webrtc/modules/video_coding/codecs/test/videoprocessor_integrationtest.cc:737:
SetRateProfilePars(&rate_profile, 0, 90, 30, 0);
I don't fully understand this change. Could you explain why the target bitrate
changes from 50 to 90?

[pbos-webrtc, 2015-10-22 19:51:50]

On 2015/10/22 09:44:14, stefan-webrtc (holmer) wrote:
>
https://codereview.webrtc.org/1409143005/diff/100001/webrtc/modules/video_cod...
> File webrtc/modules/video_coding/codecs/test/videoprocessor_integrationtest.cc
> (right):
> 
>
https://codereview.webrtc.org/1409143005/diff/100001/webrtc/modules/video_cod...
> webrtc/modules/video_coding/codecs/test/videoprocessor_integrationtest.cc:737:
> SetRateProfilePars(&rate_profile, 0, 90, 30, 0);
> I don't fully understand this change. Could you explain why the target bitrate
> changes from 50 to 90?

The linux_msan error looks like a libvpx bug to me, decoding seems to read from
an uninitialized part of the frame buffer. libvpx should fill this buffer during
decode, right?

[pbos-webrtc, 2015-10-22 19:54:59]

On 2015/10/22 19:51:50, pbos-webrtc wrote:
> On 2015/10/22 09:44:14, stefan-webrtc (holmer) wrote:
> >
>
https://codereview.webrtc.org/1409143005/diff/100001/webrtc/modules/video_cod...
> > File
webrtc/modules/video_coding/codecs/test/videoprocessor_integrationtest.cc
> > (right):
> > 
> >
>
https://codereview.webrtc.org/1409143005/diff/100001/webrtc/modules/video_cod...
> >
webrtc/modules/video_coding/codecs/test/videoprocessor_integrationtest.cc:737:
> > SetRateProfilePars(&rate_profile, 0, 90, 30, 0);
> > I don't fully understand this change. Could you explain why the target
bitrate
> > changes from 50 to 90?
> 
> The linux_msan error looks like a libvpx bug to me, decoding seems to read
from
> an uninitialized part of the frame buffer. libvpx should fill this buffer
during
> decode, right?

To me, it looks like:

    #6 0x7fb850004db9 in vpx_realloc_frame_buffer
chromium/src/third_party/libvpx_new/source/libvpx/vpx_scale/generic/yv12config.c:199:11

Generates the uninitialized memory (because realloc shouldn't clear memory,
unless vpx_realloc_frame_buffer have different semantics), then the loop filter
seems to read from there before the memory is initialized.

[marpan, 2015-10-23 15:57:10]

https://codereview.webrtc.org/1409143005/diff/100001/webrtc/modules/video_cod...
File webrtc/modules/video_coding/codecs/test/videoprocessor_integrationtest.cc
(right):

https://codereview.webrtc.org/1409143005/diff/100001/webrtc/modules/video_cod...
webrtc/modules/video_coding/codecs/test/videoprocessor_integrationtest.cc:737:
SetRateProfilePars(&rate_profile, 0, 90, 30, 0);
On 2015/10/22 09:44:14, stefan-webrtc (holmer) wrote:
> I don't fully understand this change. Could you explain why the target bitrate
> changes from 50 to 90?

resize logic was changed to allow for 3/4 resizing first (before going down to
1/2 if things are still bad), so for this test at 50k, 2 resize actions occur. 

Either needed to increase bitrate and keep 1 resize action (and also keep same
expected RC metric values), or keep the bitrate=50k and change some expected
values in RC metrics(line 749). Went back to the latter change now (as in
patch#3).

[marpan, 2015-10-26 15:22:47]

Committed patchset #8 (id:140001) manually as
09b38f3ca055100f38ddc3280a21febb4bac9797 (presubmit successful).

[commit-bot: I haz the power, 2015-10-26 15:22:56]

Patchset 8 (id:??) landed as
https://crrev.com/09b38f3ca055100f38ddc3280a21febb4bac9797
Cr-Commit-Position: refs/heads/master@{#10409}