Revert of Roll libvpx: cc9c0c:8dc6ea

Revert of Roll libvpx: cc9c0c:8dc6ea (patchset #3 id:40001 of https://codereview.chromium.org/845453005/)

Reason for revert:
Suspected to cause multiple tests failure on win vista bot

http://build.chromium.org/p/chromium.win/builders/Vista%20Tests%20%281%29?numbuilds=50
http://build.chromium.org/p/chromium.win/builders/Vista%20Tests%20%281%29/builds/52775

Original issue's description:
> Roll libvpx: cc9c0c:8dc6ea
> 
> Turn on AVX2 optimizations
> https://codereview.chromium.org/863143002
> 
> Add them to GN build file:
> https://codereview.chromium.org/865893004
> 
> Correct CPU detection:
> https://codereview.chromium.org//864393002
> 
> Win GN build flag:
> https://codereview.chromium.org//817293008
> 
> R=tomfinegan@chromium.org
> 
> Committed: https://crrev.com/efd856c2a46810e742f7aba0130d534783db6b2a
> Cr-Commit-Position: refs/heads/master@{#312712}

TBR=tomfinegan@chromium.org,thakis@chromium.org,johannkoenig@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Committed: https://chromium.googlesource.com/chromium/src/+/8ad2a87e9d2ca440383f54a358ef294841242dca

[binjin, 2015-01-23 11:14:26]

Created Revert of Roll libvpx: cc9c0c:8dc6ea

[commit-bot: I haz the power, 2015-01-23 11:15:58]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/869973003/1

[commit-bot: I haz the power, 2015-01-23 11:16:39]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-01-23 11:16:40]

Failed to apply patch for DEPS:
While running git apply --index -3 -p1;
  error: patch failed: DEPS:40
  error: repository lacks the necessary blob to fall back on 3-way merge.
  error: DEPS: patch does not apply

Patch:       DEPS
Index: DEPS
diff --git a/DEPS b/DEPS
index
31aa8b26cf01ca6a3be4635ac8d994cdad3a6809..1db563c1a6c6d6f0facd6fdf73daa2607bbbd7e9
100644
--- a/DEPS
+++ b/DEPS
@@ -40,7 +40,7 @@
   'pdfium_git': 'https://pdfium.googlesource.com',
   'skia_git': 'https://skia.googlesource.com',
   'boringssl_git': 'https://boringssl.googlesource.com',
-  'libvpx_revision': '8dc6ea93de62df4fb4d8bcc6156c3ad4df158bb3',
+  'libvpx_revision': 'cc9c0c5198e416173f480fb1f11e052f5c29a95b',
   'sfntly_revision': '1bdaae8fc788a5ac8936d68bf24f37d977a13dac',
   'skia_revision': '2dca817edb0d741b7d8930009417bf9bd071808d',
   # Three lines of non-changing comments so that

[commit-bot: I haz the power, 2015-01-23 11:17:00]

The CQ bit was unchecked by commit-bot@chromium.org

[binjin, 2015-01-23 11:31:45]

The CQ bit was checked by binjin@chromium.org

[binjin, 2015-01-23 11:31:47]

The CQ bit was unchecked by binjin@chromium.org

[commit-bot: I haz the power, 2015-01-23 11:34:09]

Patchset 2 (id:??) landed as
https://crrev.com/8ad2a87e9d2ca440383f54a358ef294841242dca
Cr-Commit-Position: refs/heads/master@{#312817}

[binjin, 2015-01-23 11:34:16]

Committed patchset #2 (id:40001) manually as
8ad2a87e9d2ca440383f54a358ef294841242dca (presubmit successful).

[Johann, 2015-01-23 16:16:13]

Ugh, is there any way to get access to the bot, or a vista machine? The tests
just time out silently. No idea what's up.

[Johann, 2015-01-23 16:20:40]

lgtm, btw

[binjin, 2015-01-23 16:53:49]

On 2015/01/23 16:16:13, Johann wrote:
> Ugh, is there any way to get access to the bot, or a vista machine? The tests
> just time out silently. No idea what's up.

Just as Michael Achenbach's comment at https://codereview.chromium.org/845453005
, there are another avx change yesterday which breaks the windows xp/vista bots
constantly. Maybe the root causes are indeed similar?

[Johann, 2015-01-23 18:14:32]

On 2015/01/23 16:53:49, binjin wrote:
> On 2015/01/23 16:16:13, Johann wrote:
> > Ugh, is there any way to get access to the bot, or a vista machine? The
tests
> > just time out silently. No idea what's up.
> 
> Just as Michael Achenbach's comment at
https://codereview.chromium.org/845453005
> , there are another avx change yesterday which breaks the windows xp/vista
bots
> constantly. Maybe the root causes are indeed similar?

Unfortunately the fix there appears related to whether they use AVX on the
target. Since the build appears to be passing, the next possibility is that the
AVX instructions that are generated are being run when the should not. I would
expect to see "invalid instruction" crashes in the log, not timeouts. An
experiment would be to simply disable the possibility of using the AVX functions
by manipulating our feature detecting code. Unfortunately the test->verify cycle
is huge because it requires committing the code to the deps branch, doing a deps
roll, and waiting for the xp bots to pick it up and fail. I would prefer to test
a fix.