Update the Opus test hash for OPUS_FIXED_POINT.

Update the Opus test hash for OPUS_FIXED_POINT.

The hash has to be generated from an ARM platform, since the x86 hash does not match.

Committed: https://crrev.com/62379b0de272c8b86174c919e5b50ba66c9e87a3
Cr-Commit-Position: refs/heads/master@{#439525}

[mbjorge, 2016-12-16 22:28:33]

mbjorge@chromium.org changed reviewers:
+ chcunningham@chromium.org

[mbjorge, 2016-12-16 22:28:33]

jc, how did this value get calculated? Is this getting exercised anywhere
successfully?

[chcunningham, 2016-12-16 22:37:52]

On 2016/12/16 22:28:33, mbjorge wrote:
> jc, how did this value get calculated? Is this getting exercised anywhere
> successfully?

I calculated the original hash values by forcing libopus to use fixed point on
my goobuntu machine. I'm surprised to see your hashes differ -- what platform
are you using? Is it expected that fixed point calculations would vary across
platforms? 

Its a known issue that we do not have proper test coverage for arm. I've filed
this bug when I made the initial change:
https://bugs.chromium.org/p/chromium/issues/detail?id=672291

Please start that issue or comment to express your support :)

[mbjorge, 2016-12-16 22:43:03]

On 2016/12/16 at 22:37:52, chcunningham wrote:
> On 2016/12/16 22:28:33, mbjorge wrote:
> > jc, how did this value get calculated? Is this getting exercised anywhere
> > successfully?
> 
> I calculated the original hash values by forcing libopus to use fixed point on
my goobuntu machine. I'm surprised to see your hashes differ -- what platform
are you using? Is it expected that fixed point calculations would vary across
platforms? 
> 
> Its a known issue that we do not have proper test coverage for arm. I've filed
this bug when I made the initial change:
> https://bugs.chromium.org/p/chromium/issues/detail?id=672291
> 
> Please start that issue or comment to express your support :)

This is on the Chromecast devices, which are arm based. I am not sure if it's
expected for there to be differences across platforms though

[chcunningham, 2016-12-16 23:47:04]

On 2016/12/16 22:43:03, mbjorge wrote:
> On 2016/12/16 at 22:37:52, chcunningham wrote:
> > On 2016/12/16 22:28:33, mbjorge wrote:
> > > jc, how did this value get calculated? Is this getting exercised anywhere
> > > successfully?
> > 
> > I calculated the original hash values by forcing libopus to use fixed point
on
> my goobuntu machine. I'm surprised to see your hashes differ -- what platform
> are you using? Is it expected that fixed point calculations would vary across
> platforms? 
> > 
> > Its a known issue that we do not have proper test coverage for arm. I've
filed
> this bug when I made the initial change:
> > https://bugs.chromium.org/p/chromium/issues/detail?id=672291
> > 
> > Please start that issue or comment to express your support :)
> 
> This is on the Chromecast devices, which are arm based. I am not sure if it's
> expected for there to be differences across platforms though

I've just re-run the test on Goobuntu with opus fixed-point forced to true. The
test using the hashes you're chaning all passed (using the checked in hashes). 

Interestingly, I had some other tests fail (details below). Do all pipeline
integration tests pass for you once you've applied this change? Can you
reproduce my goobuntu results? Are you using the same libopus as me (the one
checked in to chromium tip of tree)?


----


3 tests failed:
    PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed
(../../media/test/pipeline_integration_test.cc:1068)
    PipelineIntegrationTest.BasicPlaybackOpusWebmTrimmingHashed
(../../media/test/pipeline_integration_test.cc:1092)
    PipelineIntegrationTest.BasicPlaybackOpusWebmTrimmingHashed_MediaSource
(../../media/test/pipeline_integration_test.cc:1116)

Here's the details from one of the failures

$ out/Default/media_unittests
--gtest_filter=*PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed
IMPORTANT DEBUGGING NOTE: batches of tests are run inside their
own process. For debugging a test inside a debugger, use the
--gtest_filter=<your_test_name> flag along with
--single-process-tests.
Using sharding settings from environment. This is shard 0/1
Using 1 parallel jobs.
Note: Google Test filter =
PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed
[==========] Running 1 test from 1 test case.
[----------] Global test environment set-up.
[----------] 1 test from PipelineIntegrationTest
[ RUN      ] PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed
../../media/test/pipeline_integration_test.cc:1075: Failure
Value of: GetAudioHash()
  Actual: "-4.57,-5.66,-6.52,-6.29,-4.37,-3.60,"
Expected: kOpusEndTrimmingHash_1
Which is: "-4.57,-5.66,-6.52,-6.30,-4.37,-3.61,"
../../media/test/pipeline_integration_test.cc:1081: Failure
Value of: GetAudioHash()
  Actual: "-11.90,-11.10,-8.26,-7.12,-7.85,-9.99,"
Expected: kOpusEndTrimmingHash_2
Which is: "-11.91,-11.11,-8.27,-7.13,-7.86,-10.00,"
../../media/test/pipeline_integration_test.cc:1088: Failure
Value of: GetAudioHash()
  Actual: "-13.30,-14.37,-13.70,-11.69,-10.20,-10.48,"
Expected: kOpusEndTrimmingHash_3
Which is: "-13.31,-14.38,-13.70,-11.71,-10.21,-10.49,"
[  FAILED  ] PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed (135 ms)
[----------] 1 test from PipelineIntegrationTest (135 ms total)

[----------] Global test environment tear-down
[==========] 1 test from 1 test case ran. (135 ms total)
[  PASSED  ] 0 tests.
[  FAILED  ] 1 test, listed below:
[  FAILED  ] PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed

 1 FAILED TEST
[35599:35600:1216/154103.198127:950954610645:ERROR:kill_posix.cc(84)] Unable to
terminate process group 35601: No such process
[1/1] PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed (135 ms)
1 test failed:
    PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed
(../../media/test/pipeline_integration_test.cc:1068)
Tests took 0 seconds.

[mbjorge, 2016-12-17 00:16:55]

On 2016/12/16 at 23:47:04, chcunningham wrote:
> On 2016/12/16 22:43:03, mbjorge wrote:
> > On 2016/12/16 at 22:37:52, chcunningham wrote:
> > > On 2016/12/16 22:28:33, mbjorge wrote:
> > > > jc, how did this value get calculated? Is this getting exercised
anywhere
> > > > successfully?
> > > 
> > > I calculated the original hash values by forcing libopus to use fixed
point on
> > my goobuntu machine. I'm surprised to see your hashes differ -- what
platform
> > are you using? Is it expected that fixed point calculations would vary
across
> > platforms? 
> > > 
> > > Its a known issue that we do not have proper test coverage for arm. I've
filed
> > this bug when I made the initial change:
> > > https://bugs.chromium.org/p/chromium/issues/detail?id=672291
> > > 
> > > Please start that issue or comment to express your support :)
> > 
> > This is on the Chromecast devices, which are arm based. I am not sure if
it's
> > expected for there to be differences across platforms though
> 
> I've just re-run the test on Goobuntu with opus fixed-point forced to true.
The test using the hashes you're chaning all passed (using the checked in
hashes). 
> 
> Interestingly, I had some other tests fail (details below). Do all pipeline
integration tests pass for you once you've applied this change? Can you
reproduce my goobuntu results? Are you using the same libopus as me (the one
checked in to chromium tip of tree)?
> 
> 
> ----
> 
> 
> 3 tests failed:
>     PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed
(../../media/test/pipeline_integration_test.cc:1068)
>     PipelineIntegrationTest.BasicPlaybackOpusWebmTrimmingHashed
(../../media/test/pipeline_integration_test.cc:1092)
>     PipelineIntegrationTest.BasicPlaybackOpusWebmTrimmingHashed_MediaSource
(../../media/test/pipeline_integration_test.cc:1116)
> 
> Here's the details from one of the failures
> 
> $ out/Default/media_unittests
--gtest_filter=*PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed
> IMPORTANT DEBUGGING NOTE: batches of tests are run inside their
> own process. For debugging a test inside a debugger, use the
> --gtest_filter=<your_test_name> flag along with
> --single-process-tests.
> Using sharding settings from environment. This is shard 0/1
> Using 1 parallel jobs.
> Note: Google Test filter =
PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed
> [==========] Running 1 test from 1 test case.
> [----------] Global test environment set-up.
> [----------] 1 test from PipelineIntegrationTest
> [ RUN      ] PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed
> ../../media/test/pipeline_integration_test.cc:1075: Failure
> Value of: GetAudioHash()
>   Actual: "-4.57,-5.66,-6.52,-6.29,-4.37,-3.60,"
> Expected: kOpusEndTrimmingHash_1
> Which is: "-4.57,-5.66,-6.52,-6.30,-4.37,-3.61,"
> ../../media/test/pipeline_integration_test.cc:1081: Failure
> Value of: GetAudioHash()
>   Actual: "-11.90,-11.10,-8.26,-7.12,-7.85,-9.99,"
> Expected: kOpusEndTrimmingHash_2
> Which is: "-11.91,-11.11,-8.27,-7.13,-7.86,-10.00,"
> ../../media/test/pipeline_integration_test.cc:1088: Failure
> Value of: GetAudioHash()
>   Actual: "-13.30,-14.37,-13.70,-11.69,-10.20,-10.48,"
> Expected: kOpusEndTrimmingHash_3
> Which is: "-13.31,-14.38,-13.70,-11.71,-10.21,-10.49,"
> [  FAILED  ] PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed (135
ms)
> [----------] 1 test from PipelineIntegrationTest (135 ms total)
> 
> [----------] Global test environment tear-down
> [==========] 1 test from 1 test case ran. (135 ms total)
> [  PASSED  ] 0 tests.
> [  FAILED  ] 1 test, listed below:
> [  FAILED  ] PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed
> 
>  1 FAILED TEST
> [35599:35600:1216/154103.198127:950954610645:ERROR:kill_posix.cc(84)] Unable
to terminate process group 35601: No such process
> [1/1] PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed (135 ms)
> 1 test failed:
>     PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed
(../../media/test/pipeline_integration_test.cc:1068)
> Tests took 0 seconds.


> Do all pipeline integration tests pass for you once you've applied this
change? 
Yes indeed

> Can you reproduce my goobuntu results? 
Yes I can. I set use_opus_fixed_point = true in third_party/opus/BUILD.gn, built
media_unittests, and
PipelineIntegrationTest.BasicPlaybackOpusPrerollExceedsCodecDelay passes using
the currently checked in value, and I get those same 3 failures.


> Are you using the same libopus as me (the one checked in to chromium tip of
tree)?
Yup, 1.1.3 from third_party/opus

So it seems that the hashes are indeed different between arm vs x86 :/

[chcunningham, 2016-12-19 18:28:03]

On 2016/12/17 00:16:55, mbjorge wrote:
> On 2016/12/16 at 23:47:04, chcunningham wrote:
> > On 2016/12/16 22:43:03, mbjorge wrote:
> > > On 2016/12/16 at 22:37:52, chcunningham wrote:
> > > > On 2016/12/16 22:28:33, mbjorge wrote:
> > > > > jc, how did this value get calculated? Is this getting exercised
> anywhere
> > > > > successfully?
> > > > 
> > > > I calculated the original hash values by forcing libopus to use fixed
> point on
> > > my goobuntu machine. I'm surprised to see your hashes differ -- what
> platform
> > > are you using? Is it expected that fixed point calculations would vary
> across
> > > platforms? 
> > > > 
> > > > Its a known issue that we do not have proper test coverage for arm. I've
> filed
> > > this bug when I made the initial change:
> > > > https://bugs.chromium.org/p/chromium/issues/detail?id=672291
> > > > 
> > > > Please start that issue or comment to express your support :)
> > > 
> > > This is on the Chromecast devices, which are arm based. I am not sure if
> it's
> > > expected for there to be differences across platforms though
> > 
> > I've just re-run the test on Goobuntu with opus fixed-point forced to true.
> The test using the hashes you're chaning all passed (using the checked in
> hashes). 
> > 
> > Interestingly, I had some other tests fail (details below). Do all pipeline
> integration tests pass for you once you've applied this change? Can you
> reproduce my goobuntu results? Are you using the same libopus as me (the one
> checked in to chromium tip of tree)?
> > 
> > 
> > ----
> > 
> > 
> > 3 tests failed:
> >     PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed
> (../../media/test/pipeline_integration_test.cc:1068)
> >     PipelineIntegrationTest.BasicPlaybackOpusWebmTrimmingHashed
> (../../media/test/pipeline_integration_test.cc:1092)
> >     PipelineIntegrationTest.BasicPlaybackOpusWebmTrimmingHashed_MediaSource
> (../../media/test/pipeline_integration_test.cc:1116)
> > 
> > Here's the details from one of the failures
> > 
> > $ out/Default/media_unittests
> --gtest_filter=*PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed
> > IMPORTANT DEBUGGING NOTE: batches of tests are run inside their
> > own process. For debugging a test inside a debugger, use the
> > --gtest_filter=<your_test_name> flag along with
> > --single-process-tests.
> > Using sharding settings from environment. This is shard 0/1
> > Using 1 parallel jobs.
> > Note: Google Test filter =
> PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed
> > [==========] Running 1 test from 1 test case.
> > [----------] Global test environment set-up.
> > [----------] 1 test from PipelineIntegrationTest
> > [ RUN      ] PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed
> > ../../media/test/pipeline_integration_test.cc:1075: Failure
> > Value of: GetAudioHash()
> >   Actual: "-4.57,-5.66,-6.52,-6.29,-4.37,-3.60,"
> > Expected: kOpusEndTrimmingHash_1
> > Which is: "-4.57,-5.66,-6.52,-6.30,-4.37,-3.61,"
> > ../../media/test/pipeline_integration_test.cc:1081: Failure
> > Value of: GetAudioHash()
> >   Actual: "-11.90,-11.10,-8.26,-7.12,-7.85,-9.99,"
> > Expected: kOpusEndTrimmingHash_2
> > Which is: "-11.91,-11.11,-8.27,-7.13,-7.86,-10.00,"
> > ../../media/test/pipeline_integration_test.cc:1088: Failure
> > Value of: GetAudioHash()
> >   Actual: "-13.30,-14.37,-13.70,-11.69,-10.20,-10.48,"
> > Expected: kOpusEndTrimmingHash_3
> > Which is: "-13.31,-14.38,-13.70,-11.71,-10.21,-10.49,"
> > [  FAILED  ] PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed (135
> ms)
> > [----------] 1 test from PipelineIntegrationTest (135 ms total)
> > 
> > [----------] Global test environment tear-down
> > [==========] 1 test from 1 test case ran. (135 ms total)
> > [  PASSED  ] 0 tests.
> > [  FAILED  ] 1 test, listed below:
> > [  FAILED  ] PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed
> > 
> >  1 FAILED TEST
> > [35599:35600:1216/154103.198127:950954610645:ERROR:kill_posix.cc(84)] Unable
> to terminate process group 35601: No such process
> > [1/1] PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed (135 ms)
> > 1 test failed:
> >     PipelineIntegrationTest.BasicPlaybackOpusOggTrimmingHashed
> (../../media/test/pipeline_integration_test.cc:1068)
> > Tests took 0 seconds.
> 
> 
> > Do all pipeline integration tests pass for you once you've applied this
> change? 
> Yes indeed
> 
> > Can you reproduce my goobuntu results? 
> Yes I can. I set use_opus_fixed_point = true in third_party/opus/BUILD.gn,
built
> media_unittests, and
> PipelineIntegrationTest.BasicPlaybackOpusPrerollExceedsCodecDelay passes using
> the currently checked in value, and I get those same 3 failures.
> 
> 
> > Are you using the same libopus as me (the one checked in to chromium tip of
> tree)?
> Yup, 1.1.3 from third_party/opus
> 
> So it seems that the hashes are indeed different between arm vs x86 :/

Bummer. This change LGTM - we don't ever used fixed_point on x86, so I'm not
into making x86 specific defines.

[chcunningham, 2016-12-19 18:28:14]

https://codereview.chromium.org/2579423002/diff/1/media/test/pipeline_integra...
File media/test/pipeline_integration_test.cc (right):

https://codereview.chromium.org/2579423002/diff/1/media/test/pipeline_integra...
media/test/pipeline_integration_test.cc:130: #if defined(OPUS_FIXED_POINT)
Can you add a little comment here?

NOTE: Hashes are specific to ARM devices. x86 will not match.

[mbjorge, 2016-12-19 18:33:39]

On 2016/12/19 at 18:28:14, chcunningham wrote:
>
https://codereview.chromium.org/2579423002/diff/1/media/test/pipeline_integra...
> File media/test/pipeline_integration_test.cc (right):
> 
>
https://codereview.chromium.org/2579423002/diff/1/media/test/pipeline_integra...
> media/test/pipeline_integration_test.cc:130: #if defined(OPUS_FIXED_POINT)
> Can you add a little comment here?
> 
> NOTE: Hashes are specific to ARM devices. x86 will not match.

Done

[mbjorge, 2016-12-19 18:34:25]

Description was changed from

==========
Update the Opus test hash for OPUS_FIXED_POINT.

The hash got out of sync; started breaking on cast devices when we
rolled our chromium version.
==========

to

==========
Update the Opus test hash for OPUS_FIXED_POINT.

The hash has to be generated from an ARM platform, since the x86 hash does not
match.
==========

[mbjorge, 2016-12-19 18:35:00]

The CQ bit was checked by mbjorge@chromium.org

[mbjorge, 2016-12-19 18:35:02]

The patchset sent to the CQ was uploaded after l-g-t-m from
chcunningham@chromium.org
Link to the patchset: https://codereview.chromium.org/2579423002/#ps20001
(title: "Add note that hashes are ARM specific.")

[commit-bot: I haz the power, 2016-12-19 18:35:26]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-12-19 19:41:25]

CQ is committing da patch.

Bot data: {"patchset_id": 20001, "attempt_start_ts": 1482172500922490,
"parent_rev": "4e2e3e19db1489f9f9a8e62a74d5ab30b9356122", "commit_rev":
"998fe02160841d962d495307cc6a3a906d5a652c"}

[commit-bot: I haz the power, 2016-12-19 19:41:53]

Description was changed from

==========
Update the Opus test hash for OPUS_FIXED_POINT.

The hash has to be generated from an ARM platform, since the x86 hash does not
match.
==========

to

==========
Update the Opus test hash for OPUS_FIXED_POINT.

The hash has to be generated from an ARM platform, since the x86 hash does not
match.

Review-Url: https://codereview.chromium.org/2579423002
==========

[commit-bot: I haz the power, 2016-12-19 19:41:54]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2016-12-19 19:45:54]

Description was changed from

==========
Update the Opus test hash for OPUS_FIXED_POINT.

The hash has to be generated from an ARM platform, since the x86 hash does not
match.

Review-Url: https://codereview.chromium.org/2579423002
==========

to

==========
Update the Opus test hash for OPUS_FIXED_POINT.

The hash has to be generated from an ARM platform, since the x86 hash does not
match.

Committed: https://crrev.com/62379b0de272c8b86174c919e5b50ba66c9e87a3
Cr-Commit-Position: refs/heads/master@{#439525}
==========

[commit-bot: I haz the power, 2016-12-19 19:45:55]

Patchset 2 (id:??) landed as
https://crrev.com/62379b0de272c8b86174c919e5b50ba66c9e87a3
Cr-Commit-Position: refs/heads/master@{#439525}