Add support creating and saving a new reference file.

Add support creating and saving a new reference file.

Since many tests switched from using finishAudioTest to doing a manual
comparison between the actual results and the expected results in a
wav file, we lost the ability to create easily the reference wav file.

This adds a simple method to allow saving the actual result to a
file.  This is intended to be run in a browser where the actual result
is automatically downloaded for the developer to examine.  This is part
of the test, so it's automatic. When run as a layout test, no file is saved.

This has no effect on the layout tests, so each affected test was
run manually in a browser to see that the file was generated and a
quick check was done comparing the new result with the expected result.

For the record on a linux machine:

detune-modulation: max error 1 bit, 95 differences.
playback-modulation: max error 1 bit, 15 differences.
gain: no differences
webm-decode: max error 1 bit, 465 differences
osc-custom: no differences
osc-sawtooth: no differences
osc-sine: no differences
osc-square: no differences
osc-triangle: no differences


BUG=696773
TEST=none

Review-Url: https://codereview.chromium.org/2728613003
Cr-Commit-Position: refs/heads/master@{#454693}
Committed: https://chromium.googlesource.com/chromium/src/+/9d77b75aa9748aa22b1136823e0e4d1ace6afd6e

[Raymond Toy, 2017-03-01 18:31:52]

rtoy@chromium.org changed reviewers:
+ hongchan@chromium.org

[Raymond Toy, 2017-03-01 18:31:53]

PTAL at this WIP

Which style do you like? I think I like the defined task to generate the
reference. It's more work to do this, but it's more obvious on how to generate
the reference file.

I manually tested saving the reference file in both cases.  The generated
reference is different from the existing expected file, but I think the new
reference would still pass the tests.  At least on my linux machine.

[hongchan, 2017-03-03 15:53:54]

Looks good. I also prefer to have a task rather than a commented-out function.

[Raymond Toy, 2017-03-03 18:31:49]

PTAL at this updated version where we always save the file when run from a
browser (testRunner undefined).

If you like this, I'll go and add this to all the other tests have need expected
results.

[hongchan, 2017-03-03 18:39:25]

https://codereview.chromium.org/2728613003/diff/20001/third_party/WebKit/Layo...
File
third_party/WebKit/LayoutTests/webaudio/AudioBufferSource/audiobuffersource-playbackrate-modulation.html
(right):

https://codereview.chromium.org/2728613003/diff/20001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/webaudio/AudioBufferSource/audiobuffersource-playbackrate-modulation.html:64:
}
This section can be abstracted a function:

bool generateReferenceFileForBrowser(buffer, name);

[hongchan, 2017-03-03 18:39:57]

On 2017/03/03 18:39:25, hongchan wrote:
>
https://codereview.chromium.org/2728613003/diff/20001/third_party/WebKit/Layo...
> File
>
third_party/WebKit/LayoutTests/webaudio/AudioBufferSource/audiobuffersource-playbackrate-modulation.html
> (right):
> 
>
https://codereview.chromium.org/2728613003/diff/20001/third_party/WebKit/Layo...
>
third_party/WebKit/LayoutTests/webaudio/AudioBufferSource/audiobuffersource-playbackrate-modulation.html:64:
> }
> This section can be abstracted a function:
> 
> bool generateReferenceFileForBrowser(buffer, name);

I meant to be abstracted "into" a function.

[Raymond Toy, 2017-03-03 18:43:10]

https://codereview.chromium.org/2728613003/diff/20001/third_party/WebKit/Layo...
File
third_party/WebKit/LayoutTests/webaudio/AudioBufferSource/audiobuffersource-playbackrate-modulation.html
(right):

https://codereview.chromium.org/2728613003/diff/20001/third_party/WebKit/Layo...
third_party/WebKit/LayoutTests/webaudio/AudioBufferSource/audiobuffersource-playbackrate-modulation.html:64:
}
On 2017/03/03 18:39:25, hongchan wrote:
> This section can be abstracted a function:
> 
> bool generateReferenceFileForBrowser(buffer, name);

Yes and no.  For old Audit, Should is just a function.  But for new Audit, we'd
have to pass in the should function.  Just pass in Should (or should) as an
additional arg?

[hongchan, 2017-03-03 18:47:24]

On 2017/03/03 18:43:10, Raymond Toy wrote:
>
https://codereview.chromium.org/2728613003/diff/20001/third_party/WebKit/Layo...
> File
>
third_party/WebKit/LayoutTests/webaudio/AudioBufferSource/audiobuffersource-playbackrate-modulation.html
> (right):
> 
>
https://codereview.chromium.org/2728613003/diff/20001/third_party/WebKit/Layo...
>
third_party/WebKit/LayoutTests/webaudio/AudioBufferSource/audiobuffersource-playbackrate-modulation.html:64:
> }
> On 2017/03/03 18:39:25, hongchan wrote:
> > This section can be abstracted a function:
> > 
> > bool generateReferenceFileForBrowser(buffer, name);
> 
> Yes and no.  For old Audit, Should is just a function.  But for new Audit,
we'd
> have to pass in the should function.  Just pass in Should (or should) as an
> additional arg?

if (generateRefereceFileForBrowser(buffer, name)) {
  Should()...
}

Can't we do something like this? I think we should move away from passing task
or should into the utility methods.

[Raymond Toy, 2017-03-03 18:51:11]

On 2017/03/03 18:47:24, hongchan wrote:
> On 2017/03/03 18:43:10, Raymond Toy wrote:
> >
>
https://codereview.chromium.org/2728613003/diff/20001/third_party/WebKit/Layo...
> > File
> >
>
third_party/WebKit/LayoutTests/webaudio/AudioBufferSource/audiobuffersource-playbackrate-modulation.html
> > (right):
> > 
> >
>
https://codereview.chromium.org/2728613003/diff/20001/third_party/WebKit/Layo...
> >
>
third_party/WebKit/LayoutTests/webaudio/AudioBufferSource/audiobuffersource-playbackrate-modulation.html:64:
> > }
> > On 2017/03/03 18:39:25, hongchan wrote:
> > > This section can be abstracted a function:
> > > 
> > > bool generateReferenceFileForBrowser(buffer, name);
> > 
> > Yes and no.  For old Audit, Should is just a function.  But for new Audit,
> we'd
> > have to pass in the should function.  Just pass in Should (or should) as an
> > additional arg?
> 
> if (generateRefereceFileForBrowser(buffer, name)) {
>   Should()...
> }
> 
> Can't we do something like this? I think we should move away from passing task
> or should into the utility methods.

This works.

[Raymond Toy, 2017-03-03 19:28:40]

Description was changed from

==========
Add support creating and saving a new reference file.

Since many tests switched from using finishAudioTest to doing a manual
comparison between the actual results and the expected results in a
wav file, we lost the ability to create easily the reference wav file.

This adds a simple method to allow saving the actual result to a
file.  This is intended to be run in a browser and the developer needs
to use the developer console to run downloadAudioData to save the
result.  It is expected that due care is taken to make sure the
reference file is truly the new expected result.

Any tests modified did not change functionality; they were modified
with comments on how to create the reference and updated if necessary
to make the actual results available.

BUG=696773
TEST=manually created result files for the tests that were modified herein.
==========

to

==========
Add support creating and saving a new reference file.

Since many tests switched from using finishAudioTest to doing a manual
comparison between the actual results and the expected results in a
wav file, we lost the ability to create easily the reference wav file.

This adds a simple method to allow saving the actual result to a
file.  This is intended to be run in a browser where the actual result
is automatically downloaded for the developer to examine.  This is part
of the test, so it's automatic. When run as a layout test, no file is saved.

This has no effect on the layout tests, so each affected test was
run manually in a browser to see that the file was generated and a
quick check was done comparing the new result with the expected result.

BUG=696773
TEST=none
==========

[Raymond Toy, 2017-03-03 19:51:34]

Description was changed from

==========
Add support creating and saving a new reference file.

Since many tests switched from using finishAudioTest to doing a manual
comparison between the actual results and the expected results in a
wav file, we lost the ability to create easily the reference wav file.

This adds a simple method to allow saving the actual result to a
file.  This is intended to be run in a browser where the actual result
is automatically downloaded for the developer to examine.  This is part
of the test, so it's automatic. When run as a layout test, no file is saved.

This has no effect on the layout tests, so each affected test was
run manually in a browser to see that the file was generated and a
quick check was done comparing the new result with the expected result.

BUG=696773
TEST=none
==========

to

==========
Add support creating and saving a new reference file.

Since many tests switched from using finishAudioTest to doing a manual
comparison between the actual results and the expected results in a
wav file, we lost the ability to create easily the reference wav file.

This adds a simple method to allow saving the actual result to a
file.  This is intended to be run in a browser where the actual result
is automatically downloaded for the developer to examine.  This is part
of the test, so it's automatic. When run as a layout test, no file is saved.

This has no effect on the layout tests, so each affected test was
run manually in a browser to see that the file was generated and a
quick check was done comparing the new result with the expected result.

For the record on a linux machine:

detune-modulation: max error 1 bit, 95 differences.
playback-modulation: max error 1 bit, 15 differences.
gain: no differences
webm-decode: max error 1 bit, 465 differences
osc-custom: no differences
osc-sawtooth: no differences
osc-sine: no differences
osc-square: no differences
osc-triangle: no differences


BUG=696773
TEST=none
==========

[Raymond Toy, 2017-03-03 19:53:42]

PTAL. I've changed all tests using testharness and also requiring a reference
file.

[hongchan, 2017-03-03 20:15:22]

lgtm

[Raymond Toy, 2017-03-03 21:47:01]

The CQ bit was checked by rtoy@chromium.org

[commit-bot: I haz the power, 2017-03-03 21:47:15]

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, 2017-03-03 22:46:48]

CQ is committing da patch.

Bot data: {"patchset_id": 100001, "attempt_start_ts": 1488577621373820,
"parent_rev": "853b5ae0c2f47d981dd8b390d321a886e465c005", "commit_rev":
"9d77b75aa9748aa22b1136823e0e4d1ace6afd6e"}

[commit-bot: I haz the power, 2017-03-03 22:47:26]

Description was changed from

==========
Add support creating and saving a new reference file.

Since many tests switched from using finishAudioTest to doing a manual
comparison between the actual results and the expected results in a
wav file, we lost the ability to create easily the reference wav file.

This adds a simple method to allow saving the actual result to a
file.  This is intended to be run in a browser where the actual result
is automatically downloaded for the developer to examine.  This is part
of the test, so it's automatic. When run as a layout test, no file is saved.

This has no effect on the layout tests, so each affected test was
run manually in a browser to see that the file was generated and a
quick check was done comparing the new result with the expected result.

For the record on a linux machine:

detune-modulation: max error 1 bit, 95 differences.
playback-modulation: max error 1 bit, 15 differences.
gain: no differences
webm-decode: max error 1 bit, 465 differences
osc-custom: no differences
osc-sawtooth: no differences
osc-sine: no differences
osc-square: no differences
osc-triangle: no differences


BUG=696773
TEST=none
==========

to

==========
Add support creating and saving a new reference file.

Since many tests switched from using finishAudioTest to doing a manual
comparison between the actual results and the expected results in a
wav file, we lost the ability to create easily the reference wav file.

This adds a simple method to allow saving the actual result to a
file.  This is intended to be run in a browser where the actual result
is automatically downloaded for the developer to examine.  This is part
of the test, so it's automatic. When run as a layout test, no file is saved.

This has no effect on the layout tests, so each affected test was
run manually in a browser to see that the file was generated and a
quick check was done comparing the new result with the expected result.

For the record on a linux machine:

detune-modulation: max error 1 bit, 95 differences.
playback-modulation: max error 1 bit, 15 differences.
gain: no differences
webm-decode: max error 1 bit, 465 differences
osc-custom: no differences
osc-sawtooth: no differences
osc-sine: no differences
osc-square: no differences
osc-triangle: no differences


BUG=696773
TEST=none

Review-Url: https://codereview.chromium.org/2728613003
Cr-Commit-Position: refs/heads/master@{#454693}
Committed:
https://chromium.googlesource.com/chromium/src/+/9d77b75aa9748aa22b1136823e0e...
==========

[commit-bot: I haz the power, 2017-03-03 22:47:27]

Committed patchset #6 (id:100001) as
https://chromium.googlesource.com/chromium/src/+/9d77b75aa9748aa22b1136823e0e...