Fix breakages while performance_browser_tests was not running on Mac.

Fix breakages while performance_browser_tests was not running on Mac.

In our vain attempt to re-enable performance_browser_tests on the Mac
bots, a number of breakages seem to have happened over the months where
we were not even aware the tests weren't running.

This change fixes all of the following:

1. Adds a BUILD dependency on //chrome:chrome_app. Without this, the
test binary would be missing many runtime dependencies (e.g., the whole
"Chrome Framework" bundle).

2. Migrate MachPortsTest to use BrowserTestBase::embedded_test_server()
instead of its own server of "fake content" for its browser tests.

3. Add base::ScopedAllowIO for a test utility class that creates and
tears down its own set of threads. It's not clear why the thread
restriction is only checked on Mac (perhaps it's in the way the
browser test's main thread is set up?). Nevertheless, it's perfectly
reasonable for the main thread to block on stopping these other threads:
After the test has run, the threads are shut down before the data
analysis begins, since we want to be sure the data sets are no longer
mutating. Also, added PRESUBMIT.py exception for this use case.

BUG=697444, 722367
TBR=nduca

Review-Url: https://codereview.chromium.org/2901113003
Cr-Commit-Position: refs/heads/master@{#475773}
Committed: https://chromium.googlesource.com/chromium/src/+/8e0e80c3608ac2a08e87ead7d103578f6751f2f6

[miu, 2017-05-24 02:45:44]

miu@chromium.org changed reviewers:
+ cmp@chromium.org, phajdan.jr@chromium.org, xjz@chromium.org

[miu, 2017-05-24 02:45:44]

cmp: PTAL at chrome/test/perf/mach_ports_performancetest.cc. Yes, your test has
not been running for a long time (see crbug for details).

phajdan.jr: PTAL at PRESUBMIT.py exception and chrome/test/BUILD.gn change.

xjz: PTAL at media/cast changes.

[miu, 2017-05-24 02:45:56]

The CQ bit was checked by miu@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2017-05-24 02:47:07]

Dry run: 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, 2017-05-24 06:47:58]

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

[commit-bot: I haz the power, 2017-05-24 06:47:59]

Dry run: Try jobs failed on following builders:
  linux_android_rel_ng on master.tryserver.chromium.android (JOB_FAILED,
https://build.chromium.org/p/tryserver.chromium.android/builders/linux_androi...)

[xjz, 2017-05-24 17:35:15]

media/cast LGTM.

[Paweł Hajdan Jr., 2017-05-26 15:21:52]

LGTM

Note cmp@ is likely inactive at this point.

[miu, 2017-05-30 20:15:40]

miu@chromium.org changed reviewers:
+ nduca@chromium.org
- cmp@chromium.org

[miu, 2017-05-30 20:15:41]

Replaced cmp with nduca (for chrome/test/perf/...). Also, due to the fact that
this has been keeping the bots red for so long (and the fix is trivial), I'm
going to submit TBR now.

[miu, 2017-05-30 20:15:52]

Description was changed from

==========
Fix breakages while performance_browser_tests was not running on Mac.

In our vain attempt to re-enable performance_browser_tests on the Mac
bots, a number of breakages seem to have happened over the months where
we were not even aware the tests weren't running.

This change fixes all of the following:

1. Adds a BUILD dependency on //chrome:chrome_app. Without this, the
test binary would be missing many runtime dependencies (e.g., the whole
"Chrome Framework" bundle).

2. Migrate MachPortsTest to use BrowserTestBase::embedded_test_server()
instead of its own server of "fake content" for its browser tests.

3. Add base::ScopedAllowIO for a test utility class that creates and
tears down its own set of threads. It's not clear why the thread
restriction is only checked on Mac (perhaps it's in the way the
browser test's main thread is set up?). Nevertheless, it's perfectly
reasonable for the main thread to block on stopping these other threads:
After the test has run, the threads are shut down before the data
analysis begins, since we want to be sure the data sets are no longer
mutating. Also, added PRESUBMIT.py exception for this use case.

BUG=697444,722367
==========

to

==========
Fix breakages while performance_browser_tests was not running on Mac.

In our vain attempt to re-enable performance_browser_tests on the Mac
bots, a number of breakages seem to have happened over the months where
we were not even aware the tests weren't running.

This change fixes all of the following:

1. Adds a BUILD dependency on //chrome:chrome_app. Without this, the
test binary would be missing many runtime dependencies (e.g., the whole
"Chrome Framework" bundle).

2. Migrate MachPortsTest to use BrowserTestBase::embedded_test_server()
instead of its own server of "fake content" for its browser tests.

3. Add base::ScopedAllowIO for a test utility class that creates and
tears down its own set of threads. It's not clear why the thread
restriction is only checked on Mac (perhaps it's in the way the
browser test's main thread is set up?). Nevertheless, it's perfectly
reasonable for the main thread to block on stopping these other threads:
After the test has run, the threads are shut down before the data
analysis begins, since we want to be sure the data sets are no longer
mutating. Also, added PRESUBMIT.py exception for this use case.

BUG=697444,722367
TBR=nduca
==========

[miu, 2017-05-30 20:15:59]

The CQ bit was checked by miu@chromium.org

[commit-bot: I haz the power, 2017-05-30 20:16:31]

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, 2017-05-30 23:38:44]

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

[commit-bot: I haz the power, 2017-05-30 23:38:45]

Try jobs failed on following builders:
  linux_android_rel_ng on master.tryserver.chromium.android (JOB_FAILED,
https://build.chromium.org/p/tryserver.chromium.android/builders/linux_androi...)

[miu, 2017-05-31 01:38:38]

The CQ bit was checked by miu@chromium.org

[commit-bot: I haz the power, 2017-05-31 01:39:36]

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, 2017-05-31 03:36:00]

CQ is committing da patch.

Bot data: {"patchset_id": 1, "attempt_start_ts": 1496194718476560, "parent_rev":
"4ac11ed7d02420bda817ec1441af56f2b56a0a02", "commit_rev":
"8e0e80c3608ac2a08e87ead7d103578f6751f2f6"}

[commit-bot: I haz the power, 2017-05-31 03:36:10]

Description was changed from

==========
Fix breakages while performance_browser_tests was not running on Mac.

In our vain attempt to re-enable performance_browser_tests on the Mac
bots, a number of breakages seem to have happened over the months where
we were not even aware the tests weren't running.

This change fixes all of the following:

1. Adds a BUILD dependency on //chrome:chrome_app. Without this, the
test binary would be missing many runtime dependencies (e.g., the whole
"Chrome Framework" bundle).

2. Migrate MachPortsTest to use BrowserTestBase::embedded_test_server()
instead of its own server of "fake content" for its browser tests.

3. Add base::ScopedAllowIO for a test utility class that creates and
tears down its own set of threads. It's not clear why the thread
restriction is only checked on Mac (perhaps it's in the way the
browser test's main thread is set up?). Nevertheless, it's perfectly
reasonable for the main thread to block on stopping these other threads:
After the test has run, the threads are shut down before the data
analysis begins, since we want to be sure the data sets are no longer
mutating. Also, added PRESUBMIT.py exception for this use case.

BUG=697444,722367
TBR=nduca
==========

to

==========
Fix breakages while performance_browser_tests was not running on Mac.

In our vain attempt to re-enable performance_browser_tests on the Mac
bots, a number of breakages seem to have happened over the months where
we were not even aware the tests weren't running.

This change fixes all of the following:

1. Adds a BUILD dependency on //chrome:chrome_app. Without this, the
test binary would be missing many runtime dependencies (e.g., the whole
"Chrome Framework" bundle).

2. Migrate MachPortsTest to use BrowserTestBase::embedded_test_server()
instead of its own server of "fake content" for its browser tests.

3. Add base::ScopedAllowIO for a test utility class that creates and
tears down its own set of threads. It's not clear why the thread
restriction is only checked on Mac (perhaps it's in the way the
browser test's main thread is set up?). Nevertheless, it's perfectly
reasonable for the main thread to block on stopping these other threads:
After the test has run, the threads are shut down before the data
analysis begins, since we want to be sure the data sets are no longer
mutating. Also, added PRESUBMIT.py exception for this use case.

BUG=697444,722367
TBR=nduca

Review-Url: https://codereview.chromium.org/2901113003
Cr-Commit-Position: refs/heads/master@{#475773}
Committed:
https://chromium.googlesource.com/chromium/src/+/8e0e80c3608ac2a08e87ead7d103...
==========

[commit-bot: I haz the power, 2017-05-31 03:36:11]

Committed patchset #1 (id:1) as
https://chromium.googlesource.com/chromium/src/+/8e0e80c3608ac2a08e87ead7d103...