DevTools test harness: add logToStderr method on testRunner.

DevTools test harness: add logToStderr method on testRunner.

it is very hard to troubleshoot a flaky asynchronous DevTools test without
being able to log something directly to the console output.
With this patch DevTools window is able to do that. Also logging into
Error stream doesn't skew the normal test output so the flaky test
could be tested in a cycle until the first error.

BUG=

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=271024

[apavlov, 2014-05-14 10:11:54]

"2" used in place of "To" is non-common both in Blink and Chromium (and is
generally not endorsed.)

Also, the "err" part should not be capitalized in "stderr", since it's a
standard stream name.

[loislo, 2014-05-14 10:49:20]

On 2014/05/14 10:11:54, apavlov wrote:
> "2" used in place of "To" is non-common both in Blink and Chromium (and is
> generally not endorsed.)
> 
> Also, the "err" part should not be capitalized in "stderr", since it's a
> standard stream name.

comments addressed

[yurys, 2014-05-14 11:28:18]

Wouldn't console.error produce desired output?

[loislo, 2014-05-14 12:05:00]

On 2014/05/14 11:28:18, yurys wrote:
> Wouldn't console.error produce desired output?

1) console.* methods work only for the inspected page.
2) console.* calls from DevTools page go to the /dev/null.
3) In the current implementation of test harness we collect all console output
and pushes it to the inspected page. And this skew the output.
4) We can't dump protocol messages into the inspected page though protocol
because recursion.
5) if we collect the messages first and push them later when dumping mechanics
is switched off we may loose them in case of crash
also in that case the messages wouldn't be ordered by time.

[loislo, 2014-05-14 12:06:16]

On 2014/05/14 12:05:00, loislo wrote:
> On 2014/05/14 11:28:18, yurys wrote:
> > Wouldn't console.error produce desired output?
> 
> 1) console.* methods work only for the inspected page.
> 2) console.* calls from DevTools page go to the /dev/null.
> 3) In the current implementation of test harness we collect all console output
> and pushes it to the inspected page. And this skew the output.
> 4) We can't dump protocol messages into the inspected page though protocol
> because recursion.
> 5) if we collect the messages first and push them later when dumping mechanics
> is switched off we may loose them in case of crash
> also in that case the messages wouldn't be ordered by time.

wouldn't be ordered by time with the messages from the inspected page.

[yurys, 2014-05-15 12:54:46]

lgtm

[loislo, 2014-05-15 13:10:16]

The CQ bit was checked by loislo@chromium.org

[commit-bot: I haz the power, 2014-05-15 13:10:29]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/loislo@chromium.org/288663003/40001

[commit-bot: I haz the power, 2014-05-15 14:54:13]

FYI, CQ is re-trying this CL (attempt #1).
Please consider checking whether the failures are real,
and report flakes to chrome-troopers@google.com.
The failing builders are:
  chromium_presubmit on tryserver.chromium
(http://build.chromium.org/p/tryserver.chromium/builders/chromium_presubmit/bu...)
  ios_dbg_simulator on tryserver.chromium
(http://build.chromium.org/p/tryserver.chromium/builders/ios_dbg_simulator/bui...)

[commit-bot: I haz the power, 2014-05-15 14:58:07]

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

[commit-bot: I haz the power, 2014-05-15 14:58:07]

Try jobs failed on following builders:
  chromium_presubmit on tryserver.chromium
(http://build.chromium.org/p/tryserver.chromium/builders/chromium_presubmit/bu...)

[pfeldman, 2014-05-15 14:59:11]

https://codereview.chromium.org/288663003/diff/40001/content/shell/renderer/t...
File content/shell/renderer/test_runner/test_runner.cc (right):

https://codereview.chromium.org/288663003/diff/40001/content/shell/renderer/t...
content/shell/renderer/test_runner/test_runner.cc:534: if (runner_)
Is there a reason to not LOG(ERROR) here?

[loislo, 2014-05-16 11:19:15]

On 2014/05/15 14:59:11, pfeldman wrote:
>
https://codereview.chromium.org/288663003/diff/40001/content/shell/renderer/t...
> File content/shell/renderer/test_runner/test_runner.cc (right):
> 
>
https://codereview.chromium.org/288663003/diff/40001/content/shell/renderer/t...
> content/shell/renderer/test_runner/test_runner.cc:534: if (runner_)
> Is there a reason to not LOG(ERROR) here?

The only reason is to keep the method visible in test_runner.h header.
Otherwise it will become a method on a private class in test_runner.cc

[loislo, 2014-05-16 11:30:39]

comments addressed

[pfeldman, 2014-05-16 12:04:24]

lgtm

[loislo, 2014-05-16 12:18:12]

The CQ bit was checked by loislo@chromium.org

[commit-bot: I haz the power, 2014-05-16 12:18:26]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/loislo@chromium.org/288663003/60001

[commit-bot: I haz the power, 2014-05-16 14:06:00]

FYI, CQ is re-trying this CL (attempt #1).
Please consider checking whether the failures are real,
and report flakes to chrome-troopers@google.com.
The failing builders are:
  android_dbg on tryserver.chromium
(http://build.chromium.org/p/tryserver.chromium/builders/android_dbg/builds/18...)
  linux_chromium_chromeos_rel on tryserver.chromium
(http://build.chromium.org/p/tryserver.chromium/builders/linux_chromium_chrome...)

[commit-bot: I haz the power, 2014-05-16 15:11:35]

Change committed as 271024