Moved SuicideOnChannelErrorFilter to content.

Moved SuicideOnChannelErrorFilter to content.

BUG=none
TEST=trybots


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=131429

[asharif1, 2012-04-09 19:00:13]

jam, PTAL.

[jam, 2012-04-09 19:08:54]

lgtm, thanks

[commit-bot: I haz the power, 2012-04-09 19:25:22]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/asharif@chromium.org/10027027/1

[commit-bot: I haz the power, 2012-04-09 20:54:18]

Change committed as 131429

[nsylvain, 2012-04-10 00:23:36]

I think this change is causing important problems on the bots. (And I can
reproduce locally)

For example:

chrome-bot@mini44-m4 bash$ ./content_unittests
--gtest_filter=WebRTCAudioDeviceTest.*
Note: Google Test filter = WebRTCAudioDeviceTest.*
[==========] Running 7 tests from 1 test case.
[----------] Global test environment set-up.
[----------] 7 tests from WebRTCAudioDeviceTest
[ RUN      ] WebRTCAudioDeviceTest.TestValidInputRates
chrome-bot@mini44-m4 bash$


For some reasons this test is making content_unittest crash. The worse thing is
that since you exit with 0, it makes content_unittest return 0, which means it's
not an error.  This is causing content_unittests to be green on the bot, and yet
it does not run to completion.

[jam, 2012-04-10 00:40:09]

doh, we should revert this then. sorry Ahmad I didn't expect this. i'll
sync on my linux machine and track it further

reverting now..

On Mon, Apr 9, 2012 at 5:23 PM, <nsylvain@google.com> wrote:

> I think this change is causing important problems on the bots. (And I can
> reproduce locally)
>
> For example:
>
> chrome-bot@mini44-m4 bash$ ./content_unittests
> --gtest_filter=**WebRTCAudioDeviceTest.*
> Note: Google Test filter = WebRTCAudioDeviceTest.*
> [==========] Running 7 tests from 1 test case.
> [----------] Global test environment set-up.
> [----------] 7 tests from WebRTCAudioDeviceTest
> [ RUN      ] WebRTCAudioDeviceTest.**TestValidInputRates
> chrome-bot@mini44-m4 bash$
>
>
> For some reasons this test is making content_unittest crash. The worse
> thing is
> that since you exit with 0, it makes content_unittest return 0, which
> means it's
> not an error.  This is causing content_unittests to be green on the bot,
> and yet
> it does not run to completion.
>
>
http://codereview.chromium.**org/10027027/<http://codereview.chromium.org/100...
>

[asharif1, 2012-04-10 02:48:49]

Sorry about this.

I really think problems like these should be caught by the commit queue
and/or trybots. Should I have done something different than git try?

Why does the CQ/trybots fail because of flakiness but not because of real
problems? I thought the reason we had flaky tests was to extend the
coverage to catch real issues.

On Mon, Apr 9, 2012 at 5:23 PM, <nsylvain@google.com> wrote:

> I think this change is causing important problems on the bots. (And I can
> reproduce locally)
>
> For example:
>
> chrome-bot@mini44-m4 bash$ ./content_unittests
> --gtest_filter=**WebRTCAudioDeviceTest.*
> Note: Google Test filter = WebRTCAudioDeviceTest.*
> [==========] Running 7 tests from 1 test case.
> [----------] Global test environment set-up.
> [----------] 7 tests from WebRTCAudioDeviceTest
> [ RUN      ] WebRTCAudioDeviceTest.**TestValidInputRates
> chrome-bot@mini44-m4 bash$
>
>
> For some reasons this test is making content_unittest crash. The worse
> thing is
> that since you exit with 0, it makes content_unittest return 0, which
> means it's
> not an error.  This is causing content_unittests to be green on the bot,
> and yet
> it does not run to completion.
>
> http://codereview.chromium.**org/10027027/
>

[nsylvain, 2012-04-10 15:50:12]

On Mon, Apr 9, 2012 at 7:48 PM, asharif <asharif@chromium.org> wrote:

> Sorry about this.
>
> I really think problems like these should be caught by the commit queue
> and/or trybots. Should I have done something different than git try?
>
> Why does the CQ/trybots fail because of flakiness but not because of real
> problems? I thought the reason we had flaky tests was to extend the
> coverage to catch real issues.
>

The commit queue/try bots does only 1 thing, "run tests and check if they
return 0 or non-zero", and based on that it will determine if the patch is
good or not.   If the test executable returned 0, then it's a pass.  Let's
make sure we don't return 0 on failure otherwise all our buildbot
infrastructure will be useless.

Nicolas


> On Mon, Apr 9, 2012 at 5:23 PM, <nsylvain@google.com> wrote:
>
>> I think this change is causing important problems on the bots. (And I can
>> reproduce locally)
>>
>> For example:
>>
>> chrome-bot@mini44-m4 bash$ ./content_unittests
>> --gtest_filter=**WebRTCAudioDeviceTest.*
>> Note: Google Test filter = WebRTCAudioDeviceTest.*
>> [==========] Running 7 tests from 1 test case.
>> [----------] Global test environment set-up.
>> [----------] 7 tests from WebRTCAudioDeviceTest
>> [ RUN      ] WebRTCAudioDeviceTest.**TestValidInputRates
>> chrome-bot@mini44-m4 bash$
>>
>>
>> For some reasons this test is making content_unittest crash. The worse
>> thing is
>> that since you exit with 0, it makes content_unittest return 0, which
>> means it's
>> not an error.  This is causing content_unittests to be green on the bot,
>> and yet
>> it does not run to completion.
>>
>> http://codereview.chromium.**org/10027027/
>>
>
>