Blacklist all builders who specify add_tests_as_compile_targets=False.

Blacklist all builders who specify add_tests_as_compile_targets=False.

Builders which specify add_tests_as_compile_targets=False in their recipes
causes targets not listed inside build/all.gyp to not be built. This is a
problem with isolated testing. Bug about fixing this is http://crbug.com/503241.
Once fixed, http://crbug.com/498446 can be addressed and this CL can be
reverted, so that all tests can be archived so they can be run via Swarming.

R=thakis@chromium.org
BUG=98637

[M-A Ruel, 2015-06-22 20:23:59]

Interestingly, CrWinAsan doesn't specify add_tests_as_compile_targets=False so I
didn't add it.

The goal is to not break these builders while isolating the rest of the tests
via automated tools. Once it's possible to build all targets properly, we'll add
these back.

[Nico, 2015-06-23 16:53:25]

I'm not sure if this is necessary. The majority of test binaries are explicitly
referenced from build/all.gyp. The ones that aren't, should be. It's ok if I'm
on the hook for making sure this is true until http://crbug.com/503241 is done
(this also gives me an incentive to fix that bug). Does that sound ok to you?

On 2015/06/22 20:23:59, M-A Ruel wrote:
> Interestingly, CrWinAsan doesn't specify add_tests_as_compile_targets=False so
I
> didn't add it.

It seems to build the default target ("") anyways:
http://build.chromium.org/p/chromium.fyi/builders/CrWinClang/builds/1563/step...

> The goal is to not break these builders while isolating the rest of the tests
> via automated tools. Once it's possible to build all targets properly, we'll
add
> these back.

[M-A Ruel, 2015-06-23 17:26:37]

On 2015/06/23 16:53:25, Nico wrote:
> I'm not sure if this is necessary. The majority of test binaries are
explicitly
> referenced from build/all.gyp. The ones that aren't, should be.

All the tests are in build/all.gyp. It's the "*_run" that aren't. "*_run" will
get away with GN in some semi-distant future but in the meantime, due to
http://crbug.com/503241, this patch simplifies my life by not breaking the Clang
bots.


> It's ok if I'm
> on the hook for making sure this is true until http://crbug.com/503241 is done
> (this also gives me an incentive to fix that bug). Does that sound ok to you?

Yes.

> On 2015/06/22 20:23:59, M-A Ruel wrote:
> > Interestingly, CrWinAsan doesn't specify add_tests_as_compile_targets=False
so
> I
> > didn't add it.
> 
> It seems to build the default target ("") anyways:
>
http://build.chromium.org/p/chromium.fyi/builders/CrWinClang/builds/1563/step...
> 
> > The goal is to not break these builders while isolating the rest of the
tests
> > via automated tools. Once it's possible to build all targets properly, we'll
> add
> > these back.

[M-A Ruel, 2015-06-26 15:17:19]

Ping

[M-A Ruel, 2015-06-29 19:22:04]

On 2015/06/26 15:17:19, M-A Ruel wrote:
> Ping

This is problematic as people will start isolating tests that will break on
clang builders.

[Nico, 2015-06-29 19:25:02]

On 2015/06/29 19:22:04, M-A Ruel wrote:
> On 2015/06/26 15:17:19, M-A Ruel wrote:
> > Ping
> 
> This is problematic as people will start isolating tests that will break on
> clang builders.

What's the ping for exactly? I thought the outcome here (comment 2/3) was that
we won't do this but instead add missing _run entries to build/all.gypi (?)

[Marc-Antoine Ruel (Google), 2015-06-29 22:41:50]

On 2015/06/29 19:25:02, Nico wrote:
> On 2015/06/29 19:22:04, M-A Ruel wrote:
> > On 2015/06/26 15:17:19, M-A Ruel wrote:
> > > Ping
> > 
> > This is problematic as people will start isolating tests that will break on
> > clang builders.
> 
> What's the ping for exactly? I thought the outcome here (comment 2/3) was that
> we won't do this but instead add missing _run entries to build/all.gypi (?)

We misunderstood ourselves. I meant this CL should be in and then reverted once
https://code.google.com/p/chromium/issues/detail?id=503241 is fixed. I see this
as a short term gap. I don't want to forcibly pull all the _run target in, as
this would archive all targets unnecessarily, which would increase the duration
of the archive step.

[Nico, 2015-07-01 15:45:00]

On 2015/06/29 22:41:50, Marc-Antoine Ruel (Google) wrote:
> On 2015/06/29 19:25:02, Nico wrote:
> > On 2015/06/29 19:22:04, M-A Ruel wrote:
> > > On 2015/06/26 15:17:19, M-A Ruel wrote:
> > > > Ping
> > > 
> > > This is problematic as people will start isolating tests that will break
on
> > > clang builders.
> > 
> > What's the ping for exactly? I thought the outcome here (comment 2/3) was
that
> > we won't do this but instead add missing _run entries to build/all.gypi (?)
> 
> We misunderstood ourselves. I meant this CL should be in and then reverted
once
> https://code.google.com/p/chromium/issues/detail?id=503241 is fixed. I see
this
> as a short term gap. I don't want to forcibly pull all the _run target in, as
> this would archive all targets unnecessarily, which would increase the
duration
> of the archive step.

It would only affect builders building 'all', which no bots other than the clang
bots do, no (?)

[M-A Ruel, 2015-07-01 15:58:06]

On phone atm so can't confirm 100% but i think no other bot build all *and*
use swarming/run tests.

Wrote by human
Le 2015-07-01 11:45 AM, <thakis@chromium.org> a écrit :

> On 2015/06/29 22:41:50, Marc-Antoine Ruel (Google) wrote:
>
>> On 2015/06/29 19:25:02, Nico wrote:
>> > On 2015/06/29 19:22:04, M-A Ruel wrote:
>> > > On 2015/06/26 15:17:19, M-A Ruel wrote:
>> > > > Ping
>> > >
>> > > This is problematic as people will start isolating tests that will
>> break
>>
> on
>
>> > > clang builders.
>> >
>> > What's the ping for exactly? I thought the outcome here (comment 2/3)
>> was
>>
> that
>
>> > we won't do this but instead add missing _run entries to build/all.gypi
>> (?)
>>
>
>  We misunderstood ourselves. I meant this CL should be in and then reverted
>>
> once
>
>> https://code.google.com/p/chromium/issues/detail?id=503241 is fixed. I
>> see
>>
> this
>
>> as a short term gap. I don't want to forcibly pull all the _run target
>> in, as
>> this would archive all targets unnecessarily, which would increase the
>>
> duration
>
>> of the archive step.
>>
>
> It would only affect builders building 'all', which no bots other than the
> clang
> bots do, no (?)
>
> https://codereview.chromium.org/1202573002/
>

To unsubscribe from this group and stop receiving emails from it, send an email
to chromium-reviews+unsubscribe@chromium.org.

[Nico, 2015-07-01 18:02:45]

In that case, making All depend on all the _run targets should be fine,
right?

On Wed, Jul 1, 2015 at 8:58 AM, Marc-Antoine Ruel <maruel@chromium.org>
wrote:

> On phone atm so can't confirm 100% but i think no other bot build all
> *and* use swarming/run tests.
>
> Wrote by human
> Le 2015-07-01 11:45 AM, <thakis@chromium.org> a écrit :
>
>> On 2015/06/29 22:41:50, Marc-Antoine Ruel (Google) wrote:
>>
>>> On 2015/06/29 19:25:02, Nico wrote:
>>> > On 2015/06/29 19:22:04, M-A Ruel wrote:
>>> > > On 2015/06/26 15:17:19, M-A Ruel wrote:
>>> > > > Ping
>>> > >
>>> > > This is problematic as people will start isolating tests that will
>>> break
>>>
>> on
>>
>>> > > clang builders.
>>> >
>>> > What's the ping for exactly? I thought the outcome here (comment 2/3)
>>> was
>>>
>> that
>>
>>> > we won't do this but instead add missing _run entries to
>>> build/all.gypi (?)
>>>
>>
>>  We misunderstood ourselves. I meant this CL should be in and then
>>> reverted
>>>
>> once
>>
>>> https://code.google.com/p/chromium/issues/detail?id=503241 is fixed. I
>>> see
>>>
>> this
>>
>>> as a short term gap. I don't want to forcibly pull all the _run target
>>> in, as
>>> this would archive all targets unnecessarily, which would increase the
>>>
>> duration
>>
>>> of the archive step.
>>>
>>
>> It would only affect builders building 'all', which no bots other than
>> the clang
>> bots do, no (?)
>>
>> https://codereview.chromium.org/1202573002/
>>
>

To unsubscribe from this group and stop receiving emails from it, send an email
to chromium-reviews+unsubscribe@chromium.org.