V8 Buildbot: Pass isolate hashes from builder to tester.

V8 Buildbot: Pass isolate hashes from builder to tester.

BUG=chromium:535160
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=297076

[Michael Achenbach, 2015-10-09 07:40:32]

machenbach@chromium.org changed reviewers:
+ kjellander@chromium.org, tandrii@chromium.org

[Michael Achenbach, 2015-10-09 07:40:32]

PTAL - we'll only isolate on builders. Testers will take the hashes trigger and
collect without further logic.

[kjellander_chromium, 2015-10-09 08:05:02]

kjellander@chromium.org changed reviewers:
+ phajdan.jr@chromium.org

[kjellander_chromium, 2015-10-09 08:05:03]

Since I haven't done this myself yet, I'm going to loop in Pawel for advice
instead.

https://codereview.chromium.org/1403463002/diff/1/scripts/slave/recipe_module...
File scripts/slave/recipe_modules/v8/builders.py (right):

https://codereview.chromium.org/1403463002/diff/1/scripts/slave/recipe_module...
scripts/slave/recipe_modules/v8/builders.py:46: 'isolated_tests': {'default':
'[dummy hash for default]'},
This looks like test data to me, should it really be here?
Or is it because you use old-style triggering?

I don't see any of the builders in
scripts/slave/recipe_modules/chromium_tests/chromium_linux.py do this...
It seems the testers call find_isolated_tests instead, after extracting the
build.

[Michael Achenbach, 2015-10-09 08:06:53]

I'm not planning to do what chromium does. The plan is to not extract builds at
all on testers (this is not necessary as the hashes is all we need).

When the testers are light-weight enough, we can run multiple on the same slave.

[Michael Achenbach, 2015-10-09 08:13:56]

(I'm also planning to not even checkout on testers)

[Paweł Hajdan Jr., 2015-10-09 08:22:10]

LGTM

https://codereview.chromium.org/1403463002/diff/1/scripts/slave/recipe_module...
File scripts/slave/recipe_modules/v8/builders.py (right):

https://codereview.chromium.org/1403463002/diff/1/scripts/slave/recipe_module...
scripts/slave/recipe_modules/v8/builders.py:46: 'isolated_tests': {'default':
'[dummy hash for default]'},
On 2015/10/09 at 08:05:03, kjellander (chromium) wrote:
> This looks like test data to me, should it really be here?

FWIW I'm fine with this. Alternatively, maybe I didn't understand your concern,
Henrik.

> Or is it because you use old-style triggering?
> 
> I don't see any of the builders in
scripts/slave/recipe_modules/chromium_tests/chromium_linux.py do this...
> It seems the testers call find_isolated_tests instead, after extracting the
build.

find_isolated_tests has its drawbacks as well (e.g. too much magic when it scans
the tree as opposed to being explicit about which isolated targets we want - I
actually prefer the latter, which seems to be done in this CL).

[Michael Achenbach, 2015-10-09 08:25:00]

https://codereview.chromium.org/1403463002/diff/1/scripts/slave/recipe_module...
File scripts/slave/recipe_modules/v8/builders.py (right):

https://codereview.chromium.org/1403463002/diff/1/scripts/slave/recipe_module...
scripts/slave/recipe_modules/v8/builders.py:46: 'isolated_tests': {'default':
'[dummy hash for default]'},
I'd like to do this smarter. Right now, there is logic in the builder that scans
the tests to be run on the testers in order to calculate what to isolate. I
should reuse the same logic to simulate what was isolated in the tests. Will do
that in some follow up, otherwise I'd have to guess correctly for each builder
what it should simulate.

[Michael Achenbach, 2015-10-09 08:25:28]

The CQ bit was checked by machenbach@chromium.org

[commit-bot: I haz the power, 2015-10-09 08:25:41]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1403463002/1
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1403463002/1

[commit-bot: I haz the power, 2015-10-09 08:28:37]

Committed patchset #1 (id:1) as
http://src.chromium.org/viewvc/chrome?view=rev&revision=297076

[kjellander_chromium, 2015-10-09 08:55:49]

On 2015/10/09 08:22:10, Paweł Hajdan Jr. wrote:
> LGTM
> 
>
https://codereview.chromium.org/1403463002/diff/1/scripts/slave/recipe_module...
> File scripts/slave/recipe_modules/v8/builders.py (right):
> 
>
https://codereview.chromium.org/1403463002/diff/1/scripts/slave/recipe_module...
> scripts/slave/recipe_modules/v8/builders.py:46: 'isolated_tests': {'default':
> '[dummy hash for default]'},
> On 2015/10/09 at 08:05:03, kjellander (chromium) wrote:
> > This looks like test data to me, should it really be here?
> 
> FWIW I'm fine with this. Alternatively, maybe I didn't understand your
concern,
> Henrik.

I just thought it looked strange with "dummy data" in the builder configuration,
that's all.

> > Or is it because you use old-style triggering?
> > 
> > I don't see any of the builders in
> scripts/slave/recipe_modules/chromium_tests/chromium_linux.py do this...
> > It seems the testers call find_isolated_tests instead, after extracting the
> build.
> 
> find_isolated_tests has its drawbacks as well (e.g. too much magic when it
scans
> the tree as opposed to being explicit about which isolated targets we want - I
> actually prefer the latter, which seems to be done in this CL).

Thanks for the input. I clearly need to catch up on all this as well in order to
get swarming to work for us.

[kjellander_chromium, 2015-10-09 08:55:55]

lgtm

[Michael Achenbach, 2015-10-09 09:06:50]

> I just thought it looked strange with "dummy data" in the builder
configuration,
> that's all.
> 

This is because I try to simulate for the tester, what the builder also outputs
in the simulation here:
https://codereview.chromium.org/1403463002/diff/1/scripts/slave/recipes/v8.ex...

I could call it also anything else, but idealy, I'd not need that code at all.
Will work on that.

[Michael Achenbach, 2015-10-09 09:07:55]

Works. See the isolated_tests property on our staging tester:
http://build.chromium.org/p/client.v8/builders/V8%20Linux%20-%20swarming%20st...