Run mozilla and webkit of fast quickcheck archs as well

Run mozilla and webkit of fast quickcheck archs as well

BUG=none
R=svenpanne@chromium.org
LOG=n

Committed: https://code.google.com/p/v8/source/detail?r=20056

[Sven Panne, 2014-03-19 09:35:35]

lgtm

https://codereview.chromium.org/204363002/diff/1/Makefile
File Makefile (right):

https://codereview.chromium.org/204363002/diff/1/Makefile#newcode375
Makefile:375: SUPERFASTTESTMODES = ia32.release,x64.release
What's next? RIDICULOUSLYFASTMODES? LUDICROUSLYFASTMODES? :-D

[Jakob Kummerow, 2014-03-19 09:52:09]

DBC: 
Putting the additional logic "if ia32.release then run mozilla+webkit in
addition to the default tests" into run_test.py (guarded by the --quickcheck
flag) would (1) result in a small speed improvement since long-running tests
belonging to different suites would be run in parallel, and (2) improve the UX a
bit because there wouldn't be two lines saying ">>> Running tests for
ia32.release".

Also, I'm not sure there's value in running those additional tests for two
architectures. Have there been examples recently where one of them failed on
x64.release but not on ia32.release?

[jochen (gone - plz use gerrit), 2014-03-19 10:38:55]

On 2014/03/19 09:52:09, Jakob wrote:
> DBC: 
> Putting the additional logic "if ia32.release then run mozilla+webkit in
> addition to the default tests" into run_test.py (guarded by the --quickcheck
> flag) would (1) result in a small speed improvement since long-running tests
> belonging to different suites would be run in parallel, and (2) improve the UX
a
> bit because there wouldn't be two lines saying ">>> Running tests for
> ia32.release".

I think spreading the logic what archs do what between the Makefile and
run-tests.py is worse.


> 
> Also, I'm not sure there's value in running those additional tests for two
> architectures. Have there been examples recently where one of them failed on
> x64.release but not on ia32.release?

no idea. Sven, you proposed this change, wdyt?

[Sven Panne, 2014-03-19 10:44:29]

On 2014/03/19 10:38:55, jochen wrote:
> I think spreading the logic what archs do what between the Makefile and
> run-tests.py is worse.

Agreed.

> > Also, I'm not sure there's value in running those additional tests for two
> > architectures. Have there been examples recently where one of them failed on
> > x64.release but not on ia32.release?
> 
> no idea. Sven, you proposed this change, wdyt?

Hmmm, I think Jakob is right and running it for one of the Intel platform will
very probably be enough. It doesn't have to be foolproof, just catch e.g. some
ES6-related changes etc., which are normally platform-independent.

[jochen (gone - plz use gerrit), 2014-03-19 10:48:52]

Committed patchset #2 manually as r20056 (presubmit successful).