Bug 193734 – webkitpy: Wait longer when launching WebKitTestRunner

• Status: RESOLVED
• Resolution: FIXED
• Component: Tools / Tests

[Jonathan Bedard]

Originally, I only saw this problem when booting simulators. Recently, I saw this timeout hit well after the simulators were booted. The wait time was a bit arbitrary in the first place, I'd like to increase it.

[Jonathan Bedard]

Created attachment 359924 [details]
Patch

[WebKit Commit Bot]

Comment on attachment 359924 [details]
Patch

Clearing flags on attachment: 359924

Committed r240356: <https://trac.webkit.org/changeset/240356>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Radar WebKit Bug Importer]

<rdar://problem/47491778>

[Daniel Bates]

Comment on attachment 359924 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=359924&action=review

> Tools/Scripts/webkitpy/port/simulator_process.py:98
> +        with Timeout(15, RuntimeError('Timed out waiting for pid {} to connect at port {}'.format(self._pid, self._target_host.listening_port()))):

I did a double take when I saw this change. This timeout value seems to be following Moore's law. Check it out:

3 seconds @ <http://trac.webkit.org/r211370> (01/30/2017)
6 seconds @ <http://trac.webkit.org/r214896 > (04/04/2017)
*NOW* 15 seconds  (01/23/2019)

We are wasting engineering hours progressively increasing this value. I think we should just save us 10 years and pick ~3072 seconds :P. All joking aside, too bad we cannot do anything better.

[Jonathan Bedard]

(In reply to Daniel Bates from comment #5)
> Comment on attachment 359924 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=359924&action=review
> 
> > Tools/Scripts/webkitpy/port/simulator_process.py:98
> > +        with Timeout(15, RuntimeError('Timed out waiting for pid {} to connect at port {}'.format(self._pid, self._target_host.listening_port()))):
> 
> I did a double take when I saw this change. This timeout value seems to be
> following Moore's law. Check it out:
> 
> 3 seconds @ <http://trac.webkit.org/r211370> (01/30/2017)
> 6 seconds @ <http://trac.webkit.org/r214896 > (04/04/2017)
> *NOW* 15 seconds  (01/23/2019)
> 
> We are wasting engineering hours progressively increasing this value. I
> think we should just save us 10 years and pick ~3072 seconds :P. All joking
> aside, too bad we cannot do anything better.

I wish I had a better answer. This is closely related to problems booting simulators, basically, simulators are quite slow. We were getting into cases where bots where 30 minutes into a test run and hit this timeout because the machine was stuck gathering crashlogs from another test run while trying to start up the test runner.