[Webkit-unassigned] [Bug 63610] [NRWT] make nrwt able to test the same order like orwt

bugzilla-daemon at webkit.org bugzilla-daemon at webkit.org
Fri Jul 8 11:40:14 PDT 2011


--- Comment #16 from Dirk Pranke <dpranke at chromium.org>  2011-07-08 11:40:14 PST ---
(In reply to comment #15)
> (In reply to comment #12)
> > (From update of attachment 99971 [details] [details])
> > Okay, I think I get it now. This patch is trying to do two things ...
> > 
> > 1) Move the HTTP and websocket tests to the end
> > 
> > 2) Create shards of tests, 1000 at a time.
> > 
> > It is true that ORWT does (1). 
> > 
> > I think (2) is wholly new, correct? ORWT didn't have a "sharding" concept
> While it is true that ORWT didn't have the concept of sharding, it did restart DumpRenderTree after every 1000 test. This behaviour made ORWT quite deterministic that is why we initiated the work on this.

Ah, the --nthly flag. Yeah, we should probably add that to NRWT.

> The transition to NRWT however showed that using NRWT with --child-processes=1 the result is almost the same (except some minor flakyness) as with ORWT, so this feature is not that important as it was evaluated at first.
> Although the mutual exclusion of tests which need httpd is important for the Qt bots since our concept with ORWT was to run multiple bots in parallel on the same server without using a vm or a chroot.
> I do not fully understand the concept of shards regarding the http tests yet, but if BUG64092 gets fixed I think this bug really does not have any particular purpose.

Glad to hear it. That fix should land shortly :).

Shards are the units of parallelism - how work is divided up between multiple workers. Tests in a shard are run consecutively by the same worker, but different shards can be run by different workers (either serially or concurrently). Ideally every test would be its own shard, but some of our tests are broken and have inter-test dependencies, so we use the sharding concept to get around this.

Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

More information about the webkit-unassigned mailing list