[Webkit-unassigned] [Bug 37262] new-run-webkit-tests should estimate test completion time

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=37262





--- Comment #2 from Dirk Pranke <dpranke at chromium.org>  2010-04-08 12:36:14 PST ---
Note that in the perl run-webkit-tests, the HTTP tests were put at the end
intentionally because they were slowest. The original chrome run-webkit-tests
put HTTP tests at the beginning, since that shard was usually the slowest. But,
I would have thought that with Ojan's more recent sharding efforts, it would be
more uniform. 

If you run with --experimental-fully-parallel, then the distribution should
become more uniform, and if you run with --randomize-order, that would also
make it more (arguably completely) uniform. One or both of these should
probably be the default, but they risk making the testing more
nondeterministic. (Although I suspect that the timing issues when running in
parallel under load do enough to make things nondeterministic anyway that it's
moot).

Perhaps we should have the script output a seed number when running with
--randomize-order (visible with --log config), so that the shuffling could at
least be reproduced if necessary.

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