buildbot status

[Dan Kegel]

I've turned buildbot error email to this list back on, even though
I anticipate a burst of patch conflicts if/when Alexandre
commits a fix for the global rpcrt4 regression which doesn't
seem to affect his machine.  (I revert that locally so I can continue
testing wine.)

The patch series parser was adjusted slightly to not try to run patches
labelled [x/y try2] as singleton patches.  That eliminates one more source
of spurious failure messages.

There's now an ATI buildbot.

The transposed grid page, http://buildbot.kegel.com/tgrid , now shows
20 rows by default instead of 5.  (You can still see more by adding
e.g. ?length=60 .)

It also shows the testbot patch number and description of each patch,
so it's easier to see what's going on at a glance.

The buildbot is now smart enough to not rerun all tests when
only one directory's tests change.  Instead, it runs just that one
directory of tests five times.  (Except for cmd, which seems to be O(n^2)
in test length, and badly needs Frédéric's goto cache.)

It uses config.cache to shave a few seconds off the configure step.
It uses cpufreq-set -g performance on the AMD box to shave several
minutes off the cycle time
(and cpufreq-set -g conservative on the others, my office doesn't have
good AC :-)

Backlog can get up to an hour or two when long patch series are submitted.
A few more volunteer bots would help; Austin has one almost ready to hook up
( you can preview it at http://austin987.dyndns.org:8010 ).

There are several more blacklisted tests, partly in anticipation of
Austin's new bot, which ran into a problem or two.

A MacOSX 10.6 bot would be good.  Charles Davis is looking at
that, but other volunteers would be welcome, one is not really enough.

For more info, see http://wiki.winehq.org/BuildBot