Can we really setup "continuous" TCK testing? How long does it take 2
VMs to complete a TCK run?
Shouldn't running the TCK prereq a Continuous Build first, by moving the
automated builds that Jarek runs over to these machines and only run a
build when there are svn changes in that branch (instead of a time based
approach)? That would only generate builds on active branches (like 2.1
and trunk) while increasing the overall usage of these 2 machines to
show that we're really using them.
Also, which releases of Geronimo would we setup for automated TCK runs?
-Donald
Kevan Miller wrote:
As many of you know, we have two apache.org machines for TCK testing of
Geronimo (phoebe and selene). We used the machines for certification of
our 2.1.3 release. However, this testing was run manually. It's time to
get continuous, automatic TCK testing running on these machines.
We had the basic setup running on GBuild machines. IIRC, this was built
around AntHill -- Jason Dillon knows it best. There are other testing
systems available (e.g. Hudson) which could be used for this task.
What do others think? What underlying technology should we use? Who
wants to get involved?
I think this discussion should be on our dev@ mailing list. TCK should
be for test specific discussions.
--kevan
