Hi Kristian,
Some comments inline...
Kristian Waagan wrote:
Hello,
I'd like to investigate whether the Apache Derby project can get some
ASF machine resources assigned, because I feel we have some tasks that
could be better performed in a project environment (as opposed to a
user environment).
I agree that moving our continuous integration environment onto
community machines would be a good idea. That includes the
commit-triggered build/test cycles, nightly build/test cycles, and doc
builds.
One of the good things about a project environment, is that more
people have the opportunity to fix problems and get involved. It is
not clear to me yet whether access to the resource should be limited
to PMC members, committers or a different group of contributers.
If this work is scoped to the Derby project, it makes sense to me to
limit access to Derby committers.
The initial issue I'd like to solve is building and publishing the
manuals. This process is currently pretty "closed" (i.e., just a few
persons can monitor and fix problems with the process).
+1
Other things on the list of possible tasks:
- Clover (needs license, ASF might have one we can use)
Code coverage and intelligent test selection on commits.
A full run of the test suite on a weekly basis?
- Hudson
Continuous builds. Note that Ole is doing a great job of this today
and I'm not suggesting stopping that effort!
I'm not sure if we can expect to run the full test suite, but maybe
we can run a small subset of the tests, and it looks exciting to
combine Clover and Hudson.
I think you're suggesting that Ole migrate the build/test framework from
Sun machines onto a community machine?
Let me spend a few sentences on why I'm suggesting Clover.
Clover is a code coverage tool. It uses source level instrumentation
to get the information it needs when running tests. Compilation is
handled through the regular project ant scripts (i.e., that will take
care of using the correct compiler etc).
As far as I'm aware, it can answer questions like [1]:
- what is our overall code coverage?
- which test(s) contribute to the coverage of this code?
- which tests should be rerun first when we change this part of the
code? (i.e. which tests are most likely to fail after a code change)
- which areas of the code should be better tested? (complexity/size
and missing coverage)
Clover sounds like a good tool. But I'd recommend separating the Clover
discussion into a separate email thread.
I think we may be able to get a Solaris zone to perform one or more of
these tasks.
The first step would involve at least the following steps:
1) Obtain a Solaris zone.
2) Agree on who should have access to the zone (PMC chair will be the
initial owner).
I don't understand why the PMC chair needs to be so heavily involved. I
can promise you that this PMC chair will pass the responsibility onto
more knowledgeable people in a New York minute!
Thanks,
-Rick
3) Initial configuration of the zone (software, users, etc).
4) Describe and implement the automation task for building and
publishing the manuals.
5) Figure out the usage restrictions on the zone (if any).
I volunteer to investigate this and hopefully get a zone created.
Before I take this to the PMC and/or infra, I'd like some feedback
from the community.
Is this something that will add value to the project?
Regards,
