I like the idea of raising awareness of available patches but I
think we want to hold someone up that is doing a bunch of work with
having to verify all patches that have been submitted and how the
patches might effect the new work.
One way (I think) to address the concern is to have more than one of
us apply the patch, run the tests and verify that all tests pass.
Assuming all tests pass then the patch should be OK and we put a
comment into the JIRA to that effect. Only problem I can see is that
some of the internal BEA tests would fail but these tests are on
their way into the OpenJPA code base correct?
BTW - BAU == ?
On Aug 30, 2006, at 8:07 AM, Kevin Sutter wrote:
> I'm looking to start a conversation on how we can coordinate
> verification of patch submissions. As you have probably noticed,
> and I have been working on a patch for OPENJPA-15 (
> http://issues.apache.org/jira/browse/OPENJPA-15). Granted, we
> had a
> couple of miscues as we continue to learn the patching and
> processes. But, we have also had to regenerate the patch a few
> because of changes being integrated into the affected files before
> the patch
> could be verified.
> I'm wondering if this is BAU in the open-source community, or
> something we can do to help this process?
> I know there was a discussion about having the ability to flag a
> JIRA report
> when a patch is attached. I don't believe I've seen resolution of
> request. Flagging a JIRA report might help some, but it still
> doesn't help
> with a timely verification of the patch. At least it would
> awareness that patches are available for verification. And, any
> new change
> activities should check for possible patches before hitting the
> Since this OpenJPA community is still highly reliant on the
> original authors
> of the contribution, I guess this question is mostly directed at
> them. Once
> more of us non-BEA committers become familiar enough with the code
> base, we
> can help with this verification process to help spread the wealth.
> But, in
> the mean time, it would seem that we need to figure out some
> process for
> verifying patches before the next set of changes gets committed
> the patch null-and-void.