[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
Re: [eclipselink-dev] EclipseLink 2.0 unfrozen
|
Everyone,
Just a minor suggestion that we may be able to minimize the time
that a build is frozen in the future (to avoid build and/or test errors)
by establishing a minimum commit procedure - and only checking in if the
affected components' tests are 0,0
I propose we do something like the following for all transactions -
which will add 30-120 min to your checkin time depending on the whether
your machine is a stock P4 or a mainstream i7.
1) - verify that all core, jpa, sdo moxy (non-testing at
least) projects compile in eclipse (select all and refresh)
2) - run trunk>ant
3) - run at least the LRG for the product you are affecting
Example for JPA 2.0 work that does not affect core
trunk>ant test-jpa
For core API work like in DatabaseMapping.java I
would also run
trunk>ant test-core
4) "publish" results in your SVN commit comment (it will show
up in Fisheye http://fisheye2.atlassian.com/changelog/eclipselink )- so
developers that update can be aware that there are "expected" errors -
Ideally the numbers would be 0,0
Test Results
[JPA LRG] off of SVN rev# 5188
[junit] Tests run: 1311, Failures: 0, Errors: 0,
Time elapsed: 1,220.58 sec
5) be aware that to be "Absolutely" sure that your TXN is ok at
least for SE testing on your specific DB and JRE - you will need to
refresh a 2nd time after you commit - if there were checkins between
when you refreshed and committed. Why? because SVN will only force you
to do a 2nd refresh if there was a modification to something you have
checked out - it will not pick up on other dependencies. (so run tests
before refreshing (optional), run tests after a refresh and run tests a
2nd/3rd time after an optional 2nd update).
Here is an example
http://fisheye2.atlassian.com/changelog/eclipselink/?cs=5189
This may add time to each checkin but it will minimize the time
that other developers need to spend verifying that new failures or build
errors are not due to their own changes when they do an update to the
latest SVN rev#..
Test failures may still occur for cross project dependencies,
unexpected errors specific to the JRE (PPC vs Intel) or DB target
(oracle on dev, but mysql on the build server) and possibly the server
EE tests - which usually are only run nightly on a subset of the LRG.
thank you
/michael
Peter Krogh wrote:
Sorry for the inconvenience...
_______________________________________________
eclipselink-dev mailing list
eclipselink-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipselink-dev