Re: svn commit: r979283 - in /qpid/trunk/qpid/java/common/src/main/java/org/apache/qpid/transport: Session.java SessionDelegate.java

2010-07-26 Thread Rajith Attapattu
Andrew, Have you tested this change with the C++ test profiles ? Anytime the 0-10 code path is changed, please make sure to test with the C++ test profiles. For example a recent checkin cause the JMSProperty test to fail (I believe in all test profiles). While it's understandable that humans

Re: svn commit: r979283 - in /qpid/trunk/qpid/java/common/src/main/java/org/apache/qpid/transport: Session.java SessionDelegate.java

2010-07-26 Thread Robert Godfrey
I agree we should be checking all code against both brokers... but I think it will soon become unreasonable to expect that before every checkin we test all combinations. At the moment we *should* be checking the following scenarios on a Java client checkin: Java InVM broker Java External broker

Re: svn commit: r979283 - in /qpid/trunk/qpid/java/common/src/main/java/org/apache/qpid/transport: Session.java SessionDelegate.java

2010-07-26 Thread Rajith Attapattu
Rob, I agree about the complexity involved in running the test profiles and I agree with you about the set of test profiles you mentioned in the email. While the above criteria could be relaxed for certain changes, for other non trivial changes that has a history of issues the above should be

Re: svn commit: r979283 - in /qpid/trunk/qpid/java/common/src/main/java/org/apache/qpid/transport: Session.java SessionDelegate.java

2010-07-26 Thread Andrew Kennedy
Yes, I ran the tests for the most recent set of commits against the cpp test profile under the ant build system, which I hope would have highlighted any issues with the C++ broker versus the 0-10 changes in the Java client? Sorry about the JMSTestProperties, I checked in a fix for one profile,

Re: svn commit: r979283 - in /qpid/trunk/qpid/java/common/src/main/java/org/apache/qpid/transport: Session.java SessionDelegate.java

2010-07-26 Thread Rajith Attapattu
My own view is that on checkin we should need only to be running *unit* tests (which are in rather short supply to be fair).  We should have a CI environment where *system* / *integration* tests are being run constantly with all possible profiles being tested.  IMHO unit tests for the client

Re: svn commit: r979283 - in /qpid/trunk/qpid/java/common/src/main/java/org/apache/qpid/transport: Session.java SessionDelegate.java

2010-07-26 Thread Rajith Attapattu
On Mon, Jul 26, 2010 at 11:06 AM, Andrew Kennedy andrewinternatio...@gmail.com wrote: Yes, I ran the tests for the most recent set of commits against the cpp test profile under the ant build system, which I hope would have highlighted any issues with the C++ broker versus the 0-10 changes in

Re: svn commit: r979283 - in /qpid/trunk/qpid/java/common/src/main/java/org/apache/qpid/transport: Session.java SessionDelegate.java

2010-07-26 Thread Justin Ross
On Mon, 26 Jul 2010, Rajith Attapattu wrote: My own view is that on checkin we should need only to be running *unit* tests (which are in rather short supply to be fair).  We should have a CI environment where *system* / *integration* tests are being run constantly with all possible profiles