I tested it again, and even in Eclipse I have some of the strange failures I have noticed (ie, some DN is not present in the partitions table).
I have suspected some problem with the DN parser (thread safety issues), but after having read the code, I have dismissed it. Now, after having added some logs in the method which check the presence of a DN in the partition table, I'm suspecting some collision in the server-integ tests, if for instance you run some tests which remove a partition before the one which suppose to find it. This is not something which occurs everytime. For instance, if I run the failing tests alone, they are all ok. It's when I run the full build that I get always the same error. Anyway, I also found another error, obtained when building the code with an IBM JVM (I will fill a JIRA for this one later). There is obviously something wrong, and it's probable depending on the computer you run the test one. My computer is pretty fast compared to Andrea's and Kiran's ones. Pierre-Arnaud run the test on a mac, so it's not equivalent (not the same jvm, not the same system). I also noticed in the past few months that I had to run the tests twice in order to get the server to be built. I simply ignored the problem, even if I _knew_ there were something wrong about it, but now, it's time to face the facts : the build is not stable... -- Regards, Cordialement, Emmanuel Lécharny www.iktek.com
