I have a test that is suddenly failing in 2.1.0 but only when I do a site build.
The test is a little wierd in that it is testing some code that makes the 2nd thread to enter a method pause with a thread wait until the 1st thread to enter is finished. To do this the test makes up a couple of runnables, starts one, waits a bit and starts the other and makes sure the 2nd thread in takes longer to run than it otherwise would. This wasn't something I got right the first time. (The way Java decides what thread to run and when isn't immediately obvious to my mind at least.) I know that this isn't the final word on testing thread stuff. But the test works on multi and single CPU workstations under Windows XP and on a VmWare vm under Linux (RHE 4.2) so I haven't worried about it. Its been okay for a year or so under Maven 2.0.7 and 2.0.8. Further, it runs on all those machines when not doing the site build. (mvn clean install site) My first thought has to do with reports using instrumentation and such which causes the tests run multiple times. So I have turned off the Dashboard Report and the Cobertura Report first. Before I go through all the pain of taking out reports until it starts working I was wondering if this rung any bells with any of you. Clues are good. Answers are better. Whatever someone finds. (Oh, it runs for 32 minutes before failing so taking out reports trial and error takes forever. There is a parent POM and a whole bunch of its children building a RAR and an EAR.) Thanks. -- Lee Meador
