On Fri, 22 Sep 2006, Steve Loughran <[EMAIL PROTECTED]> wrote: > The first subsidiary build works, but the second is failing
It gets further now, and fails in an expected manner since the test is supposed to fail under Gump. What I've done is that I added a <mkdir> for the <work> entry of the failing build. I thought we had left this problem behind with JDK 1.4+ but it seems to be raising its ugly head again. In JDK 1.3 you needed to add <mkdir>s for each <work> that was created during the build, if you need classes from those directories in <java> or non-forked <junit>s. The reason was that the VM would drop all CLASSPATH entries that didn't exist at startup time. What we have now seems to be a variant thereof. I don't think it is related to <antunit>, but then again we haven't seen the problem with any other task lately. Stefan --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
