On 2010-11-07, Sander Temme wrote: > So M3 picks up M2_HOME and infers the classworlds location from it?
If it is set, yes. If M2_HOME is not set, it will infer it from the location of the wrapper script - assuming the regulear layout of a maven distribution (i.e. M2_HOME is one directory up closer to the root than the directory holding thw wrapper script). > So will it run with classworlds 1.1 (which came with Maven 2) instead > of the plexus-classworlds-2.2.3.jar that comes with it? No. That's why Gump requires M3_HOME and sets M2_HOME to the value provided as M3_HOME when running mvn 3.x. We could have do it differently inside Gump - not providing M2_HOME at all to the forked process when running mvn and have the wrapper scripts figure out the rest. This might provide better results with system installed versions of mvn. Still we'd need a way to provide the absolute path of the wrapper scripts since both (mvn 2.x and 3.x) have the same name. > The midnight run will tell. Looks fine from here. You'll need git in order to get far with Ant based builds since you really want JUnit to be there. If you ever get the whole of testbase working, you'll likely see build module update errors with NUnit. So far I haven't invested time into figuring out the problem - bazaar merge talks about uncommitted changes which I don't understand since Gump doesn't touch the checked out trees. Stefan --------------------------------------------------------------------- To unsubscribe, e-mail: general-unsubscr...@gump.apache.org For additional commands, e-mail: general-h...@gump.apache.org