Archie Cobbs wrote: > In the Mavue test gnu/testlet/java/lang/Class/init.java, this > test seems to think it's OK to initialize a class before the > first active use. But this contradicts JVMS $5.5, which says > initialization may happen "only as a result of ...". > > There are some other checks that can be added as well. > See attached revised version of this file.
Looks good. > Also: I like the suggestion someone had of putting a moratorium > on Classpath changes until we can clean up all the Mauve checks. That was me. > I.e., we need a "mauve-classpath" file that contains all the tests > an otherwise "perfect" Classpath-based VM should pass/not pass, etc. Exactly! I think it can be a great motivational tool to have a set of known good tests that can be gradually expanded, instead of the current haphazard fixing. It would also help to be able to compare the different VMs (and motivate the ones that have bugs to fix them). > Finally, who do we send Mauve patches to? See http://sources.redhat.com/mauve/, or you can ask Tom Tromey for cvs access. Regards, Jeroen _______________________________________________ Classpath mailing list [email protected] http://lists.gnu.org/mailman/listinfo/classpath
