Vladimir Ivanov has already written tests at the classfile level, which javac doesn't generate. They do not all pass yet, which is why they are not on by default.
Let me know if anyone wants to study the details. Note that there is already a bug filed to remove the bridging and covariant return subsets of tests since the vm no longer needs to support that. And we know we still need to visit in detail the expected appropriate exceptions that get thrown. thanks, Karen On Jul 25, 2013, at 10:25 AM, Joel Borggrén-Franck wrote: > Hi Amy, > > On Jul 24, 2013, at 3:30 AM, Amy Lu <amy...@oracle.com> wrote: > >> Thank you Dan ! >> >> Please see my comments inline... >> >> On 7/24/13 5:12 AM, Dan Smith wrote: >>> Hi. >>> >>> Per a request from Joel, I've taken a look at DefaultStaticTestData. I >>> don't really have the full context here, but I'm assuming that the >>> annotations get translated into tests that guarantee 1) the result of >>> Class.getMethods is exactly (no more -- excepting Object methods -- and no >>> less) those methods named by MethodDesc annotations; and 2) the result of >>> Class.getDeclaredMethods is exactly (no more, no less) those methods that >>> are marked "declared=YES". >>> >>> The expected results seem accurate. I would personally focus testing more >>> on different inheritance shapes and less on different combinations of >>> (unrelated) method declarations or presence/absence type variables (!?), >>> but it's a valid test in any case. >>> >>> There ought to be some testing for scenarios that javac won't generate, >>> like conflicting default method declarations. >>> >> Testing on "javac" is out of this scope, it's covered by langtools tests, >> say test/tools/javac/defaultMethods/ > > I sort of agree with Dan here. This wouldn't be testing of javac, rather > testing that Core Reflection works for combinations that javac doesn't > currently emit. However I think that is an excellent candidate for a follow > up test, we can address that after these test are finished. I will file a bug > for this. > > cheers > /Joel
