Peter Donald wrote: > On Thu, 21 Feb 2002 01:59, Leif Mortenson wrote: > >>I stared at these and don't see the problem. Any ideas? Here is the >>current build output that I am seeing: >> > > Unfortunately it is jikes that is buggered up. You were defining a m_manager > class in > > >org.apache.avalon.excalibur.component.test.ExcaliburComponentManagerTestCase.AbstractBadCircularKid > > which made the m_manager defined in > > org.apache.avalon.excalibur.component.test.ExcaliburComponentManagerTestCase
AbstractBadCircularKid defined its own m_manager field which should have been in scope in the method which was failing. In most places, the compiler will look at fields in the currect class before searching for fields in the outer class. In this case though, the javac compiler was getting confused because there were two possible choices for m_manager. Jikes was correctly choosing this.m_manager over ExcaliburComponentManagerTestCase.this.m_manager. Given that though the code is hard to read and should be fixed :-) > > inaccessible. So javac was correct complain while jikes just generated bad > bytecode ;) But the test ran correctly when build from Jikes, so the bytecodes were correct. If they were bad, then the Validator would have provided an error when the class was loaded. Javac should never be compiling anything that does not run. Thanks, Leif -- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
