------- Additional Comments From greenrd at greenrd dot org 2005-05-27 01:00 ------- Oops - my diagnosis was slightly incorrect.
The stack trace is in fact correct, and it was klass.name that was null, not klass itself. But my conclusion is still correct - the first loop runs merrily over the end of the stack trace, because it skips too many entries before it starts. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21775
