Per Bothner wrote: > Looking at the code, it is clear that the first candidate is the > correct one, and the second is bogus. The problem is in the function > find_applicable_accessible_methods_list, which searches the > surrounding classes of an inner class. This is contrary to the JLS, > but is fixed by my patch to find_applicable_accessible_methods_list, > which I'm fairly confident about. Cool - thanks for looking at this, as there are some long-standing PRs about this issue: http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&pr=1348&database=gcc http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&pr=1932&database=gcc You may want to run the test cases in those URLs against your patch. regards [ bryce ]