java unexpected failures with inexplicable error messages

[Tom Tromey]

>>>>> "Zack" == Zack Weinberg <zack@codesourcery.com> writes:

Zack> I'm seeing TestEarlyGC and TestLeak fail in my builds, with
Zack> runtime backtraces like this:

Zack> (1) Are these current expected failures, or did I break something?

Neither; these are fairly new bugs in libgcj.  I tracked them down to
a String change I made a while back.  I must have missed the
regression somehow.  TestLeak clearly shows up in my database as
failing on Apr 1; strangely, TestEarlyGC doesn't start failing until
the 20th.

Anyway, I have a patch for the problem that I'm testing now.  I'll
check it in tomorrow, assuming that the test suite really has no
regressions (and the two improvements) tonight.

Zack> (2) What's the deal with this "loaded class X was in fact named X"
Zack>     error message?  Shouldn't those two strings be, y'know, different?

_Jv_equalUtf8Consts compares the hash values first.  The bug was that
we ended up with an incorrect hash value for one of the utf8const
objects.

Tom