Re: Is Java broken on the trunk?

[Tom Tromey <tromey at redhat dot com>]

>>>>> "Hans" == Boehm, Hans <hans_boehm@hp.com> writes:

Hans> I also don't think I tested without threads but with hash
Hans> synchronization.  Can we just arrange to turn off hash
Hans> synchronization in the no-threads case?  No-threads
Hans> synchronization is pretty fast anyway :-) .

That's probably what I'll do.  I want to reproduce the problem before
making any changes though.

Hans> David Mosberger and I ran into a vaguely related issue
Hans> yesterday.  The 3.0 tree on Itanium seems to default to the
Hans> no-threads case.  This seems clearly wrong for Java.  Does it
Hans> also imply that the default C++ runtime or generated code is not
Hans> thread-safe?  That would also be clearly wrong.

Yes.  People who really want to use Java have to explicitly enable
threads.  The entire compiler must be configured with threads, as the
exception handling support requires knowledge of the thread system.
At least, that's how it was in the past.  I don't know whether this
changed with the new EH code.

Ideally, perhaps, this code would also use thin locks.

While I agree that this default is a poor choice for Java, I think the
expectation is that system builders will be savvy enough to turn on
the flag.

Tom