GCC/GCJ versions?...

GCC/GCJ versions?...

[Nathanael Nerode]

I notice you're using the tree-ssa version of GCJ (with extra patches?).

Is this essential, or is it possible to use the mainline version which 
will become GCC 3.4?  If mainline doesn't work, are the changes 
backportable?

It would certainly be nice if 3.4 could Do The Right Thing.

-- 
Nathanael Nerode  <neroden at gcc.gnu.org>
http://home.twcny.rr.com/nerode/neroden/fdl.html

Re: GCC/GCJ versions?...

[Tom Tromey]

>>>>> "Nathanael" == Nathanael Nerode <neroden@twcny.rr.com> writes:

Nathanael> I notice you're using the tree-ssa version of GCJ (with
Nathanael> extra patches?).  Is this essential, or is it possible to
Nathanael> use the mainline version which will become GCC 3.4?  If
Nathanael> mainline doesn't work, are the changes backportable?

Nathanael> It would certainly be nice if 3.4 could Do The Right Thing.

The gcc trunk would be fine.  The patches should back-port pretty
easily.

I'm going to start putting some of the patches in pretty soon.
Probably after next week.

There are some problems though.  First, I don't have the rationale for
a couple of the patches.  Maybe they're in the archives of our
internal mailing list; I'll look there, but if the info isn't there
then... ugh.  Maybe they'll turn out not to be necessary :-)

The miranda patch turns out to have a bug, which we discovered quite
recently.  If you apply it to the gcc trunk, I think you won't be able
to rebuild libgcj.  So that patch needs reworking.

And, finally, the patch to disable the gcj verifier is unacceptable.
But due to bugs in that verifier we can't compile eclipse with gcj.
Fixing it looks like quite a bit of work (hence all the discussion
about putting C++ into the front ends a while back...).  I'm not sure
when this will happen.  This is the most likely candidate for missing
the 3.4 timeline.

Tom