Current Status

Tom Tromey tromey@cygnus.com
Tue Feb 16 15:32:00 GMT 1999


>>>>> "Terry" == Terry Sikes <tsikes@netcom.com> writes:

Terry> Something should be done to raise the visibility of gcj in
Terry> general, it seems to me.

I personally think the visibility will be raised as a side effect of
the runtime release.  I think a fair number of people are waiting for
it to be actually usable.

I still can't say when the runtime release will be, sorry.

Terry> It appears that gcj has essentially been rolled into egcs, and
Terry> is now just a part of that project?

Yes, the gcj front end is part of egcs.

Terry> One question I have about gcj is how well modularized and
Terry> documented is the GC implementation?  Will it be easy for third
Terry> parties to implement different GC subsystems?

Right now gcj assumes that the runtime is using a conservative GC.  On
the runtime side, the GC implementation is pretty well modularized,
but it isn't really documented.  (That isn't such a big deal since
there are only 3 or 4 entry points.)

The GC stuff will probably go through some revisions later on.  For
instance, right now the Boehm GC glue code in the runtime knows how to
mark objects.  I'd rather have the runtime itself do this, and just
call into the glue code to do the actual marking.  (That is, the
current factoring is wrong.)

Terry> I'm running Red Hat 5.2, is gcj included with the egcs 1.0.3 release?

As far as I know gcj hasn't been in any public egcs release.

Tom


More information about the Java mailing list