GCJ and generics

Andrew Haley aph@redhat.com
Mon Oct 20 17:39:00 GMT 2003


Bryce McKinlay writes:
 > On Oct 15, 2003, at 4:52 AM, Tom Tromey wrote:
 > 
 > >>>>>> "Cedric" == Cedric Berger <cedric@berger.to> writes:
 > >
 > > Cedric> I think generics define multiple functions with the same 
 > > arguments
 > > Cedric> but different return values. Could that be the problem?
 > >
 > > Yes, see PR 9861.
 > >
 > > We'll have to change the mangling to make this work.
 > > We should probably tack this on the new ABI work -- i.e., not declare
 > > the new ABI finished until this is done.
 > 
 > With the new ABI we won't be married to any particular name mangling 
 > scheme - that is, the name mangling can be changed without breaking 
 > binary compatibility, so it isn't a major issue.

I suspect that what Tom meant was that we would like to fix this
problem, but we need the new ABI to do it.  Anyway, you're quite
right: we'll need to fix our mangling so that we generate unique
names.

 > Names will only have to be a) unique within each compilation unit
 > and b) descriptive enough to aid debugging. Changing to Java's name
 > mangling would be a logical choice. That would confuse GDB but the
 > new ABI is going to confuse it anyway.

Not much, I wouldn't think.

Andrew.



More information about the Java mailing list