This is the mail archive of the
java-patches@sources.redhat.com
mailing list for the Java project.
Re: PATCH: libgcj -vs- libstdc++-v3
- To: tromey at cygnus dot com
- Subject: Re: PATCH: libgcj -vs- libstdc++-v3
- From: Bryce McKinlay <bryce at albatross dot co dot nz>
- Date: Fri, 13 Oct 2000 17:53:34 +1300
- CC: java-patches at sourceware dot cygnus dot com, Olivier Louchart-Fletcher <olivier dot louchart at netregistry dot au dot com>
- References: <39E66555.47286ED@albatross.co.nz> <87n1g9cqqm.fsf@creche.cygnus.com>
Patch is checked in.
Tom Tromey wrote:
> Bryce> libgcj won't work with compilers configured with the "old"
> Bryce> libstdc++ any more, unless we link libgcj against libstdc++,
> Bryce> because the "old" libstdc++ doesn't have a separate libsupc++
> Bryce> at all.
>
> Will v3 be the default in gcc 3.0? That is what matters, I think.
Yes - this is part of the release criteria IIRC.
> Bryce> A better fix to the explicit -lsupc++ would be to change the
> Bryce> gcj front end to always implicitly -lsupc++ when linking? Is
> Bryce> there any potential disadvantage to doing that?
>
> Will package builders put libsupc++ into the -dev packages or into the
> ordinary runtime packages? That might be a disadvantage.
>
> I think we should follow what C++ does and link it into our library.
> That seems safest.
ok. We probibly should really link in all of it then, not just the
symbols that libgcj wants. That should be easy when libgcj is in the gcc
tree. That way, linking C++ object files usng gcj will allways work
without having to explicitly -lsupc++ or -lstdc++ (unless you really do
need stdc++, of course)
regards
[ bryce ]