This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Status of new C++ abi
On Tue, Sep 26, 2000 at 08:36:03PM -0700, Mark Mitchell wrote:
> >>>>> "Zack" == Zack Weinberg <zackw@stanford.edu> writes:
>
> Zack> What is the status of the new ABI? Referring here to the
>
> Hey, aren't you a chemist now? :-)
Yes, but I'm still reading the mailing list and I hope to kick in the
occasional patch as I have time. [*Very* occasional patch, looks like.]
> Zack> compiler side only, not libstdc++3. I was under the
> Zack> impression that -fnew-abi was automatically on if you
> Zack> enabled libstdc++3, but I seem to have been mistaken.
>
> You are. The new ABI is in the state of implemented-but-untested --
> things should work, but there hasn't been enough real world
> experimentation to know. The plan is to get switched over to V3
> first; then switch to the new ABI. This will minimize the number of
> variables changing at once. I'd been hoping to switch to V3 ages ago,
> but the library folks have hit some snags. Benjamin is working on
> this even as we speak, to the best of my knowledge.
Okay, so the situation is a lot more complicated than I thought - no
surprise.
In case you're wondering, I attempted to move all the C++-specific
code out of libgcc into libstdc++ (I can argue for this change in
detail if you want) and ran into all kinds of confusion over which ABI
is in use.
Incidentally, the internal definition of struct cp_eh_info doesn't
match the definition in exceptions.cc.
zw