Using C++ in gcc (was Re: [RFC] type safe trees)

[Matt Austern]

On Jun 25, 2004, at 8:31 AM, Richard Kenner wrote:

>     This would only be an issue if you are using C++ features like 
> exception
>     handling and class definitions that require vtables and
>     constructors/destructors to be handled by the linker.
>
> Right, that's what I said.
>
>     The proposal was to use only a tiny set of C++ features.
>
> The *current* proposal.  But I think everybody agrees that once we take
> that first step, it's going to get easier and easier to add more.  
> There's
> the real question, given human nature, of whether it's best not to take
> that first step.
>
> That's basically what I see as the heart of the debate.

I'm still not sure whether I think that it's a good idea to use C++ in
gcc.  I'm pretty sure of these two things, though:
  (1) I don't believe the argument above is a good reason not
        to use C++.  The argument above has two parts: that once
        we start using any parts of C++ there will be no  reason not
        to use all of the language, and that this will have bad
        consequences because some parts of the C++ language
        are unsuitable for our purposes.  Either part of that
        argument might be true, but both can't be.
  (2) Writing gcc in C++ is a better idea than writing it in an
        invented language of our own that gets turned into C by
        specially written tools.  (And that's true whether or not we
        call what we're doing an invented language, and whether
        or not we give that new language a name.)   All the
        arguments against C++ are even more true for this.

If we think that pure C meets our needs, then that's fine.  If we
don't, then we should be looking for an alternative language
that might serve us better.  C++ is one reasonable candidate.
There are others.

			--Matt