Re: g++

[Bruno H. <br...@cl...>]

Sam wrote:
> > I think C99 allows non-constant elements in aggregate initializers.
> > C++ also allows non-constant elements in static aggregate initializers.
>
> so you are saying that it's not that they suddenly forbade this, but
> that it has always been technically forbidden but was available as an
> extension from some C vendors and has now been standardized in c++ and
> c99.

Right. And before ANSI C you could not even assign structs like this:

    struct foo instance1, instance2;
    instance1 = instance2;

You had to use bcopy() on the addresses. (Although you could 'return'
structs, and then the compiler would generate the bcopy() of the return
value for you.)

> > Yes I think the ability to work around a severe compiler problem with
> > just a little patch in a preprocessor is a good reason to keep
> > 'varbrace' alive.
>
> Just a second!
> if both C++ and C99 permit this _and_ also variables in the middle of
> the block, then why don't we just drop support for pre-C99?

So few compilers do support C99. In particular, to my knowledge, MSVC 7
isn't C99 compliant. (It has none of <stdbool.h>, <stdint.h>, <inttypes.h>).
So dropping C99 means dropping the WIN32_NATIVE port of clisp. Which I
won't agree to.

Bruno