An Ansi-C question or... Is egcs/gcc Ansi compliant?

An Ansi-C question or... Is egcs/gcc Ansi compliant?

[Enrique I.R.]

Hi.

Just playing with const:

const int i=0;
   i++;
   printf("%i\n",i);

You get a warning only (increment of read-only variable) and then both compiles
(egcs&gcc) and 'i' is modified at runtime. I was perplexed and checked this with
other compilers (OSF's, vc++, bcb) and couldn't even compile this.

Why ignore egcs/gcc the const keyword such way?
Is there anyway to get a real const with egcs?

PS: I don't know this message is really on-topic here. If you know of a mailing
list this could fit better, please tell me.

(for USENET postings) Delete 'CUTTHIS' from my e-mail to reply.
        -----------
        Enrique I.R.
 Freelance Sofware Developer.
   Gran Canaria. Spain.
  --aka CtrInt on #Linux--

Re: An Ansi-C question or... Is egcs/gcc Ansi compliant?

[Joe Buck]

> Just playing with const:
> 
> const int i=3D0;
>    i++;
>    printf("%i\n",i);
> 
> You get a warning only (increment of read-only variable) and then both =
> compiles
> (egcs&gcc) and 'i' is modified at runtime. I was perplexed and checked =
> this with
> other compilers (OSF's, vc++, bcb) and couldn't even compile this.

egcs is technically correct, in that all that the ANSI/ISO standard
requires is a diagnostic here, and a warning qualifies as a diagnostic.

However, I agree with you that const violations of this kind should
be errors, not warnings ... especially if egcs then goes on to put i
in a read-only section and the program crashes at runtime!

> Why ignore egcs/gcc the const keyword such way?

The keyword is not ignored, you get a warning.

> Is there anyway to get a real const with egcs?

It's a real const.

Re: An Ansi-C question or... Is egcs/gcc Ansi compliant?

[Martin v. Loewis]

> Just playing with const:
[...]
> Why ignore egcs/gcc the const keyword such way?

The keyword is not ignored: you get a warning. This is ANSI compliant;
the compiler is required to diagnose the error, it may still chose to
compile the program (as an extension).

> Is there anyway to get a real const with egcs?

If you absolutely want compilation to fail, you should pass the flag
-pedantic-errors

> PS: I don't know this message is really on-topic here.

This was a good choice. Since you've suspected a bug,
egcs-bugs@cygnus.com would have been even better.

Regards,
Martin