Subject: Re: Emacs and sys_errlist woes
To: John F. Woods <>
From: None <>
List: current-users
Date: 12/11/1993 09:28:40
> I take it that gcc is refusing to compile these programs because of
> const char * versus char * type clashes (I'm two weeks behind again, so
> I can't just peek at the source :-).  Is it possible to have gcc regard
> this particular error to be a warning rather than a fatal error, to smooth
> over this transition (after all (putting on the Language Lawyer hat),
> "undefined behavior" includes anything up to and including doing the right
> thing, all that's required is a ``diagnostic'')?  (Of course, *I'm* not
> going to volunteer to crawl around in gcc's guts, nosir!  :-)

(obviously a lisp programmer.  :-)

I think this would eliminate the transition rather than smooth it.  If
we do that, we might as well scrap the whole "const" thing entirely.

It is obviously the core team's policy to advance the technology even
when it causes some pain.  Although I occasionally resent this policy,
I understand the motivation behind it, and I respect it.

On the other hand, this "const" thing isn't truly being enforced since
there are no type checks done at link time.  For instance, right now
any file which doesn't include stdio.h can declare sys_errlist as a
non-const and get away with it.  I saw some files in emacs-19.22 that
were like this in fact, but I used grep to find all of them.

Email:           | Brown University
PGP Key: finger       | Dept of Computer Science