Re: Buffer size limitation in insdel.c

[Eli Zaretskii <eliz@gnu.org>]

> From: Lars Magne Ingebrigtsen <larsi@gnus.org>
> Date: Wed, 22 Sep 2010 14:12:02 +0200
> 
> >   { EMACS_INT total_size = Z_BYTE - BEG_BYTE + GAP_SIZE + nbytes_added;
> >     if (total_size < 0
> > 	/* Don't allow a buffer size that won't fit in a Lisp integer.  */
> > 	|| total_size != XINT (make_number (total_size))
> > 	/* Don't allow a buffer size that won't fit in an int
> > 	   even if it will fit in a Lisp integer.
> > 	   That won't work because so many places still use `int'.  */
> > 	|| total_size != (EMACS_INT) (int) total_size)
> >       error ("Buffer exceeds maximum size");
> >
> > "bzr annotate" says this was committed a year ago by Stefan.
> >
> > Any objections to removing this limitation on 64-bit machines?
> 
> If you saw my experiment with -Wconversion, we have 5400 potential type
> conversion errors, where a lot (a *lot*) of them involved EMACS_INT/int
> confusion, and a lot of those were buffer-related.
> 
> So until that's fixed, I think it's a valid limitation.  Emacs is
> probably just going to segfault if you load a buffer that's bigger than
> (er) 2GB.

Then why do other places that bitch about buffer size being exceeded
don't do the same?