On 01/18/2013 10:54 PM, Eli Zaretskii wrote:
> It's not a bug, it's a fundamental limitation of how Emacs accesses
> buffers.  Any valid buffer position _must_ fit into the width of an
> Emacs integer.

Sure, but Vitalie's not asking for large buffer positions,
just for large file offsets.  Emac's a bit squirrelly in this
area, as it represents large file offsets by using floating
point numbers, but if it's going to be squirrelly it should
be *consistently* squirrelly, so I installed the attached patch
into the trunk as bzr 111554.

If 64-bit EMACS_INT were the default Vitalie wouldn't have had this
problem.  We have a simple fix (change configure so that --with-wide-int
is the default, see Bug#8794) but this most likely would slow Emacs a bit.
Stefan was skeptical and asked for benchmarks, but maybe
I should boost the priority of this.  Or, more drastically,
maybe we should bite the bullet and add bignums (Bug#8611);
it can't be *that* hard.