Re: [RFC] bytecomp: simple source-level optimization

[Dmitry Antipov]

On 06/23/2014 01:56 PM, David Kastrup wrote:

> The latter is a bad idea.  You cannot reliably compare integers with
> eq in Common Lisp, cf
> <URL:http://www.cs.cmu.edu/Groups/AI/html/cltl/clm/node74.html>:

I know.

> The current Elisp implementation in Emacs (though possibly not in
> XEmacs, and definitely not in GUILE) _does_ implement all integers as
> immediate values, but that may well change in future:

Since such a change will break a lot of things under lisp/ directory,
I don't see practical considerations to make this move in foreseeable
future. Nevertheless, it may be worth trying to enhance bytecomp to
emit warning if at least one operand of 'eq' is a compile-time numeric
constant.

Dmitry