Re: Floating-point constant folding in Emacs byte compiler

[Stefan Monnier]

> Yes, because strings share the same EQ issue with floats.  And because
> we sometimes try to "optimize" that.  For example, in make_pure_string.

It's actually subtly different: with floats, EQ (and functions that
use it) is the *only* way to distinguish two equal floats.
With strings it's not the case because you can modify the object in
place (e.g. by changing its text-properties).

So for strings, the semantics is much less undefined: we're not free to
merge two strings just because they're `equal`, whereas for floats
we are.


        Stefan