Re: master d582356: * src/fns.c (Frandom): Handle bignum `limit`s

[Eli Zaretskii <eliz@gnu.org>]

> From: Pip Cet <pipcet@gmail.com>
> Date: Sat, 6 Mar 2021 09:44:18 +0000
> Cc: Stefan Monnier <monnier@iro.umontreal.ca>, emacs-devel@gnu.org
> 
> I thought this code in code_conversion_save was safe:
> 
>       Lisp_Object name
>         = Fgenerate_new_buffer_name (Vcode_conversion_workbuf_name, Qnil);
>       workbuf = Fget_buffer_create (name, Qt);
> 
> but I had misread the second argument to Fget_buffer_create: it's
> inhibit-hooks, not run-hooks.
> 
> So I'm not sure whether code_conversion_save is allowed to call Lisp.

I'd rather it didn't, for more than one reason.  But we can side-step
this by making Fgenerate_new_buffer_name use random-fixnum, which is
still a pure-C implementation.

> It would really help to document the "doesn't call Lisp" and "doesn't
> quit" restrictions somewhere (but I'm not volunteering...)

I agree, on both counts.  We are making Emacs ever more complex, and
keeping correctness as we go, let alone avoiding performance
degradations, becomes more and more problematic.  So much so that you
can easily observe what control engineers call "limit-cycle problem"
in Emacs development, whereby the stability of Emacs doesn't become
steadily better with time, but instead shows small oscillations, even
though we fix bugs at a very high rate, including in old code.

> As an alternative, we could simply use get_random() % 1000000 and
> accept that the first 737418-ish buffer names are microscopically more
> likely to be used on 32-bit narrow-int systems.

Yes, we could do that.  Perhaps it's even better, and we could re-use
the same logic as in Frandom regarding the proximity to INTMASK.