Paul Eggert <eggert@cs.ucla.edu> writes:

> I've been looking into this and have several patches along these lines.
> None of them address message-unique-id directly yet (I plan to tackle
> this soon) but they do address the general problem area. The basic idea
> is to use a new make-nonce primitive.

Thanks!  I have read your patchset, which looks good to me.

I have also attached my latest patch for `message-unique-id', but I'm
not married to it if you have something better in mind.  It could easily
be updated to use `make-nonce' though.

>  (defun math-init-random-base ()
[...snip...]
> +  (declare (obsolete nil "29.1")))

This is a nit, but perhaps this could be simplified to:

    (declare-obsolete-function-alias 'math-init-random-base
       #'ignore "29.1)

> diff --git a/src/sysdep.c b/src/sysdep.c
> index 4786c8fa4f..5117460fc0 100644
> --- a/src/sysdep.c
> +++ b/src/sysdep.c
> @@ -2159,6 +2159,22 @@ seed_random (void *seed, ptrdiff_t seed_size)
>    set_random_seed (arg);
>  }
>
> +/* Set BUF, of size BUFSIZE, to random data derived from system entropy.  */
> +
> +void
> +get_entropy (void *buf, ptrdiff_t bufsize)
> +{
> +  char *p = buf, *lim = p + bufsize;
> +  while (p < lim)
> +    {
> +      ssize_t gotten = getrandom (p, lim - p, 0);
> +      if (0 <= gotten)
> +	p += gotten;
> +      else if (errno != EINTR)
> +	report_file_error ("Getting random data", Qnil);
> +    }
> +}

If we claim that the random data is suitable for cryptographic purposes,
should we be using the GRND_RANDOM flag here?

On Linux, flags 0 and GRND_RANDOM are equivalent (I read the most recent
kernel code to verify this).  But I have no idea about other platforms.