Re: [Emacs-diffs] master 3b48f99: Tune pseudovector allocation assuming Qnil == 0

Re: [Emacs-diffs] master 3b48f99: Tune pseudovector allocation assuming Qnil == 0

[Stefan Monnier]

> +  /* Only the first lisplen slots will be traced normally by the GC.
> +     But since Qnil == 0, we can memset Lisp_Object slots as well.  */
> +  memset (v->contents, 0, zerolen * word_size);

Is Qnil==0 true for all builds?  Maybe it's better to do

   if (Qnil == 0)
      <newcode>
   else
      <oldcode>


-- Stefan

Re: [Emacs-diffs] master 3b48f99: Tune pseudovector allocation assuming Qnil == 0

[Paul Eggert]

Stefan Monnier wrote:
> Is Qnil==0 true for all builds?

Yes, that's been true since Jan 10.  I looked, and there are some other places 
where we can profitably prefer memset to assigning Qnil by hand; I did this in 
commit b7f83adda5a32140811e8e7decc4394d64cada3d just now.

> Maybe it's better to do
>
>    if (Qnil == 0)
>       <newcode>
>    else
>       <oldcode>

It's a bit simpler to verify that Qnil is zero, as in the abovementioned commit.