Re: emacs-28 1a1b206: Adapt the recent 'num_processors' change to MS-Windows

Re: emacs-28 1a1b206: Adapt the recent 'num_processors' change to MS-Windows

[Stefan Kangas]

eliz@gnu.org (Eli Zaretskii) writes:

> +/* Emulate Gnulib's 'num_processors'.  We cannot use the Gnulib
> +   version because it unconditionally calls APIs that aren't available
> +   on old MS-Windows versions.  */

I'm surprised that the above is not fixed in Gnulib itself.

Does it make sense that Emacs has better compatibility for old versions
of Windows than the GNU portability library?

Re: emacs-28 1a1b206: Adapt the recent 'num_processors' change to MS-Windows

[Paul Eggert]

On 10/11/21 6:41 AM, Stefan Kangas wrote:

> Does it make sense that Emacs has better compatibility for old versions
> of Windows than the GNU portability library?

Yes, if Emacs supports obsolete MS-Windows versions that the rest of the 
GNU world doesn't.

If Eli has the time, perhaps he could merge the relevant obsolete 
MS-windows code into the corresponding Gnulib module. Although this 
might make the Emacs source cleaner overall, it isn't urgent as it 
likely wouldn't benefit other GNU projects (which is the point of Gnulib).

Re: emacs-28 1a1b206: Adapt the recent 'num_processors' change to MS-Windows

[Eli Zaretskii]

> From: Paul Eggert <eggert@cs.ucla.edu>
> Date: Mon, 11 Oct 2021 08:21:26 -0700
> Cc: Eli Zaretskii <eliz@gnu.org>, emacs-devel@gnu.org
> 
> On 10/11/21 6:41 AM, Stefan Kangas wrote:
> 
> > Does it make sense that Emacs has better compatibility for old versions
> > of Windows than the GNU portability library?
> 
> Yes, if Emacs supports obsolete MS-Windows versions that the rest of the 
> GNU world doesn't.

Not "the rest of the GNU world", Gnulib.  It's the decision of the
Gnulib developers to drop support for older MS-Windows versions.  Not
all the GNU projects followed suit.

> If Eli has the time, perhaps he could merge the relevant obsolete 
> MS-windows code into the corresponding Gnulib module. Although this 
> might make the Emacs source cleaner overall, it isn't urgent as it 
> likely wouldn't benefit other GNU projects (which is the point of Gnulib).

I'll probably add to the Emacs version of this function the relevant
parts of the Gnulib code -- on the master branch, not on emacs-28,
where I think stability (the current code works for months without any
problems) tramps flexibility.  But submitting that to Gnulib is a lost
cause, IME, since Gnulib folks aren't interested in complicating their
code for the benefit of Windows versions they consider EOLed.  I had
my share of arguments about that with Bruno and others, and have no
intention to argue about that again.  The Gnulib folks made their
decision, and that's fine with me.

Re: emacs-28 1a1b206: Adapt the recent 'num_processors' change to MS-Windows

[Eli Zaretskii]

> From: Stefan Kangas <stefankangas@gmail.com>
> Date: Mon, 11 Oct 2021 06:41:18 -0700
> 
> Does it make sense that Emacs has better compatibility for old versions
> of Windows than the GNU portability library?

These are two separate development teams with two different
perspectives.  It's a small wonder that the decisions don't coincide.