Re: [Emacs-diffs] trunk r117941: Default to stack objects on non-GNU/Linux, non-DOS_NT platforms.

[Dmitry Antipov <dmantipov@yandex.ru>]

On 09/25/2014 12:15 PM, Eli Zaretskii wrote:

> It was done in complete silence, without prior discussions wrt
> whether to make this the default.  (Yes, I suggested that, but no
> one replied, and we never discussed it.)

IMO 1) any discussions whether to make USE_STACK_LISP_OBJECTS the default
makes no sense until we know whether it's worth the complexities at all.
For the latter, we need a lot of feedback from users, preferably with
the very different usage patterns and workloads.  The simplest (and the
only reliable) method is to enable it by default and see what happens.
OTOH I'm not against making it opt-out, as you suggested.

> It arbitrarily excludes the native MS-Windows builds from this
> feature, for no good reasons: the 64-bit Windows build has no
> problems with it

IMO 2) responsible developer should not enable any code she/he can't
test.  If you are rather sure that it should work on 64-bit MS-Windows,
feel free to add this class of systems to an appropriate #ifdef (to be
honest, this is simpler and friendlier than writing such an indignant
e-mails).

> In general, every feature that exists only on some platforms is a
> Bad Thing, as it introduces maintenance problems and in particular
> makes people who work on different platforms unable to usefully
> compare what they see and solve problems reported by others.

USE_STACK_LISP_OBJECT is not such a feature.  It's rather under
construction and not yet populated to all (most?) platforms we aim
to support.  If you want to help with that, you're always welcome.

> In sum, I don't understand these sneaky practices, and I wish they'd
> stopped.

Sneaky?  I would call it too brave.

Dmitry