Re: --disable-features in 109483

[Chong Yidong <cyd@gnu.org>]

Dmitry Antipov <dmantipov@yandex.ru> writes:

> On 08/08/2012 07:33 AM, Chong Yidong wrote:
>
>> You are missing a NEWS file.
>
> And doc/lispref/anti.texi too :-).

This kind of change probable not worth documenting in the Antinews.  But
anyway, the usual procedure is to document in NEWS first, then update
the documentation later during the run up to the pretest.

> 1) Testing: check whether the work on FEATURE1, FEATURE2, ... FEATUREN
> do not broke the build without FEATURE1, ... and without all of these
> features.
>
> 2) Building small OS distribution to fit in CD, 100Mb or so. Such
> a distributions usually prefers vi{m} and/or small Emacs clones,
> but who knows; such a distribution is not necessary X-less.
>
> 3) Using minimalistic desktop environments on top of the old hardware.

(2) and (3) are very theoretical.  If you are compiling on an old
machine which doesn't doesn't have dbus libraries etc., configure
already automatically omits them, so there's no benefit from this flag.
As for people doing cross-compilation, like Debian packagers, they are
already sophisticated enough to figure out the appropriate configure
flags.

Anyway, disabling support for various *shared* libraries to get a
"performance improvement" is an instance of false optimization if there
ever was one.

(1) is a reasonable use, but only because you can do stuff like
--disable-features --with-rsvg to re-enable rsvg.  But this also
indicates turning off X too.  Also, I think the option should be renamed
to --without-all, since it actually affects the --with/--without options
rather than the --enable/--disable options.