Bugzilla Bug 184588

The site-init file 60gtk-doc-gentoo.el contains the statement (load-library
"gtk-doc") which will load gtk-doc.el at Emacs startup time. It would be
desirable to postpone loading of gtk-doc.el until it is first needed, which can
be done using Emacs's autoload mechanism.

See <http://overlays.gentoo.org/proj/emacs/wiki/sitefile%20problems> for
further details.

Created an attachment (id=124225) [edit]
60gtk-doc-gentoo.el

Modified site-init file using autoload.

This file is just installed; it's created and maintained by upstream.  Please
file an upstream bug with the fix, and post the URL in the URL field above.

(In reply to comment #2)
> This file is just installed; it's created and maintained by upstream.

This is true for gtk-doc.el for which I don't propose any change.

However, 60gtk-doc-gentoo.el is the Gentoo site initialisation file. It resides
in ${FILESDIR} and is _not_ provided by upstream.

You're right, I was looking at the wrong file.


I don't use emacs, but I've run this by the local emacs expert, and he objects
to the proposed change as being "invasive".  I'm not sure what to do at this
point; can you maybe explain why you're doing this, instead of what all the
rest of the emacs site files do (namely require)?

(In reply to comment #4)
> I don't use emacs, but I've run this by the local emacs expert, and he objects
> to the proposed change as being "invasive".  I'm not sure what to do at this
> point; can you maybe explain why you're doing this, instead of what all the
> rest of the emacs site files do (namely require)?

O.K., let's look at the differences between the old and the new site file:
1. (load-library "gtk-doc") is replaced by autoload statements for the
   interactive functions.
2. Two global keybindings which would otherwise be done in gtk-doc.el itself.

I don't understand which part of this would be considered "invasive". Please
note that the standard procedure for the core packages included with Emacs
itself is to use the autoload mechanism (look into
/usr/share/emacs/*/lisp/loaddefs.el).

It is true that many of the Gentoo site files still contain require statements;
it is on our agenda to change this wherever possible. See the URL given in
comment #0 and tracker bug #174469 for details.

(In addition, in the case of gtk-doc it would not be possible to use "require"
since the file doesn't contain a corresponding "provide" statement.)

(I realize it doesn't have a provide statement; that can be fixed, if
necessary)

In particular, he strenuously objected to setting global keybindings, and liked
the require/provde setup.  Why is require/provide being removed?  It seems to
me that globally forcing settings on all users is a very bad idea, in general,
and settings should only be pulled in when needed.

Keep in mind: I'm *not* an emacs user, and don't care one way or the other.  If
this type of global setting is the way emacs on gentoo is going, far be it from
me to stop it...

Created an attachment (id=126388) [edit]
61gtk-doc-gentoo.el

Created an attachment (id=126390) [edit]
gtk-doc-1.8-emacs-keybindings.patch

Created an attachment (id=126392) [edit]
Diff for gtk-doc-1.8.ebuild

As discussed, here is a new site-init file, a patch that removes the global key
bindings, and a patch for the ebuild itself.

Committed.