On 2010-05-28 10:04 +0200, [email protected] wrote: > Sven Joachim wrote: >> On 2010-05-28 07:45 +0200, Trent W. Buck wrote: >> >> > I set up a chroot containing GNU Emacs 23 and (almost) every elisp >> > package, with the intention of looking for libraries that shadow >> > emacs23's libraries with obsolete older versions. >> >> Thank you for doing that. >> >> > Imagine my surprise when *even with -Q*, a number of libraries are >> > shadowed (see attached transcript). Why are third-party libraries >> > being added to the load path when -Q is specified? >> >> Because of the subdirs.el that is installed by the upstream Makefile > > Ah, I heard from the dictionaries-common maintainer about this file, > but I didn't realize it was upstream's idea. > >> under /usr/share/emacs23/site-lisp. Since Debian packages are supposed >> to add their lisp directories explicitly to load-path via >> debian-pkg-add-load-path-item, it might make sense to remove this file >> from the emacs23 package. Would you like to test which packages break >> when you move /usr/share/emacs23/site-lisp/subdirs.el out of the way? > > Hmm, can you think of a good way to test for "breaks"? > > Just removing it and running emacs will find superficial breakage, but > for most values of "foo", I don't use foo-mode regularly, so I won't > be able to detect deeper breakages.
All that subdirs.el does is to add the subdirectories to load-path, so comparing the value of load-path with and and without subdirs.el should give an indication what might break. Note that with subdirs.el there are many duplicates in load-path that will probably disappear without it, e.g. you get an entry "/usr/share/emacs23/site-lisp/foo" added by the package and an entry "/usr/share/emacs/23.1/site-lisp/foo" added by subdirs.el, but these are the same directories. Sven -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
