BP> "Stephen J. Turnbull" <turnbull(a)sk.tsukuba.ac.jp&gt; writes: -BP> I don't have time to dig into this deeply today, but I did a -BP> synch up to r21-2-45 in my local workspace for XEmacs/GTK and -BP> when it goes to dump, it cannot (require 'gtk-ffi), but that -BP> file is in lisp/ui/gtk-ffi.elc* -BP> Martin, could any of the changes you made to the make process -BP> recently be causing this? >> Michael Sperber made some adjustments, and I think (I'd have to dig up >> the thread and patch) except for lisp/mule and lisp/term, which are >> special-cased, the code doesn't search subdirectories of lisp. This is >> something we've been aiming for about a while; it reduces startup time >> noticably to get rid of the stats (especially on MS systems). Yes. Well, almost: lisp/mule is special-cased, files in lisp/term are loaded via an explicit prefix. BP> I don't really, there were just 50 or so files in there at one point, but I BP> combined a lot of them before the final release, and it is now down to 13, BP> so it is probably ok to move them up a level. I think so, too, at least for now. For post-21.2, I'd like to move to a structure like this: lisp/xemacs lisp/mule lisp/gtk ... which avoids the problems (conceptual and technical) we've been having. -- Cheers =8-} Mike Friede, Völkerverständigung und überhaupt blabla