It's me FKtPp ;) writes: > with the same procedure and different xemacs start arguments, I got the > following result: > > xemacs -vanilla : no assertion error > xemacs -q : assertion error > xemacs -q --no-site-file : assertion error The difference between -vanilla and "-q -no-site-file" is whether "early" packages are used or not. Early packages are the ones in ~/.xemacs. I would start with the ones that are shadows: Load-Path Lisp Shadows: ---------------------- (/usr/share/emacs/site-lisp/mercurial /home/fktpp/.xemacs/xemacs-packages/lisp/vc/mercurial /home/fktpp/.xemacs/xemacs-packages/lisp/xemacs-base/easy-mmode /usr/local/share/xemacs-21.5-b29/lisp/easy-mmode /home/fktpp/.xemacs/xemacs-packages/lisp/xemacs-base/regexp-opt /usr/local/share/xemacs-21.5-b29/lisp/regexp-opt /home/fktpp/.xemacs/xemacs-packages/lisp/apel/alist /usr/local/share/xemacs-21.5-b29/lisp/alist /home/fktpp/.xemacs/xemacs-packages/lisp/build/build-report /usr/local/share/xemacs-21.5-b29/lisp/build-report) APEL in particular is really evil. It needs to be compiled with the binary it runs on. (Theoretically the most recent XEmacs should run bytecode compiled by anything earlier, but ....) regexp-opt is an obvious suspect when regexps are involved, of course.

That said, it should not be possible to crash XEmacs from Lisp, so there's an XEmacs bug here somewhere.