On Nov 30, 2007 2:34 PM, Jerry James <james(a)xemacs.org&gt; wrote: Here's how it fails. Way down in ecb-activate--impl (ecb.el) is a block that starts: (when ecb-running-xemacs (let ((dummy-buf-name " *dummytogetglobalmap*")) (save-excursion (set-buffer (get-buffer-create dummy-buf-name)) (add-submenu nil ecb-minor-menu) (kill-buffer dummy-buf-name))) That kill-buffer invocation kicks off the fun. ECB put a function, ecb-kill-buffer-hook (ecb.el), on kill-buffer-hook. Execution gets almost all the way to the end of that function, where it calls ecb-get-current-visible-ecb-buffers (ecb-layout.el), to make sure we aren't killing any special ECB buffers. That calls ecb-canonical-ecb-windows-list (ecb-layout.el) to give us a list to map window-buffer over. That calls ecb-canonical-windows-list (ecb-util.el) to give us a list of all windows from which we can remove the non-ECB windows. That calls ecb-window-list (ecb-util.el) on the ECB frame. Note that the function ecb-window-list calls window-list if ecb-running-emacs-21; otherwise, it executes its own version of window-list. Since we have window-list, we could throw that all away and just call window-list, too. Unfortunately, that doesn't help with the current bug. The replacement window-list code in ecb-window-list calls save-window-excursion, which calls current-window-configuration, on which ECB has put an after advice! So off to ecb-layout.el we go, where we see that if (not ecb-windows-hidden), we call ecb-get-current-visible-ecb-buffers. Ummm... haven't we been here before? And around and around and around we go until the stack is exhausted. Using our own window-list function does not help, because it also calls save-window-excursion, which calls current-window-configuration. -- Jerry James http://loganjerry.googlepages.com/ _______________________________________________ XEmacs-Beta mailing list XEmacs-Beta(a)xemacs.org http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta

On Dec 1, 2007 1:06 PM, Ville Skyttä <scop(a)xemacs.org&gt; wrote: Thanks, Ville. I've looked at Henry S. Thompson's approach. Unfortunately, his patch doesn't work for me at all. I still fall into an infinite loop pretty quickly. I thought I would try a fundamentally different approach. I haven't finished yet; there's something weird going on with the minibuffer with this patch. Before I go any farther, though, I want to know what Klaus thinks. Does this approach have any hope of working, Klaus, or do you really need to know when random code outside of ECB calls current-window-configuration and set-window-configuration? Incidentally, both XEmacs 21.4 and 21.5 have a window-list function that calls walk-windows, which uses next-window to go down the list. Therefore, the list returned by window-list is consistent with the order in which windows are visited with next-windows. I know you were wondering about that. Regards, -- Jerry James http://loganjerry.googlepages.com/ _______________________________________________ XEmacs-Beta mailing list XEmacs-Beta(a)xemacs.org http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta