Didier Verna <verna(a)inf.enst.fr&gt; writes: OK, I'll try again: First, I give *my* interpretation of how x_get_mouse_position() works (this could be totally wrong): 1) call XQueryPointer to find out over which Window (if any) the pointer currently is. ('child_window') 2) call XTranslateCoordinates() with this Window as the destination window. 'If the coordinates are contained in a mapped child of dest_w, that child is returned to child_window.' (quote from man page) 3) repeat 2) until it returns None. 4) call x_any_window_to_frame with the previous window and store the frame in an Lisp_Object. 5) get the coordinates from XTranslateCoordinates() with the frame's text widget as the destination window. The problem is that XQueryPointer can return a window that results in a Bad Window on the first call to 2). Jan suggested to move the call to x_any_window_to_frame between 1) and 2). This sounds reasonable but doesn't work. In all my tests x_any_window_to_frame(d, child_window) returns NIL at this point. If I call x_any_window_to_frame after each iteration, I only get the frame after at least two iterations of the loop. I used a freshly started XEmacs (with toolbar and scrollbars) displaying only the *scratch* buffer. The Pointer was over the text in that buffer. The above loop did 5 iterations in this case. I don't understand this :-) more precisely I have no idea how to do that. (maybe using this error_expected stuff?). Please note that the error happens in a call to XTranslateCoordinates(). Gunnar (who should really read those damned X manuals...)