This fixes [very probable cause for] crashes seen by Karl and Greg, and solves extra minibuffers problem in ediff, reported by Greg. The first fix fixes an error: I can crash XEmacs with my carefully crafted lisp snippet (tm). I bet a dollar that this is the cause for reported crashes. Kirill 1998-04-24 Kirill M. Katsnelson <kkm(a)kis.ru&gt; * frame.c (Fmake_frame): GC-protect new frame immediately upon creation. 1998-04-27 Kirill M. Katsnelson <kkm(a)kis.ru&gt; * frame.c (delete_frame_internal): Fixed spurious minibuffer frames when creating minibufferless frames (reverted to 20.3 source, change not in ChangeLog) Index: src/frame.c =================================================================== RCS file: /var/cvsroot/ntxemacs/src/frame.c,v retrieving revision 1.4 diff --unified=2 -r1.4 frame.c --- src/frame.c 1998/04/05 13:16:54 1.4 +++ src/frame.c 1998/04/26 17:33:53 ＠＠ -417,4 +417,5 ＠＠ f = allocate_frame_core (device); + XSETFRAME (frame, f); specbind (Qframe_being_created, name); ＠＠ -446,5 +447,4 ＠＠ signal_simple_error ("Invalid value for `minibuffer'", minibuf); - XSETFRAME (frame, f); update_frame_window_mirror (f); ＠＠ -1512,5 +1512,5 ＠＠ double_break_1: - if (!called_from_delete_device) + if (called_from_delete_device < 0) /* then we're being called from delete-console, and we shouldn't try to find another default-minibuffer frame for the console.