SL Baur <steve(a)xemacs.org&gt; writes: Hm, indeed same here, but then here on my sun "Benadir" doesn't crash either -- strange... Not really, I don't see a significant change in the source. I only protected the prompt in the '[kK]' cases, in all the others callN is used and provides the gcpro. This violates 'caller protects' but I guess that's acceptable here. BTW, the nested version of GCPRO isn't documented in the internals manual. I only found it by chance in lisp.h. I'll send a patch for internals.texi later this week. Anyway here's the patch: 1998-07-27 Gunnar Evermann <Gunnar.Evermann(a)nats.informatik.uni-hamburg.de&gt; * callint.c (Fcall_interactively): GCPRO prompt string before passing it to Fread_key_sequence --- callint.c.orig-21.0-b50 Wed May 6 11:46:26 1998 +++ callint.c Mon Jul 27 12:22:26 1998 ＠＠ -715,7 +715,14 ＠＠ } case 'k': /* Key sequence (vector of events) */ { - Lisp_Object tem = Fread_key_sequence (PROMPT (), Qnil, Qnil); + struct gcpro ngcpro1; + Lisp_Object tem; + Lisp_Object key_prompt = PROMPT (); + + NGCPRO1(key_prompt); + tem = Fread_key_sequence (key_prompt, Qnil, Qnil); + NUNGCPRO; + visargs[argnum] = Fkey_description (tem); /* The following makes `describe-key' not work with extent-local keymaps and such; and anyway, it's ＠＠ -728,7 +735,14 ＠＠ case 'K': /* Key sequence (vector of events), no automatic downcasing */ { - Lisp_Object tem = Fread_key_sequence (PROMPT (), Qnil, Qt); + struct gcpro ngcpro1; + Lisp_Object tem; + Lisp_Object key_prompt = PROMPT (); + + NGCPRO1(key_prompt); + tem = Fread_key_sequence (key_prompt, Qnil, Qt); + NUNGCPRO; + visargs[argnum] = Fkey_description (tem); /* The following makes `describe-key' not work with extent-local keymaps and such; and anyway, it's