[experimental PATCH] GC-Leak in tooltalk

Hi everyone, I've finally found the bug which makes XEmacs crash if compiled with SUN-CC, -gung-ho and workshop support (as reported by Georg and myself, e.g. <13638.20762.848751.664065＠verve>, <199805041403.QAA00409(a)nats7.informatik.uni-hamburg.de>). I spend two days single-stepping through the GC and FByte_code :-) The two Lisp_Objects mentioned below get collected (before dumping I guess). In the dumped binary Tooltalk_Pattern_plist_str points to some string (in the locate-file-hashtable of some string of of the load-path vars), if a GC occurs before TT-Initialization it points to nowhere and everything goes to hell... I posted this to XEmacs-beta as well because I'm a complete novice to XEmacs-C-Internals and I guess the GC-stuff is a rather dangerous place to start hacking :-) So maybe the GC and/or TT-experts can comment -- is this the right way and why doesn't this show up when not using gung-ho? Anyway this fixes the problem for me (XEmacs starts up OK and after installing the pc and vc-packages TT-communication to workshop seems to work. Gunnar 1998-05-26 Gunnar Evermann <Gunnar.Evermann(a)nats.informatik.uni-hamburg.de&gt; * tooltalk.c (vars_of_tooltalk) added staticpro for Tooltalk_Message_plist_str and Tooltalk_Pattern_plist_str --- src/tooltalk.c~ Fri May 1 03:11:07 1998 +++ src/tooltalk.c Tue May 26 15:31:36 1998 ＠＠ -1421,6 +1421,9 ＠＠ Tooltalk_Message_plist_str = build_string ("Tooltalk Message plist"); Tooltalk_Pattern_plist_str = build_string ("Tooltalk Pattern p plist"); + staticpro(&Tooltalk_Message_plist_str); + staticpro(&Tooltalk_Pattern_plist_str); + #define MAKE_CONSTANT(name) do { \ defsymbol (&Q_ ## name, #name); \ Fset (Q_ ## name, make_int (name)); \