AJ> Michael, do you also get a working XEmacs if you compile mule-charset AJ> with -O0? AJ> I don't know how to continue at the moment :-(. Martin, do you have AJ> more clues for me? ;-) I was able to reproduce the problem with the latest egcs snapshot. I am feeling quite confident this is a compiler bug, miscompiling Fmake_char. Almost any change to that function will result in a working XEmacs. How do we handle this? Anyone have any contacts with the egcs developers? I tried to isolate a small test case, but failed. This bug wants to go away at the slightest nudge. So some egcs person will have to look at the compiled code for Fmake_char. It will likely be easy for us to work around this bug, making changes that should have been made long ago in any case for performance reasons. Nevertheless, we should try to get the egcs folks to fix it. Martin