Per Bothner writes: > Does this seem like a reasonable patch? I think so. Can you please check this in? > (I would be happier if I understood why there wasn't a problem > before!) It might have been with rest_of_decl_compilation which changed on Monday the 26th: http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/toplev.c.diff?r1=1.427&r2=1.428&f=h ./A