Hello, > Haven't we seen this bug before in the sasl and socket routines? I > remember being surprised that it hadn't bit the db code. I think > something like the following was supposed to work: yes we did. And I understand now, why it never triggered before: It only happens if you compile header cache with db4 support. For gdbm and qdbm (not upstream yet) there is no such thing as ->open ... > (h->env->open)(...) I guess. :-) Thomas