Nicholas Clark <[EMAIL PROTECTED]> wrote: > mem_sys_free(pio_registered_layers);
Ah, yep. > AWOOGA! AWOOGA! GLOBAL VARIABLE. NOT IN AN INTERPRETER STRUCTURE. > And I'm not sure what the correct fix is. But it sure ain't threadsafe > currently. Interpreter (thread) destruction is done by aquiring the C<interpreter_array_mutex>. C<pio_registered_layers> should of course be freed only for the last interpreter that is that one w/o C<interpreter->parent_interpreter>. The current fix is a bit too complicated but ok. But the first thing to solve is: should the IO subsystem be shared for threads or separate. See another post WRT Cygwin thread problems. > Nicholas Clark leo