Paul Green <[EMAIL PROTECTED]> writes:
>Maybe an I/O layer should be required to explicitly declare itself
>thread-safe?  

The IO layer's ->Dup method is free to "fail" in any way it chooses
when asked to do a clone-for-thread.

>(Or whatever thread-xxx term I mean here).  Any layer not
>declared thread-safe should be removed when a new thread is created. (And
>now perl can print a warning, since it is doing something you might not like
>& might not realize is happening).

It may make sense eventually to add a flag bit and do the fail-ing in perlio.c 
but until we see a "pattern" in types of problem it is better 
if layer's own code decides what to do.

-- 
Nick Ing-Simmons
http://www.ni-s.u-net.com/

Reply via email to