Eric Haszlakiewicz <erh@nimenees.com> writes:

> > So that the right thing happens for a pthread and a non-pthread program. 
> > Other OSs have libc and libc_r. We didn't want to do that, and this 
> > behavior lets us.
> 
> 	By "right thing" you mean non-pthread program doesn't have the 
> pthread_* functions defined?  Is that really a requirement?
> If so, that seems to mean that even if we wanted to accept the performance
> hit, it would be incorrect to just dump pthread into libc.

The "right thing" here is protection of libc-internal data structures
(stdio, malloc, etc) in the presence of threads.

I think the rest of this is addressed by my earlier messages in the
thread, so I won't bother to repeat it.

        - Nathan