Gnustep-base thread creation bug. continued.

[Chris Ball]

I spent a bit of time looking at this today and realized that the way
thr-pthead.m is initialized isn't exactly my problem.  It turns out that
my build of base is using gcc's primitives directly and not the thr-pthread.m
here.  This leaves me more confused as it looks like the gcc primitives should
be ok but they are not.  If I build a version of my test program that doesn't
use the foundation class and just calls objc_thread_detach directly I am still
running into the same problems.

I'm still looking into things but at this point my best theories are:
1.) The current slackware's gcc is using thr-pthread instead of thr-posix which
would break things since the pthread init doesn't set detachstate.
2.) The is a deeper gcc problem where either __objc_exec_class isn't being
called or it isn't calling __objc_init_thread_system.

Right now I'm betting on the distribution being dodgy but until I try building a
new gcc or two I won't know for sure.  Mind you I'm pretty sure the old linux
pthread code was posix enough for this to be a problem regardless...

I'll report back as I go.

        Chris.