> I won't define it until we have a real libpthread implementation that
> interacts correctly with libc and has all the right hairy semantics.
> An extra libpthread that is not integrated with libc will never be a
> true
> POSIX implementation.
This is going to be a dumb question, but what would that take?
When I've referred before to a "total revamp of hurd libc", that's pretty
much it. I'd like to more or less start over on signals, the hurd
critical
section stuff, and probably revisit all the locking schemes from scratch.
This is necessary to do POSIX semantics for signals, cancellation, and
their pthreads interactions, correctly and robustly.