Re: fork() loses when thread self port's refcount is max'ed out

[Neal H. Walfield]

> So you have asserts that have the side effect of leaking send
> rights?

I was working under the assumption that we do not deallocate send
rights returned by mach_thread_self.  As I will now understand that
that assumption is wrong, I will correct the code.

My comment was more to justify the use of mach_thread_self; I was
expecting someone to say we should use hurd_thread_self (in fact,
Marcus did), however, the assert is checking if the thread's TSD area
is coherent which hurd_thread_self depends on.