On Thu, Oct 29, 2015 at 04:15:33PM +0000, Alan Burlison wrote:

There was an attempt to interpret POSIX that way, with which I still
disagree. If a FD is closed or reassigned then any current pending
operations on it should be terminated.

Could the esteemed sir possibly be ars^H^H^Hprevailed upon to quote the exact
place in POSIX that requires such behaviour?

This is getting ridiculous - if we are talking about POSIX-mandated behaviour
of close(), please show where is it mandated.  Using close(2) on a descriptor
that might be used by other threads is a bloody bad design in userland code -
I think everyone in this thread agrees on that.  Making that a recommended way
to do _anything_ is nuts.

Now, no userland code, however lousy it might be, should be able to screw
the system.  But that isn't the issue - our variant is providing that just
fine.

BTW, "cancel accept(2) because sending a signal is hard" is bogus anyway -
a thread in accept(3) just about to enter the kernel would get buggered
if another thread closes that descriptor and the third one does socket(2).
_IF_ you are doing that kind of "close a descriptor under other threads"
thing, you need to inform the potentially affected threads anyway, and
you'd better not rely on them being currently in kernel mode.