Consensus on MT-, AS- and AC-Safety docs.

[Torvald Riegel]

On Sat, 2013-11-23 at 10:56 -0200, Alexandre Oliva wrote:
> On Nov 22, 2013, Torvald Riegel <triegel@redhat.com> wrote:
> 
> >> >      may-leave-locks-held-if-cancelled-asynchronously,
> 
> >> leak-locks
> 
> > Those suggestions all look fine to me.
> 
> Can you explain why lockleak suggests it's the lock object that fails to
> be destructed, but leak-locks is fine as a shorthand for the above?!?
> 

I can't; I agree it's contradictory to the other discussion we had, but
it's not meant that way :)
I think that leak-locks would be fine compare to the longer sequence of
words (if we want to say that it "leaks locks").  Nonetheless, I still
think that it leaks lock _acquisitions_, and that we should say that if
we can all be comfortable with that.  For example, we could perhaps also
call it "leak-lock-acquisitions", "may-not-release-locks", or something
like that.