Cancellation investigations.

[Ulrich Drepper]

Kaz Kylheku <kaz@ashi.footprints.net> writes:

> Since the extricate pointer refers to an auto object, we must ensure that
> it's never used by thread2 after thread1 finishes. 

I've corrected this now.  Thanks.

> If locks are undesirable, something else needs to be invented to
> delay thread1 while someone is using its extricate struct.  It's
> clear that most of the time, there is no problem; eliminating a rare
> race condition should not impose a big penalty on the frequent case.

I do have plans to tackle all this locks we currently use.  But not
with the current source base.  It is time for a rewrite which I did
start when the kernel finally had some signs of better thread support.
More on this once 2.2 is out.

-- 
---------------.                          ,-.   1325 Chesapeake Terrace
Ulrich Drepper  \    ,-------------------'   \  Sunnyvale, CA 94089 USA
Red Hat          `--' drepper at redhat.com   `------------------------