Patchset to add asynchronous open/close to master

[Jeremy Allison]

On Fri, Jun 22, 2012 at 12:45:28PM +0200, Andreas Schneider wrote:
> On Thursday 21 June 2012 15:02:23 Jeremy Allison wrote:
> > On Thu, Jun 21, 2012 at 05:55:48PM -0400, simo wrote:
> > > Neat,
> > > should we add some abort() calls to check the next time glibc decides to
> > > use SETXID on anything ... who know some smart fella may decide to set
> > > it on getpid() :-P
> > 
> > I think we only need to care about the setxxx
> > calls, not any of the getxxx calls.
> > 
> > And we never call the setxx calls directly,
> > but always through the wrappers in source3/lib/util_sec.c
> > so if we make them call the syscalls directly we should be
> > ok (IMHO). Plus the wrappers always do a getXXX
> > check afterwards to ensure the set succeeded, and
> > abort the entire process if any of them failed.
> 
> Don't forget about lib/uid_wrapper which overwrites set*uid calls in make 
> test.

Yes Andreas, I'm quite aware of lib/uid_wrapper, and how
to wrap that to cope with make test. The patch I have
creates wrapper functions (linux_thread_setXXX) so that
they can be replaced inside uid_wrapper.

That's part of the patch under test of course. It wouldn't
pass make test without it.

Jeremy.