Re: ws2/tests: Test AcceptEx with a deferred socket

[Paul Vriens]

Mike Kaplinskiy wrote:
> Paul,
>
> I don't like NT4 anymore. We're breaking up.
>
> Does the attached patch fix the latest failures (not counting the
> build warnings)?
>
> Mike.
>
> On Mon, Jul 20, 2009 at 7:05 AM, Paul Vriens wrote:
> > Mike Kaplinskiy wrote:
> > > Paul (and anyone who can test on x86_64):
> > >
> > > Can you confirm that the new patch gives no failures?
> > >
> > > Mike.
> > >
> > > On Fri, Jul 17, 2009 at 2:02 AM, Paul Vriens
> > > wrote:
> > > > Mike Kaplinskiy wrote:
> > > > > Paul,
> > > > >
> > > > > Does the attached fix test failures on at least NT4? The 4000 test
> > > > > failures is ok, it's just 3999 of the same failure. The failures are
> > > > > the same because of a mistake on my part.
> > > > >
> > > > > I didn't get a chance to test this since my virtualbox is down until a
> > > > > 2.6.31-compatible version of their kernel module is out.
> > > > >
> > > > > Mike.
> > > > >
> > > > > On Thu, Jul 16, 2009 at 3:40 AM, Paul Vriens
> > > > > wrote:
> > > > > > Mike Kaplinskiy wrote:
> > > > > > > This tests the following AcceptEx scenario:
> > > > > > >  WSAAccept->CF_DEFER->AcceptEx
> > > > > > > Windows seems to return the deferred socket with AcceptEx (verified on
> > > > > > > XP & 2K3).
> > > > > > >
> > > > > > > Mike.
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > ------------------------------------------------------------------------
> > > > > > >
> > > > > > >
> > > > > > Hi Mike,
> > > > > >
> > > > > > These new tests fail on all NT4 boxes (some 6 and 1 has 4000+ failures)
> > > > > > and
> > > > > > what looks like to be all Windows x64_64 systems. The strange thing (or
> > > > > > maybe not) is that the errors are the same on all of these (one box has
> > > > > > 4000+ failures but still has those same 6).
> > > > > >
> > > > > > See:
> > > > > >
> > > > > > http://test.winehq.org/data/tests/ws2_32:sock.html
> > > > > >
> > > > > > Could you have a look?
> > > > > >
> > > > > > I can do some tests if necessary on NT4, I don't have a Windows x86_64
> > > > > > system however.
> > > > > >
> > > > > > --
> > > > > > Cheers,
> > > > > >
> > > > > > Paul.
> > > >
> > > > This is what I now get on NT4:
> > > >
> > > > .....
> > > > sock.c:2819: Test failed: Did not get both connections, got 1
> > > > sock.c:2867: Test failed: bytesReturned isn't supposed to be 1
> > > > sock: 441754 tests executed (0 marked as todo, 2 failures), 4 skipped.
> > > >
> > > > --
> > > > Cheers,
> > > >
> > > > Paul.
> >
> > Hi Mike,
> >
> > No failures on NT4 here with this one.
> >
> > --
> > Cheers,
> >
> > Paul.

Got a compile warning:

sock.c:2784: warning: passing argument 1 of 'CancelIo' makes pointer 
from integer without a cast

Still one failure with this one but that's because you reversed the 
return value checking of CancelIo.

So this does the trick for me (fixed warning and failure):

bret = CancelIo((HANDLE)listener);
         ok(bret, "Failed to cancel failed test. Bailing...\n");
         if (!bret) return;