[PATCH 0/3] Fix master to pass smbtorture smb2.acls and raw.acls tests - v2 - with Simo fixes.

[Jeremy Allison]

On Tue, Nov 20, 2012 at 08:21:12AM +0100, Stefan (metze) Metzmacher wrote:
> 
> I'd prefer to remove this completely (and only expect windows behavior)
> and add failing tests to the knowfail file.

There are some cases for which this makes sense, but I'm not sure ACLs
are one of them. Marking this "knownfail" loses coverage of a large
number of regression test cases that we get right, so long as we
expect our own default ACL back instead of the Windows one.

These cases are *important*, as they test that we get inheritance
right over all ACL cases. To throw them all out just because we
have a different default ACL on non-inheritance is a big case of
throwing the baby out with the bathwater :-).

In order to match Windows behavior here we'd need a separate
"running under test" flag in smb.conf to generate default Windows
ACLs that make no sense under normal circumstances on a POSIX
backend. I'm not sure that's better :-).

> If this behavior test is part of a larger test, we should isolate it
> as a separate test.

It is a separate test (or at least can be run as a separate test,
currently it's a sub-test under raw.acls and smb2.acls).

Jeremy.