broken smbtorture lease cases

[Steve French]

On Wed, Mar 11, 2015 at 6:16 PM, Michael Adam <obnox at samba.org> wrote:
> On 2015-03-11 at 18:06 -0400, Steve French wrote:
>> On Wed, Mar 11, 2015 at 5:36 PM, Jeremy Allison <jra at samba.org> wrote:
>> > On Wed, Mar 11, 2015 at 03:51:29PM -0500, Steve French wrote:
>> >> Looks like various smbtorture lease test cases are broken too against
>> >> 4.2 (not just the Microsoft ones with the directory rename bug and
>> >> three trivial return code problems that I have posted fixes for
>> >> earlier).  A different three failed to Windows 8.1 (multibreak, lock1
>> >> and v2_epoch2).
>> >>
>> >> I simply ran:
>> >> /usr/local/samba-4.2/bin/smbtorture  //localhost/scratch smb2.lease
>> >>
>> >> Pretty minimal smb.conf so unlikely to be that (and even fails when
>> >> kernel sharemodes/sharemode locking are disabled on share - whether
>> >> off or on):
>> >>
>> >>    ea support = yes
>> >>    store dos attributes = yes
>> >>    strict allocate = yes
>> >>    delete readonly = yes
>> >>    smb2 leases = yes
>> >>
>> >> In addition to the two failing, 3 are skipped as expected -
>> >> dynamic_share (and v2_request and v2_request_parent require directory
>> >> leases so they are skipped) but the following are not skipped and
>> >> presumably are supportable.
>> >>
>> >> test: request
>> >> time: 2015-03-10 04:12:52.548165
>> >> time: 2015-03-10 04:12:52.561230
>> >> failure: request [
>> >> ../source4/torture/smb2/lease.c:170: Incorrect status
>> >> NT_STATUS_OBJECT_NAME_NOT_FOUND - should be NT_STATUS_OK
>> >
>> > This is due to you not having the correct VFS
>> > modules loaded. I told you to look in the test
>> > case code :-).
>>
>> I am looking in the test case code ... nothing in the description of
>> the test case,
>> or in the rc handling or in the top of source4/torture/smb2/lease.c
>> that mentions
>> "vfs" or "module" or any dependencies ...
>>
>> Is there another lease.c helper file I am missing?  I am just running
>> source4/smbtorture
>
> Jeremy is referring more to how the test program
> is run in our selftest setup. I.e. see where it
> is planned in source{3,4}/selftest/tests.py, check
> the environments against which it is run
> (selftest/target/Samba{3,4}.pm), which defines
> the samba configs, etc.
>
> And then, check in selftest/knownfail and selftest/skip
> whehter the test is listed there.
>
> Looking at the source code of the actual test
> only reveals half the truth.

This gets confusing. It looks like the whole category is skipped
(which would explain a lot).

In any case - loading the four vfs modules mentioned in Samba3.pm does
cause the mystery test (smb2.lease.request) to work ie

vfs objects = acl_xattr fake_acls xattr_tdb streams_depot


There isn't much else useful about leases that I could track down in
the test setup directories you mention though. e.g.
I grep -i for lease and nothing in the selftest directories and the
whole category (smb2.lease) is in the skip category and nothing in the
knownfail.

But at least down to one (known) failure to Samba - and eventually I
could check why the others fail to Windows 8.1

A little harder than I expected to extend them though to try to match
MS test cases.

-- 
Thanks,

Steve