failing tests in the testsuite on debian

[Jeremy Allison]

On Tue, Nov 15, 2022 at 10:52:08PM +0300, Michael Tokarev wrote:
>Thank you very much Jeremy for stepping in!
>
>15.11.2022 20:42, Jeremy Allison via samba-technical wrote:
>>On Tue, Nov 15, 2022 at 06:12:44PM +0300, Michael Tokarev via samba-technical wrote:
>
>>>Certain tests do continue to fail though, and I still don't see how
>>>to address this.
>>
>>Which tests. Specific tests sometimes are expected
>>to fail.
>
>Here's my current knownfail.d list for the --quick test run:
>
>^samba3.smb2.session\ enc.(reauth.|bind.|bind_negative.*|bind_invalid_auth|encryption-aes-.*)\(nt4_dc\)
>^samba4.ntvfs.cifs.ntlm.base.unlink.unlink\(rpc_proxy\)
>^samba4.rpc.echo\ against\ rpc\ proxy\ with\ domain\ creds\(rpc_proxy\)
>
>I found no way to exclude certain tests (but the previous email from Ralph
>gives some clues).  None of the above tests seems to have reasons for failure.
>
>I'm pasting the output from the failing tests below.
>
>
>[35(230)/370 at 58s] samba3.smb2.session enc(nt4_dc)
>UNEXPECTED(failure): samba3.smb2.session enc.reauth1(nt4_dc)
>REASON: Exception: Exception: ../../source4/torture/smb2/session.c:247: (&io1)->out.file_attr was 128 (0x80), expected 32 (0x20): out.file_attr incorrect
>UNEXPECTED(failure): samba3.smb2.session enc.reauth2(nt4_dc)
>REASON: Exception: Exception: ../../source4/torture/smb2/session.c:323: (&io1)->out.file_attr was 128 (0x80), expected 32 (0x20): out.file_attr incorrect
>...(many of them)...
>
>this is ATTRIBUTE_ARCHIVE vs ATTRIBUTE_NORMAL. The test is
>expecting the newly created files to have "a" attribute but
>apparently there's none. Just a thought here - maybe it has
>something to do with this user being unable to create extended
>attributes, or maybe tmpfs (where I run the whole thing) does
>not support it.

Yes, the DOS attributes are stored in an EA.

Tests that create EA's should be run against a share
using vfs objects including vfs_xattr_tdb so they
don't depend on the underlying filesystem for EA
support.

I see the bug. The definition of [tmpenc]
has:

vfs objects = dirsort

It should be:

vfs objects = dirsort xattr_tdb

as you're running it against a filesystem
(tmpfs) that doesn't support EAs.