[PATCH] Version 2 - Additional fix and test for bug #12899 - smbclient "setmode" no longer works to clear attribute bits due to dialect upgrade

[Jeremy Allison]

On Fri, Jul 14, 2017 at 04:47:16PM -0700, Jeremy Allison wrote:
> On Fri, Jul 14, 2017 at 04:17:56PM -0700, Jeremy Allison via samba-technical wrote:
> > Steve's fix for the smbclient setmode command
> > exposed a bug in our SMB1 implementation once I
> > wrote the test for it (our SMB2 code was already
> > correct).
> > 
> > An SMB1 SMBsetatr command should ignore the
> > set attribute request if a value of *zero* is
> > sent, we were ignoring it if a value of
> > FILE_ATTRIBUTE_NORMAL was sent.
> > 
> > FILE_ATTRIBUTE_NORMAL is already correctly
> > filtered out inside the set_dosmode() function.
> > 
> > Our SMB2 code already did this correctly.
> > 
> > Fix contains the SMB1 fix and then the
> > torture test that exposed the bug.
> 
> Bah. I missed this fix meant a change in
> SMB1 server behavior (SMBsetatr 0 means
> clear all attributes, the torture code
> depends on that).
> 
> Here's the updated version that passes
> all the existing samba3.smbtorture_s3
> tests. Sorry for the mistake.

ARGGGHHHH. It may pass samba3.smbtorture_s3,
but this isn't what Windows does (tm) - as
tested against Win2KR12 SMB1 server.

Please ignore this patchset until I've
gotten everything sorted out against
"what Windows does (tm)".

Sorry for the noise.

Jeremy.