[SCM] Samba Shared Repository - branch master updated

[Andrew Bartlett]

On Tue, 2017-03-28 at 09:11 -0700, Jeremy Allison via samba-technical
wrote:
> On Tue, Mar 28, 2017 at 01:20:03PM +0200, Andrew Bartlett wrote:
> 
> > commit a47a8e41bd3acc20d40ba78449d89775bcdd73ed
> > Author: Andrew Bartlett <abartlet at samba.org>
> > Date:   Fri Mar 24 13:07:06 2017 +1300
> > 
> >     samba-tool: Ensure that samba-tool processes --name=not-
> > existing does not error
> >     
> >     Signed-off-by: Andrew Bartlett <abartlet at samba.org>
> >     Reviewed-by: Garming Sam <garming at catalyst.net.nz>
> >     BUG: https://bugzilla.samba.org/show_bug.cgi?id=12705
> 
> I wish you'd discussed this one with me first before pushing it.
> 
> This is only masking a real bug I think. Without it, We can see
> that a bin/samba process is removing the "samba" name without
> ever having registered it. And we don't know why.

Looking over the code, last night for related reasons (trying to make
ldap multi-process again, so moving message_dgm cleanup around) I think
a talloc destructor is removing the name from the names database in a
child process after a fork().

We have had, in the past, an issue where a talloc destructor (on the
autofree context in that case) removed the messaging socket in the s4
messaging code.  That was equally fun to find :-)

> Now I've got the symlink regressions fixed I'll be looking more
> carefully into this, but first I'll have to revert this for the
> tests.

I agree we need a test that asserts all the processes that we expect to
be present are alive and responding to a ping.  I actually think it
should run before every test, instead of the current pid check.

I'm sorry that we got out of sync on this one.

Andrew Bartlett

-- 
Andrew Bartlett                       http://samba.org/~abartlet/
Authentication Developer, Samba Team  http://samba.org
Samba Developer, Catalyst IT          http://catalyst.net.nz/services/samba