[PATCH] Fix bug 13465

[Christof Schmitt]

On Thu, Jul 12, 2018 at 10:10:36AM +0200, Ralph Böhme wrote:
> On Wed, Jul 11, 2018 at 02:52:14PM -0700, Christof Schmitt wrote:
> >It also works on my cluster.
> 
> Thanks for testing! Which combinations did you test? As I've learnt in this

I tested the basic commands as root. You are correct, we need more
extensive testing. This also goes back to the old discussion that we
need a clustered Samba environment in autobuild...

> excersize, we should test all possible combinations, as most things are not
> covered by autobuild:
> 
> Cluster (with include=registry):
> - all commands as root

These commands still failed for me as root:

eventlogadm -o write test
PANIC: messaging not initialized

ntlm_auth --diagnostics
PANIC: messaging not initialized

smbcontrol winbindd debuglevel
PANIC: messaging not initialized

> - all commands as user

Most fail when run as a user:

net
PANIC (pid 13441): messaging not initialized

pdbedit -L
PANIC (pid 14922): messaging not initialized

rpcclient 127.0.0.1
Unable to initialize messaging context
PANIC (pid 16959): messaging not initialized

sharesec
PANIC (pid 17320): messaging not initialized

smbcacls
Unable to initialize messaging context
PANIC (pid 17662): messaging not initialized

smbclient
Unable to initialize messaging context
PANIC (pid 18236): messaging not initialized

smbcontrol smbd debuglevel
PANIC: messaging not initialized

smbcquotas
Unable to initialize messaging context
PANIC (pid 18830): messaging not initialized

smbget
PANIC (pid 19101): messaging not initialized

smbpasswd
PANIC (pid 19461): messaging not initialized

smbtree
Unable to initialize messaging context
PANIC (pid 19900): messaging not initialized

testparm
PANIC (pid 20452): messaging not initialized

> - all commands without smb.conf
> - all commands without install prefix dir run from build dir

I have not done the other tests, probably we should fix the above cases
first. I can take a look tomorrow.

> The same on a non-clustered Samba. I guess we need a full matrix 4.7 vs
> master+patchset as some combinations are likely broken in 4.7, like smbclient.
> 
> >One question: Why did you revert the smbclient patch? In the end it does not
> >make a difference, as the popt credentials callback already initializes
> >messaging, but it would be consistent with the other tools to keep it.
> 
> The idea was to only do the manual call where it's really needed. I'm currently
> running CI where I've also reverted the last three patches for smbtree,
> smbcquotas and smbcacls.

Understood. The slightly confusing part could be that POPT_CREDENTIALS
then internally initializes messaging, but not the other POPT
definitions. On the other hand, only POPT_CREDENTIALS required
messaging...

Christof