Bug and Fix (Browse Lists)

[Matt Roberts, GRDA]

Developers,

First, please accept my thanks for a really great product.  I use Samba
extensively and really enjoy it.

I am testing the 3.0a21 for evaluation for later use in our shop and I have
found what appears to be a minor bug, and a fix that works for me.  I can't
find it documented anywhere, but here is what I have:

1.  One Samba PDC, 3.0a21, WINS support enabled.
2.  One Win2K Pro workstation SP3, configured as domain member.

Both machines are on the same Class-C subnet.  Everything works great, even
better than the 2.2 series PDC code, except for browse lists.  Attempting to
pull a list of machines in the domain, either through Network Neighborhood
gives me "The list of servers is unavailable..." and via 'smbclient -L PDC'
shows an empty browse list (not even the PDC).

I have setup both Samba and Windows according to all available documents on
the Samba site.  I am using an LDAP SAM backend, but that doesn't appear to
be related.


In the NMBd debug output (nmbd run with -d 255), I find this excerpt:

process_dgram: datagram from WS01<20> to DOMAIN<1e> IP 192.168.17.198 for
   \MAILSLOT\BROWSE of type 15 len=33
process_browse_packet: Discarding datagram from IP 192.168.17.198. Scope ()
   mismatch with our scope ((null)).
find_workgroup_on_subnet: workgroup search for DOMAIN on subnet
   192.168.17.254: found.
announce_myself_to_domain_master_browser: t (1040176764) - last(1040176605) <
   900
find_workgroup_on_subnet: workgroup search for DOMAIN on subnet
   UNICAST_SUBNET: found.
find_workgroup_on_subnet: workgroup search for DOMAIN on subnet
   UNICAST_SUBNET: found.


The part I am eyeballing is the part "Scope() mismatch with scope ((null))."

So out of curiosity, I added this line to my smb.conf file:

   netbios scope =

with no value (empty value), and everything now works like it did with 2.2!

So could there just be a null pointer somewhere that holds the NetBIOS scope
that is set to NULL initially for safety, and is just never initialized to an
empty but memory-allocated string later?

I haven't found a piece of suspect code yet, but I post this for you in case
somebody that knows the code better than I might be able to find it faster
than I can.

If this is not a bug, can somebody tell me why this is needed?  If that is
the case, I haven't found it in the doco yet.

Thanks,
Matt