Andrew Tridgell [Wed, 4 Jul 2001 07:15:53 +0000 (07:15 +0000)]
The big character set handling changeover!
This commit gets rid of all our old codepage handling and replaces it with
iconv. All internal strings in Samba are now in "unix" charset, which may
be multi-byte. See internals.doc and my posting to samba-technical for
a more complete explanation.
Andrew Bartlett [Mon, 2 Jul 2001 13:43:46 +0000 (13:43 +0000)]
This fix from Eelco Vriezekolk <eelco@nexus.com.na> is for a SIG11 bug where we
would always file_free(fsp) twice, once in close_file and once afterwoulds.
The bug was reported in SAMBA_2_2, but a code inspection shows it to be in HEAD
as well. (Unfortunetly I don't have the facilites to actualy check this, but
the change is quite simple, makes sence and compiles).
Richard Sharpe [Fri, 29 Jun 2001 23:30:01 +0000 (23:30 +0000)]
These changes cause the libsmbclient stuff to always build a non-shared library and to build a shared library only for the four OSes that we currently like.
I will probably add HP/UX and AIX support based on the CUPS stuff soon ...
Jeremy Allison [Fri, 29 Jun 2001 22:32:24 +0000 (22:32 +0000)]
Ensured all the system calls in msdfs.c go through the vfs layer.
Added vfs calls to symlink() and readlink() with appropriate configure
checks.
Jeremy.
Tim Potter [Fri, 29 Jun 2001 08:04:12 +0000 (08:04 +0000)]
Experimental support for joining a domain without creating an account in
server manager first. Just use the -U parameter to smbpasswd when joining
the domain:
Tim Potter [Fri, 29 Jun 2001 01:58:48 +0000 (01:58 +0000)]
Win9x weirdness. When setting the named pipe handle state on a Unicode
server, Win9x only sends \PIPE instead of \PIPE\. Looks suspiciously like
an off-by one bug in the Win9x dce/rpc pipe code.
Added info to some of the debug messages to get a better handle on a problem
people are reporting regarding multiple responses to queries on <1D> names.
There should only ever be one LMB but some users are seeing multiple replies
to queries for the LMB name. This is probably due to nodes on the LAN that
have NetBIOS over NetBEUI and/or IPX enabled. Previously, the debug message
did not include the IP address associated with the name. It *did* include
the source address of the packet, but in the examples I've seen all of these
were the same, eg:
[2000/06/22 11:58:25, 0] nmbd/nmbd_namequery.c:query_name_response(93)
query_name_response: Multiple (2) responses received for a query on subnet
129.130.10.136 for name NT.CIS.KSU.EDU<1d>. This response was from IP
129.130.10.24
[2000/06/22 11:58:25, 0] nmbd/nmbd_namequery.c:query_name_response(93)
query_name_response: Multiple (3) responses received for a query on subnet
129.130.10.136 for name NT.CIS.KSU.EDU<1d>. This response was from IP
129.130.10.24
[2000/06/22 11:58:25, 0] nmbd/nmbd_namequery.c:query_name_response(93)
query_name_response: Multiple (4) responses received for a query on subnet
129.130.10.136 for name NT.CIS.KSU.EDU<1d>. This response was from IP
129.130.10.24
[2000/06/22 11:58:25, 0] nmbd/nmbd_namequery.c:query_name_response(93)
query_name_response: Multiple (5) responses received for a query on subnet
129.130.10.136 for name NT.CIS.KSU.EDU<1d>. This response was from IP
129.130.10.24
Note that all of the above are reported as having come from 129.130.10.24.
This should never happen. If 129.130.10.24 is a WINS server it should
send a Negative Name Query Response for a <1D> name query (wierd but true).
So, are all of the above coming from different systems, all of which
think are the LMB? Are they all coming from one system that is, for some
strange reason, replying five times to the same query?
Anyway, I needed more info so I've changed the debug messages.
Andrew Tridgell [Thu, 28 Jun 2001 03:03:11 +0000 (03:03 +0000)]
temporarily made smbclient a non-error target so that builds will succeed when it fails. This will give richard a chance to fix problems without breaking the tree
Andrew Bartlett [Wed, 27 Jun 2001 13:58:58 +0000 (13:58 +0000)]
Jigger around with the tests a bit more
We now test security=share, security=user, security=server for both the
positive and negitive case (good/bad pw) and check that guest shares work for
share level security.
The server level security stuff seems to test positive without actualy
contating a server (another LIBSMB_PROG based smbd) - I will need to look into that...
Richard Sharpe [Wed, 27 Jun 2001 04:41:01 +0000 (04:41 +0000)]
Added back the automatic build of libsmbclient.so in head.
This is so I can find out what platforms it fails on. I will pull it again tomorrow if there are too many problems, like > 2 platforms that it fails to build on, but will pop it back in again as I resolve platforms.
Tim Potter [Tue, 26 Jun 2001 06:11:40 +0000 (06:11 +0000)]
Fixed bug introduced by changeover of security descriptor code from
malloc() to talloc(). Previously, creating an ACL containing zero ACEs
would return a non-NULL pointer to zero bytes of memory. The talloc() code
would return a NULL pointer making the ACL a NULL ACL instead of an empty
one. The difference is a NULL ACL allows all access and an empty ACL
denies all access.
We solve this by calling talloc(ctx, sizeof(SEC_ACE) * num_aces + 1).
Heh.
Tim Potter [Mon, 25 Jun 2001 06:13:27 +0000 (06:13 +0000)]
Cosmetic fixups found while playing with the server manager. Added support
for NET_SRV_SET_INFO rpc call which is made when double-clicking on a
computer in the server manager and changing the description. We always
return NT_STATUS_NOPROBLEMO as NT doesn't seem to decode any error messages
passed back.
Maybe the changed comment string could be stored in a tdb and regurgitated
instead of the "server string" smb.conf parameter?