IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
Tridge, if you have the time, you might want to look at a problem I'm having
with unix domain stream sockets. From a comment in this commit:
/* Using composite_trigger_error here causes problems with the client
* socket. Linux 2.6.8 gives me a ECONNRESET on the next read after
* writing the reply when I don't wait the 100 milliseconds. */
This is in winbind/wb_cmd_userdomgroups.c:93.
The problem I have is that I can not *immediately* send an error reply to the
client because the next receive fails. Waiting 100 milliseconds helps. It
might also be a problem with epoll(), I don't really know.
I'd appreciate if you took a brief look at this, maybe I'm doing something
wrong.
Thanks,
Volker
(This used to be commit 3e535cce74)
first. And if a request is being processed, queue it. This correctly survived
3 endless loops with wbinfo's doing different things while starting up smbd.
The number of indirections starts to become a bit scary, but what can you do
without a decent programming language that provides closures :-)
One thing that we might consider is to auto-generate async rpc requests that
return composite_context structs instead of rpc_requests. Otherwise I'd have
to write a lot of wrappers like composite_netr_LogonSamLogon_send.
The alternative would be to write two versions of wb_queue_domain_send which I
would like to avoid. This is cluttered enough already.
Volker
(This used to be commit 66c1b674f9)
so you can use them in search filters,
only for administration not used inside the winserver code
- fix the samba3 ugrade scripts to create a correct samba4 wins.ldb
metze
(This used to be commit 9f3b6746d8)
tested it, but I can not reproduce the problem I had with abartlett's initial
implementation anymore.
Fix a bug found using valgrind.
Volker
(This used to be commit 0c6c71ae3c)
* Add (limited) support for [string]
* Don't generate (and set) header elements for top level ref pointers as
they don't appear on the wire
(This used to be commit 765adaf19b)
fix the build for changes from SAMBA_4_0 branch
metze
r10541@SERNOX: metze | 2005-09-27 15:05:33 +0200
use a transaction when we allocate a new version
metze
r10549@SERNOX: metze | 2005-09-27 18:58:37 +0200
- add first start of wins pull replication
- we not yet apply records to our database but we fetch them correct form our partners
(we need conflict handling for this)
- we also need to filter out our own records!
metze
r10568@SERNOX: metze | 2005-09-28 11:33:04 +0200
move composite helpers to a seperate file, create a a seperate file for the conflict resolving logic
metze
r10571@SERNOX: metze | 2005-09-28 12:00:17 +0200
add forward declarations...to fix the build
metze
r10612@SERNOX: metze | 2005-09-29 16:11:06 +0200
we have a nbt_name now, and don't need to parse it
metze
r10614@SERNOX: metze | 2005-09-29 16:38:35 +0200
filter out our own records
metze
r10620@SERNOX: metze | 2005-09-29 18:07:08 +0200
- handle mutliple addresses in WREPL_REPL_SEND_REPLY
- make strings always valid talloc pointers
metze
r10621@SERNOX: metze | 2005-09-29 18:09:41 +0200
use debug level 2
metze
r10622@SERNOX: metze | 2005-09-29 18:48:05 +0200
- add one more debug message when we reply no record
- fix min max logic
metze
r10623@SERNOX: metze | 2005-09-29 20:49:06 +0200
build fixes...
metze
r10629@SERNOX: metze | 2005-09-30 00:11:41 +0200
- use seperate attributes for type, state, nodetype, is_static
... the winserver.c code needs some more updates to correctly,
create special group and multihomed registrations...
metze
r10640@SERNOX: metze | 2005-09-30 04:07:34 +0200
- add some short path for the composite helper functions
they will be used in the next commit
metze
r10642@SERNOX: metze | 2005-09-30 06:29:06 +0200
fix the build
metze
r10655@SERNOX: metze | 2005-09-30 17:36:49 +0200
- implement the WREPL_REPL_UPDATE* and WREPL_REPL_INFORM*
this includes the connection fliping into a client connection
for WREPL_REPL_UPDATE*
NOTE: I not yet found out how to get the w2k server to use INFORM against samba4
it uses inform against w2k and w2k3 but UPDATE against nt4 and samba4
what's left now is to be able to initiate INFORM and UPDATE requests to notify
our pull partners
metze
r10727@SERNOX: metze | 2005-10-05 14:11:05 +0200
fix the build
metze
r10770@SERNOX: metze | 2005-10-06 16:56:01 +0200
- move the table filling to a seperate function, will be reused later
- fix the build, wrepl_nbt_name fixes
- remove state -> update_state
metze
r10771@SERNOX: metze | 2005-10-06 17:04:48 +0200
add a function to create a wreplsrv_in_connection from a client connection
metze
r10772@SERNOX: metze | 2005-10-06 17:13:51 +0200
- make the connection code more generic to handle the pull cached connection,
push cached connection or given connections
- when we don't use a cached connection, disconnection when a pull_cycle is done
- fix the build and use the configured source ip
metze
r10773@SERNOX: metze | 2005-10-06 17:18:49 +0200
- add composite functions for push notification
metze
r10774@SERNOX: metze | 2005-10-06 17:23:46 +0200
- use periodic push notifycation, this is just for now
as it needs to be configurable and and be triggered when the local database
has changes since the last notify
- I also need to work out how to decide if the partner supports
persistent connections and WREPL_REPL_INFORM* messages
metze
r10923@SERNOX: metze | 2005-10-12 16:52:34 +0200
fix the build becuse of conflicts with main SAMBA_4_0 tree
metze
(This used to be commit 6d97dd6e50)
we should only use 255.255.255.255 for group records,
only when we reply, not when the name is registered.
the real address is used for replication
metze
(This used to be commit 45a774c63c)
- as the old records are broken sinse the last winsdb_dn() changes, (the dn components order was reversed)
we can use nicer attribute and objectClass names...
- use much more verbose error handling for winsdb_*
- print a debug message when we found a corrupted record
metze
(This used to be commit 82bad3f3ef)
create winsdb_record() and winsdb_message() as public functions
so that they can be used in the wrepl_server/
metze
(This used to be commit b8b48c8aa5)
- load our wins partners at start time:
# this is a sample partner record:
dn: name=SERNOX4-1,CN=PARTNERS
objectClass: wreplPartner
name: SERNOX4-9
address: 172.31.1.1
ourAddress: 172.31.9.1
type: 0x3
pullInterval: 3600
- go through all winsdb records in the database and create the wins_owner table,
but don't add ourself to it as out nbt task will update the db too, we refetch
the local max_versiion, each time we need it, (that typicaly onces per replication cycle)
metze
(This used to be commit 4490a2864e)
- store the wins owner on the record and the wins owner and expire time on each address
- we use "0.0.0.0" to mark entries which are registered at the local wins server
- we use this ldif-format:
address: 172.31.9.1;winsOwner:0.0.0.0;expireTime:20050923032337.0Z
address: 172.31.1.1;winsOwner:172.31.9.202;expireTime:20050923032330.0Z
metze
(This used to be commit 752025a0e1)
move to struct winsdb_addr, (I'll add expiry_time and the wins_owner later
tridge: can you please review the new winsdb_addr_* functions carefull, look for off-by-one bugs, etc.
metze
(This used to be commit 72eba90465)
