14 years agor10421: following on discussions with simo, I have worked out a way of
Andrew Tridgell [Thu, 22 Sep 2005 13:12:46 +0000 (13:12 +0000)]
r10421: following on discussions with simo, I have worked out a way of
allowing searches to proceed while another process is in a
transaction, then only upgrading the transaction lock to a write lock
on commit.

The solution is:

 - split tdb_traverse() into two calls, called tdb_traverse() and
   tdb_traverse_read(). The _read() version only gets read locks, and
   will fail any write operations made in the callback from the

 - the normal tdb_traverse() call allows for read or write operations
   in the callback, but gets the transaction lock, preventing
   transastions from starting inside the traverse

In addition we enforce the following rule that you may not start a
transaction within a traverse callback, although you can start a
traverse within a transaction

With these rules in place I believe all the deadlock possibilities are
removed, and we can now allow for searches to happen in parallel with

14 years agor10420: Two minor scons fixes
Jelmer Vernooij [Thu, 22 Sep 2005 12:23:44 +0000 (12:23 +0000)]
r10420: Two minor scons fixes

14 years agor10419: Remove unused prototypes of locking functions (thanks Jelmer)
Simo Sorce [Thu, 22 Sep 2005 09:06:37 +0000 (09:06 +0000)]
r10419: Remove unused prototypes of locking functions (thanks Jelmer)
omment about transactions

14 years agor10418: Fix comparison_fn_t test I busted. Add $HOME to the environment as some
Tim Potter [Thu, 22 Sep 2005 08:47:32 +0000 (08:47 +0000)]
r10418: Fix comparison_fn_t test I busted.  Add $HOME to the environment as some
tools get confused if it isn't there.

14 years agor10416: Detect some more types to get rid of spurious warnings in the
Tim Potter [Thu, 22 Sep 2005 07:17:24 +0000 (07:17 +0000)]
r10416: Detect some more types to get rid of spurious warnings in the
lib/replace/replace.h header.

Running 'scons lib/tdb' now (mostly) builds tdb.

14 years agor10415: The ldb and tdb libraries are bad examples to test out the make proto
Tim Potter [Thu, 22 Sep 2005 06:38:26 +0000 (06:38 +0000)]
r10415: The ldb and tdb libraries are bad examples to test out the make proto
code as they are marked as NOPROTO in the files.

14 years agor10414: Fix building of proto.h
Tim Potter [Thu, 22 Sep 2005 06:20:53 +0000 (06:20 +0000)]
r10414: Fix building of proto.h

14 years agor10411: we don't need the 10 times retry on rid allocation now, as
Andrew Tridgell [Thu, 22 Sep 2005 05:31:16 +0000 (05:31 +0000)]
r10411: we don't need the 10 times retry on rid allocation now, as
transactions ensure two account creations can't interfere with each

14 years agor10410: blindly update the scons file for tdb. I'm not sure how this works,
Andrew Tridgell [Thu, 22 Sep 2005 05:03:03 +0000 (05:03 +0000)]
r10410: blindly update the scons file for tdb. I'm not sure how this works,
but it definately needs transaction.c now

14 years agor10409: allow smb.conf override of ldb synchronous transactions with "ldb:nosync...
Andrew Tridgell [Thu, 22 Sep 2005 04:57:35 +0000 (04:57 +0000)]
r10409: allow smb.conf override of ldb synchronous transactions with "ldb:nosync = yes/no"

14 years agor10408: now that we are using tdb transactions we don't need any additional
Andrew Tridgell [Thu, 22 Sep 2005 04:40:23 +0000 (04:40 +0000)]
r10408: now that we are using tdb transactions we don't need any additional
locking code in the ldb_tdb backend, except for a single read lock
during searches to ensure searches don't cross transaction boundaries

The tdb transactions code would map these extra locks to noops anyway
(as locking makes no sense inside a transaction), but the work in
setting up the locking keys still costs something, and it makes the
code needlessly complex

14 years agor10407: the schannel database does not need to be synchronous (and thus crash
Andrew Tridgell [Thu, 22 Sep 2005 04:19:57 +0000 (04:19 +0000)]
r10407: the schannel database does not need to be synchronous (and thus crash
safe) as it is removed on smbd restart

14 years agor10406: added --nosync option to all ldb tools, so that you can control if
Andrew Tridgell [Thu, 22 Sep 2005 04:16:46 +0000 (04:16 +0000)]
r10406: added --nosync option to all ldb tools, so that you can control if
transactions are synchronous or not on the command line.

add LDB_FLG_NOSYNC flag to ldb_connect() so we can make our temporary
ldb databases non-synchronous

14 years agor10405: added transactions into tdb, and hook them into ldb. See my
Andrew Tridgell [Thu, 22 Sep 2005 03:56:41 +0000 (03:56 +0000)]
r10405: added transactions into tdb, and hook them into ldb. See my
samba-technical posting for more details on the transactions design.

This also adds a number of command line arguments to tdbtorture,
making it more flexible, and fixes some lock deadlock conditions in
the tdbtorture code.

14 years agor10404: make sure we use the right smb.conf in the ldap testing
Andrew Tridgell [Thu, 22 Sep 2005 03:52:32 +0000 (03:52 +0000)]
r10404: make sure we use the right smb.conf in the ldap testing

14 years agor10403: fixed the basedn for testing, and add a debug showing the size of the test...
Andrew Tridgell [Thu, 22 Sep 2005 03:51:50 +0000 (03:51 +0000)]
r10403: fixed the basedn for testing, and add a debug showing the size of the test in ldbtest

14 years agor10402: Make the RPC-SAMLOGON test pass against Win2k3 SP0 again.
Andrew Bartlett [Thu, 22 Sep 2005 01:50:58 +0000 (01:50 +0000)]
r10402: Make the RPC-SAMLOGON test pass against Win2k3 SP0 again.

I still have issues with Win2k3 SP1, and Samba4 doesn't pass it's own
test for the moment, but I'm working on these issues :-)

This required a change to the credentials API, so that the special
case for NTLM logins using a principal was indeed handled as a
special, not general case.

Also don't set the realm from a ccache, as then it overrides --option=realm=.

Andrew Bartlett

14 years agor10398: Don't do DNS lookups on short names (no .).
Andrew Bartlett [Wed, 21 Sep 2005 22:55:39 +0000 (22:55 +0000)]
r10398: Don't do DNS lookups on short names (no .).

Andrew Bartlett

14 years agor10395: While playing with winbind, I found a segfault because I had given
Volker Lendecke [Wed, 21 Sep 2005 16:35:15 +0000 (16:35 +0000)]
r10395: While playing with winbind, I found a segfault because I had given
resolve_name_send a stack-allocated method list. Duplicate it.


14 years agor10393: More type checking in scons
Jelmer Vernooij [Wed, 21 Sep 2005 15:18:11 +0000 (15:18 +0000)]
r10393: More type checking in scons
Remove unused file

14 years agor10390: Remove fstring, pstring and uint32_t from winbindd_nss.h.
Volker Lendecke [Wed, 21 Sep 2005 14:41:29 +0000 (14:41 +0000)]
r10390: Remove fstring, pstring and uint32_t from winbindd_nss.h.


14 years agor10389: Add Manifest file (required for CPAN)
Jelmer Vernooij [Wed, 21 Sep 2005 13:04:09 +0000 (13:04 +0000)]
r10389: Add Manifest file (required for CPAN)

14 years agor10388: Add version numbers (required for CPAN)
Jelmer Vernooij [Wed, 21 Sep 2005 12:57:18 +0000 (12:57 +0000)]
r10388: Add version numbers (required for CPAN)

14 years agor10387: By exporting KRB5_CONFIG pointing at a file of our choosing, we can
Andrew Bartlett [Wed, 21 Sep 2005 12:38:25 +0000 (12:38 +0000)]
r10387: By exporting KRB5_CONFIG pointing at a file of our choosing, we can
avoid DNS delays in our testsuite.  The next step will be to use
kerberos in the tests (awaits application of socketwrapper or hooks
into our socket layer).

Andrew Bartlett

14 years agor10386: Merge current lorikeet-heimdal into Samba4.
Andrew Bartlett [Wed, 21 Sep 2005 12:24:41 +0000 (12:24 +0000)]
r10386: Merge current lorikeet-heimdal into Samba4.

Andrew Bartlett

14 years agor10385: removed obsolete comment
Andrew Tridgell [Wed, 21 Sep 2005 10:46:02 +0000 (10:46 +0000)]
r10385: removed obsolete comment

14 years agor10384: add _GNU_SOURCE in tdb configure
Andrew Tridgell [Wed, 21 Sep 2005 10:45:47 +0000 (10:45 +0000)]
r10384: add _GNU_SOURCE in tdb configure

14 years agor10383: This patch is on the road to implementing servers (such as kpasswd) that
Andrew Bartlett [Wed, 21 Sep 2005 10:18:40 +0000 (10:18 +0000)]
r10383: This patch is on the road to implementing servers (such as kpasswd) that
use raw krb5, not GSSAPI.  I still keep the 'fake GSSAPI' code, but
under the module name 'fake_gssapi_krb5'.

Andrew Bartlett

14 years agor10382: In the absence of client support for the full KDC-side
Andrew Bartlett [Wed, 21 Sep 2005 10:17:56 +0000 (10:17 +0000)]
r10382: In the absence of client support for the full KDC-side
canonicalisation code, I've hacked Heimdal to use the default realm if
no other realm can be determined for a given host.

Andrew Bartlett

14 years agor10381: Be a bit stricter on pod syntax
Jelmer Vernooij [Wed, 21 Sep 2005 09:27:17 +0000 (09:27 +0000)]
r10381: Be a bit stricter on pod syntax

14 years agor10380: Use pod-style documentation rather then XML-doc, in good perl style.
Jelmer Vernooij [Wed, 21 Sep 2005 09:16:55 +0000 (09:16 +0000)]
r10380: Use pod-style documentation rather then XML-doc, in good perl style.

14 years agor10379: Add files for ldb and tdb to proto_files. The tool for building proto.h
Tim Potter [Wed, 21 Sep 2005 07:20:58 +0000 (07:20 +0000)]
r10379: Add files for ldb and tdb to proto_files.  The tool for building proto.h
is busted though.

14 years agor10378: Build config.h file from dictionary of defines and always use it.
Tim Potter [Wed, 21 Sep 2005 07:12:23 +0000 (07:12 +0000)]
r10378: Build config.h file from dictionary of defines and always use it.

14 years agor10377: Save configuration stuff to sconf.cache so it isn't annoyingly run
Tim Potter [Wed, 21 Sep 2005 05:39:18 +0000 (05:39 +0000)]
r10377: Save configuration stuff to sconf.cache so it isn't annoyingly run
at every single build.  Run 'scons configure=1' or delete sconf.cache
to force checks to be re-run.

Jelmer, I think this stuff is cached in the .sconf_cache directory but
the message is still displayed and it looks like it caches the compiled
test object file not the actual result of the test.

14 years agor10376: Argh - not quoting dictionary/hash keys is a bit of a perlism.
Tim Potter [Wed, 21 Sep 2005 04:59:02 +0000 (04:59 +0000)]
r10376: Argh - not quoting dictionary/hash keys is a bit of a perlism.

14 years agor10374: Add HAVE_* defines (on command-line or in config.h file) for scons +
Jelmer Vernooij [Wed, 21 Sep 2005 00:38:23 +0000 (00:38 +0000)]
r10374: Add HAVE_* defines (on command-line or in config.h file) for scons +
some other minor updates

14 years agor10373: Fix segfault in LookupSids.
Andrew Bartlett [Wed, 21 Sep 2005 00:27:10 +0000 (00:27 +0000)]
r10373: Fix segfault in LookupSids.

Andrew Bartlett

14 years agor10372: Having gone to all the effort to uppercase the realm, actually set the
Andrew Bartlett [Wed, 21 Sep 2005 00:15:56 +0000 (00:15 +0000)]
r10372: Having gone to all the effort to uppercase the realm, actually set the
upper-case realm.

Andrew Bartlett

14 years agor10370: only validate the re-generated binding string for hostnames with IPs
Andrew Tridgell [Tue, 20 Sep 2005 23:23:03 +0000 (23:23 +0000)]
r10370: only validate the re-generated binding string for hostnames with IPs

Jelmer, can you see a better approach to this? As far as I am aware
protocol towers don't use hostnames, they always use IP addresses

14 years agor10369: You don't need to put leading or trailing spaces on variables. It's
Tim Potter [Tue, 20 Sep 2005 23:00:45 +0000 (23:00 +0000)]
r10369: You don't need to put leading or trailing spaces on variables.  It's
a make-ism.

Use consistent (single) quoting.

14 years agor10368: when building the epm tower, don't put host names in the ip address
Andrew Tridgell [Tue, 20 Sep 2005 22:52:54 +0000 (22:52 +0000)]
r10368: when building the epm tower, don't put host names in the ip address
field, instead put a zero address. Note that zero is correct (ie. we
shouldn't do the lookup) as in the client we want to send a zero for
the server to fill in. When we make this call from the server we fill
in a real IP.

14 years agor10366: More scons fixes. Building et, asn1, lex and yacc files sort-of works now
Jelmer Vernooij [Tue, 20 Sep 2005 22:10:40 +0000 (22:10 +0000)]
r10366: More scons fixes. Building et, asn1, lex and yacc files sort-of works now

14 years agor10365: Use nsswitch/winbindd_nss.h in winbind/, update that file to the current 3_0
Volker Lendecke [Tue, 20 Sep 2005 21:37:10 +0000 (21:37 +0000)]
r10365: Use nsswitch/winbindd_nss.h in winbind/, update that file to the current 3_0


14 years agor10364: Turn gensec:gssapi on by default, except for a login of the form
Andrew Bartlett [Tue, 20 Sep 2005 21:29:29 +0000 (21:29 +0000)]
r10364: Turn gensec:gssapi on by default, except for a login of the form

This will probably break in a few configurations, so please let me
know.  I'll also work to have a way to inhibit kerberos/ntlmssp, as
this removes -k.

Andrew Bartlett

14 years agor10363: Nobody loudly screamed "noo", so commit the samba3 winbind interface to
Volker Lendecke [Tue, 20 Sep 2005 20:54:25 +0000 (20:54 +0000)]
r10363: Nobody loudly screamed "noo", so commit the samba3 winbind interface to
samba4. Ok, maybe the silence is due to timezones, but what can you do... ;-)


14 years agor10356: Make the proto generator work with scons
Jelmer Vernooij [Tue, 20 Sep 2005 17:49:19 +0000 (17:49 +0000)]
r10356: Make the proto generator work with scons

14 years agor10353: Fix typo
Volker Lendecke [Tue, 20 Sep 2005 15:43:58 +0000 (15:43 +0000)]
r10353: Fix typo

14 years agor10348: Add scons scripts for remaining subsystems. Most subsystems build now,
Jelmer Vernooij [Tue, 20 Sep 2005 11:59:03 +0000 (11:59 +0000)]
r10348: Add scons scripts for remaining subsystems. Most subsystems build now,
but final linking still fails (as does generating files asn1, et, idl and proto

14 years agor10346: here are only real ip-addresses valid, prevent dns lookups,
Stefan Metzmacher [Tue, 20 Sep 2005 11:39:40 +0000 (11:39 +0000)]
r10346: here are only real ip-addresses valid, prevent dns lookups,
when the input is invalid


14 years agor10345: Add more add-hock tests.
Andrew Bartlett [Tue, 20 Sep 2005 09:57:30 +0000 (09:57 +0000)]
r10345: Add more add-hock tests.

Andrew Bartlett

14 years agor10341: remove unused libads/ code, we'll never use this in samba4,
Stefan Metzmacher [Tue, 20 Sep 2005 08:30:30 +0000 (08:30 +0000)]
r10341: remove unused libads/ code, we'll never use this in samba4,
and have replacements for the most stuff already in the tree

discussed with abartlet


14 years agor10339: fix ndr_push_udlongr
Stefan Metzmacher [Tue, 20 Sep 2005 07:56:54 +0000 (07:56 +0000)]
r10339: fix ndr_push_udlongr


14 years agor10337: This grubby little hack is the implementation of a concept discussed
Andrew Bartlett [Tue, 20 Sep 2005 07:03:47 +0000 (07:03 +0000)]
r10337: This grubby little hack is the implementation of a concept discussed
on the kerberos mailing lists a couple of weeks ago: Don't use DNS at
all for expanding short names into long names.

Using the 'override krb5_init_context' code already in the tree, this
removes the DNS lag on a kerberos session setup/connection.

Andrew Bartlett

14 years agor10336: Add sconscript for a couple more subsystems.
Jelmer Vernooij [Tue, 20 Sep 2005 00:39:19 +0000 (00:39 +0000)]
r10336: Add sconscript for a couple more subsystems.

14 years agor10335: Build tdb tools into bin directory.
Tim Potter [Mon, 19 Sep 2005 23:39:23 +0000 (23:39 +0000)]
r10335: Build tdb tools into bin directory.

14 years agor10334: Use CPPDEFINES instead of CCFLAGS for adding preprocessor symbols.
Tim Potter [Mon, 19 Sep 2005 23:37:01 +0000 (23:37 +0000)]
r10334: Use CPPDEFINES instead of CCFLAGS for adding preprocessor symbols.

14 years agor10332: Fix the build - messaging uses UNIX_PRIVS
Jelmer Vernooij [Mon, 19 Sep 2005 23:10:28 +0000 (23:10 +0000)]
r10332: Fix the build - messaging uses UNIX_PRIVS

14 years agor10330: Add SConscript to more subsystems. Some of the tdb tools build now.
Jelmer Vernooij [Mon, 19 Sep 2005 22:01:57 +0000 (22:01 +0000)]
r10330: Add SConscript to more subsystems. Some of the tdb tools build now.
Start on custom Samba scons tools (for handling proto generation, pidl, etc)

14 years agor10329: Perhaps we shouldn't get too ahead of ourselves just yet.
Tim Potter [Mon, 19 Sep 2005 21:50:53 +0000 (21:50 +0000)]
r10329: Perhaps we shouldn't get too ahead of ourselves just yet.

14 years agor10328: Add more emacs python-mode markers.
Tim Potter [Mon, 19 Sep 2005 21:44:36 +0000 (21:44 +0000)]
r10328: Add more emacs python-mode markers.

14 years agor10327: Use CPPPATH instead of appending -Idir to CCFLAGS.
Tim Potter [Mon, 19 Sep 2005 21:37:38 +0000 (21:37 +0000)]
r10327: Use CPPPATH instead of appending -Idir to CCFLAGS.

Use a list comprehension instead of a for loop.

Add a emacs thingy to kick in python mode.  (-:

14 years agor10323: Add first bits required for getting compile with scons working. This does
Jelmer Vernooij [Mon, 19 Sep 2005 19:19:10 +0000 (19:19 +0000)]
r10323: Add first bits required for getting compile with scons working. This does
not work yet and can exist parallel with the existing build system.

14 years agor10316: More dynconfig fixes
Jelmer Vernooij [Mon, 19 Sep 2005 13:26:07 +0000 (13:26 +0000)]
r10316: More dynconfig fixes

14 years agor10315: Remove use of fstring and pstring in dynconfig.c
Jelmer Vernooij [Mon, 19 Sep 2005 12:58:43 +0000 (12:58 +0000)]
r10315: Remove use of fstring and pstring in dynconfig.c
Remove unused includes of dynconfig.h

14 years agor10314: Apply the controvertial 'server role =' patch after discussion on the list:
Andrew Bartlett [Mon, 19 Sep 2005 11:55:34 +0000 (11:55 +0000)]
r10314: Apply the controvertial 'server role =' patch after discussion on the list:

This patch removes the 'domain logon' and 'domain master' controls from
Samba4, in favour of a 'server role =' that users can actually

We can expand the list of roles as needed, and nobody has to figure out
what a 'domain master' actually means.

Andrew Bartlett

14 years agor10312: fix compiler warning
Stefan Metzmacher [Mon, 19 Sep 2005 09:57:39 +0000 (09:57 +0000)]
r10312: fix compiler warning


14 years agor10306: change these modules to use new error API
Simo Sorce [Sun, 18 Sep 2005 18:50:02 +0000 (18:50 +0000)]
r10306: change these modules to use new error API

14 years agor10305: start implementing better error handling
Simo Sorce [Sun, 18 Sep 2005 18:49:06 +0000 (18:49 +0000)]
r10305: start implementing better error handling

changed the prioivate modules API
error string are now not spread over all
modules but are kept in  a single place.

This allows a better control of memory
and error reporting.

14 years agor10304: check for basic ldb_message sanity and return appropriate
Simo Sorce [Sun, 18 Sep 2005 10:47:03 +0000 (10:47 +0000)]
r10304: check for basic ldb_message sanity and return appropriate
LDB_ERR_ value

14 years agor10303: check no attribute is given empty
Simo Sorce [Sun, 18 Sep 2005 10:46:21 +0000 (10:46 +0000)]
r10303: check no attribute is given empty

14 years agor10302: Introduce ldap like error codes
Simo Sorce [Sun, 18 Sep 2005 10:45:28 +0000 (10:45 +0000)]
r10302: Introduce ldap like error codes

14 years agor10301: fix standalone compiple after tdb changes
Simo Sorce [Sun, 18 Sep 2005 10:45:03 +0000 (10:45 +0000)]
r10301: fix standalone compiple after tdb changes

14 years agor10300: forgot to change the dsdb modules function names
Simo Sorce [Sat, 17 Sep 2005 19:29:45 +0000 (19:29 +0000)]
r10300: forgot to change the dsdb modules function names

14 years agor10299: remove the public (un)lock functions and introduce a transaction based
Simo Sorce [Sat, 17 Sep 2005 19:25:50 +0000 (19:25 +0000)]
r10299: remove the public (un)lock functions and introduce a transaction based
private ldb API

ldb_sqlite3 is already working with this model and ldb_tdb will do
as soon as tridge finishes the tdb transaction code.

currently the transactions are always implicit and wrap any single
ldb API call except searching, the transaction functions are
currently not made public on purpose.


14 years agor10296: Fix function pointer handling for older perl versions
Jelmer Vernooij [Sat, 17 Sep 2005 14:36:35 +0000 (14:36 +0000)]
r10296: Fix function pointer handling for older perl versions

14 years agor10295: Remove dependency on Data::Dumper
Jelmer Vernooij [Sat, 17 Sep 2005 14:13:36 +0000 (14:13 +0000)]
r10295: Remove dependency on Data::Dumper

14 years agor10294: Generate Makefile directly rather then thru Autoconf
Jelmer Vernooij [Sat, 17 Sep 2005 13:08:49 +0000 (13:08 +0000)]
r10294: Generate Makefile directly rather then thru Autoconf
substitution variables are now no longer used.

This is one more step towards a (hopefully) perl-based configure

14 years agor10292: This is set below from lp_server_role().
Andrew Bartlett [Sat, 17 Sep 2005 11:06:14 +0000 (11:06 +0000)]
r10292: This is set below from lp_server_role().

Andrew Bartlett

14 years agor10291: The patch optionally (off by default, not available in all cases) allows
Andrew Bartlett [Sat, 17 Sep 2005 09:46:20 +0000 (09:46 +0000)]
r10291: The patch optionally (off by default, not available in all cases) allows
Samba to use the target principal name supplied in the mechTokenMIC of
an SPNEGO negTokenInit.

This isn't a great idea for security reasons, but is how Samba3 behaves,
and allows kerberos to function more often in some environments.  It is
only available for CIFS session setups, due to the ordering of the

Andrew Bartlett

14 years agor10287: Compile compilers for build host. This fixes some bits of
Jelmer Vernooij [Sat, 17 Sep 2005 02:10:15 +0000 (02:10 +0000)]
r10287: Compile compilers for build host. This fixes some bits of
the mingw32 build (it now fails on missing inet_aton / in_addr definitions).
Find sane default for HOSTCC

14 years agor10286: This patch is ugly and disgusting, but for now it works better than the other
Andrew Bartlett [Sat, 17 Sep 2005 01:11:50 +0000 (01:11 +0000)]
r10286: This patch is ugly and disgusting, but for now it works better than the other
ideas I have had.

When I get a full list of things I want to do to a krb5_context I'll
either add gsskrb5_ wrappers, or a way of speicfying the krb5 context
per gssapi context.

(I want to ensure that the only krb5_context variables created while
executing Samba4 are via our wrapper).

Andrew Bartlett

14 years agor10283: Eliminate some more use of autoconf substitution variables.
Jelmer Vernooij [Sat, 17 Sep 2005 00:42:05 +0000 (00:42 +0000)]
r10283: Eliminate some more use of autoconf substitution variables.
Add makefile rule for '.ho' files (compiled with host compiler). This
does not allow for cross-compiling yet as that requires a HOSTLD as well.

14 years agor10277: do not ovverride LIKE, thanks to derrel I found out how to do
Simo Sorce [Fri, 16 Sep 2005 20:54:57 +0000 (20:54 +0000)]
r10277: do not ovverride LIKE, thanks to derrel I found out how to do
the same thing with a harmless user function

14 years agor10258: Fix an unused/duplicate local variable.
Tim Potter [Fri, 16 Sep 2005 07:24:36 +0000 (07:24 +0000)]
r10258: Fix an unused/duplicate local variable.

14 years agor10257: strlen returns a size_t which can be 64 bits long.
Tim Potter [Fri, 16 Sep 2005 07:19:37 +0000 (07:19 +0000)]
r10257: strlen returns a size_t which can be 64 bits long.

14 years agor10256: Fix some unhandled enumeration warnings. There's one still left,
Tim Potter [Fri, 16 Sep 2005 07:15:40 +0000 (07:15 +0000)]
r10256: Fix some unhandled enumeration warnings.  There's one still left,
RAW_SEARCH_UNIX_INFO find_fill_info(), which I think is a bug.

14 years agor10255: Fix some more 64-bit warnings.
Tim Potter [Fri, 16 Sep 2005 07:10:11 +0000 (07:10 +0000)]
r10255: Fix some more 64-bit warnings.

14 years agor10253: a fairly large tdb cleanup and re-organise. Nearly all of this change
Andrew Tridgell [Fri, 16 Sep 2005 03:52:42 +0000 (03:52 +0000)]
r10253: a fairly large tdb cleanup and re-organise. Nearly all of this change
just involves splitting up the core tdb.c code into separate files on
logical boundaries, but there are some minor functional changes as well:

 - move the 'struct tdb_context' into tdb_private.h, hiding it from
   users. This was done to allow the structure to change without
   breaking code that uses tdb.

 - added accessor functions tdb_fd(), tdb_name(), and tdb_log_fn() to
   access the elements of struct tdb_context that were used by
   external code but are no longer visible

 - simplied tdb_append() to use tdb_fetch()/tdb_store(), which is just
   as good due to the way tdb locks work

 - changed some of the types (such as tdb_off to tdb_off_t) to make
   syntax highlighting work better

 - removed the old optional spinlock code. It was a bad idea.

 - fixed a bug in tdb_reopen_all() that caused tdbtorture to sometimes
   fail or report nasty looking errors. This is the only real bug
   fixed in this commit. Jeremy/Jerry, you might like to pickup this
   change for Samba3, as that could definately affect smbd in

The aim of all of these changes is to make the tdb
transactions/journaling code I am working on easier to write. I
started to write it on top of the existing tdb.c code and it got very
messy. Splitting up the code makes it much easier to follow.

There are more cleanups we could do in tdb, such as using uint32_t
instead of u32 (suggested by metze). I'll leave those for another day.

14 years agor10252: a recent checkin from simo changed the handling of BASE and SUBTREE
Andrew Tridgell [Fri, 16 Sep 2005 03:18:49 +0000 (03:18 +0000)]
r10252: a recent checkin from simo changed the handling of BASE and SUBTREE
searches in ldb to be more ldap compliant, but broke the wins server
and the ejs ldb code. This fixes those up so 'make test' passes again.

14 years agor10251: some more work on ldb_sqlite3
Simo Sorce [Thu, 15 Sep 2005 23:10:07 +0000 (23:10 +0000)]
r10251: some more work on ldb_sqlite3

I must say that writing a new module is a very good way
to find lot of subtle bugs laying in the code

We need more tests!

commit oLschema2ldif.c to keep it safe from data losses (rm -fr :-)

update test generic to reflect the fix made on comparsion functions

14 years agor10250: the comparison is caseless so we must caseless subtract
Simo Sorce [Thu, 15 Sep 2005 23:06:57 +0000 (23:06 +0000)]
r10250: the comparison is caseless so we must caseless subtract
otherwise we get the wrong result when comparing upper
case chars with lower case chars

14 years agor10246: Remove unused function
Jelmer Vernooij [Thu, 15 Sep 2005 20:03:35 +0000 (20:03 +0000)]
r10246: Remove unused function
Move auth-specific file to auth/

14 years agor10245: Get rid of XFILE in a few places.
Jelmer Vernooij [Thu, 15 Sep 2005 19:52:13 +0000 (19:52 +0000)]
r10245: Get rid of XFILE in a few places.
Add fdprintf() and vfdprintf() helper functions.

14 years agor10238: Add a entry for sec_info to treat it as a uint32 for now.
Tim Potter [Thu, 15 Sep 2005 10:20:08 +0000 (10:20 +0000)]
r10238: Add a entry for sec_info to treat it as a uint32 for now.

14 years agor10237: fix parameter, how have I missed this...?
Stefan Metzmacher [Thu, 15 Sep 2005 09:55:16 +0000 (09:55 +0000)]
r10237: fix parameter, how have I missed this...?


14 years agor10236: fix (C) note
Simo Sorce [Thu, 15 Sep 2005 07:23:15 +0000 (07:23 +0000)]
r10236: fix (C) note

14 years agor10233: add commented PRAGMA to avoid fsyncs
Simo Sorce [Wed, 14 Sep 2005 23:14:42 +0000 (23:14 +0000)]
r10233: add commented PRAGMA to avoid fsyncs

14 years agor10232: Some work on ldb_sqlite3.
Simo Sorce [Wed, 14 Sep 2005 22:45:49 +0000 (22:45 +0000)]
r10232: Some work on ldb_sqlite3.
It is still far from being usable in samba4 but I want to commit
so that the work does not get lost by mistake.
This is also a good way to get comments if somebody is interested.

Sorry Derrell I ended up rewriting large parts of the code but I find
this style much more readable. Thanks for the hard work done. Your
work was a good reference for me.

ah the current code also shows some good numbers

sqlite3 generic test:
uid search took 0.05 seconds
real    0m12.492s
user    0m0.492s
sys     0m0.345s

with tdb we still get better numbers:
uid search took 0.46 seconds
real    0m0.892s
user    0m0.360s
sys     0m0.468s

but most of the time is spent in adding operations and I think
there's still a lot of space for improvement.


14 years agor10231: seem I flipped these, fix.
Simo Sorce [Wed, 14 Sep 2005 22:39:24 +0000 (22:39 +0000)]
r10231: seem I flipped these, fix.

14 years agor10228: Reorganizing a bit, trying to simplify. This is an attempt
Deryck Hodge [Wed, 14 Sep 2005 21:48:03 +0000 (21:48 +0000)]
r10228: Reorganizing a bit, trying to simplify.  This is an attempt
to find what's going wrong in IE formatting.

This is some better, but still IE needs help.