from Booker Bense <bbense@networking.stanford.edu>.

avoid clashes with Kth Kerberos.

  export so we don't clobber anything the user already has set

  with the shared libraries instead of static, defaults to no
* build/lib-shared.mk: if LINK_BINS_DYNAMIC is set we create a symlink to
  the .so and .so.# file along with the .a and .la files
* build/lib.mk: make sure the above links get removed on clean target
* build/top.mk: add define for LINK_BINS_DYNAMIC
* tests/scripts/defines.sh: add export for LD_LIBRARY_PATH so that tests will
  run without requiring installation of libraries when we use --enable-dynamic
  (LD_LIBRARY_PATH is always set, since it can't really hurt).

schemes are supported.  This should facilate server rootDSE
advertisement of supported schemes, etc..

Remove Kerberos project.  Looks like K4 is working again.

should become the master...

ldap_pvt_tls_config() was missing first param.

ldap_pvt_tls_set_option() expects int* as third param.

manpage carefully.  One should pass strcasecmp to ldap_sort_entries, not
ldap_sort_strcasecmp.  Ldap_sort_strcasecmp is for ldap_sort_values, NOT
ldap_sort_entries.

ldap_sort_entries work with ldap_sort_strcasecmp as the comparison
function.  Pointers were being dereferenced, and should only have been
casted.

This is technical incorrect, but we don't have the codes in place
to correctly handle empty values yet.

as an empty value.  Complain, of course, as this isn't valid LDIF.

s/dn_normalize/dn_validate/
s/dn_normalize_case/dn_normalize/

back out previous commit.

by ldap_parse_result unless that value is SUCCESS.

Old code applied sync flags to store(), however supported DBMs
require such flags to be specified during open().  The new
code now applies flags in ldbm_cache_open (which calls ldbm_open).
ldbm_cache_close() now calls ldbm_sync().  This will force
a updating of on-disk contents after each LDAP operation.
The old code either failed to sync the on-disk contents until
close or synced on every store.   Per LDBM operation syncing
*should* be safe enough... real data safety requires transactions.
Removed nosync option from BDB2 as it is not compatible with
txn support.
Also added code to disable DBM level locking as slapd is only
process acessing the databases (dbnolocking).

Fix creation of ldapadd link to ldapmodify

The loading step in slapd build was failing because of a weird
@SLAPD tag.  Removing it fixes the problem for me, but this should
be reviewed by someone who knows which @SLAPD_...@ tag was meant to
be there, if any.

  and related AC_SUBST() so that when back-tcl is compiled as a dynamic module we link
  -ltcl to the module and not slapd (this is the correct way to do this since the .la file
  handles giving the correct libs when we pass the module to it in the linker line with
  -dlopen). Also modified the perl backend in a similar way except that the PERL_CPPFLAGS
  always go to the module and never to slapd (slapd doesn't need them).
* build/mod.mk: added $(MODLIBS) to the dynamic module link line to accomodate module
  specific libraries. These should be defined in the back-*/Makefile.in file for each
  module (so far only back-tcl and back-perl need it).
* build/top.mk: modified the perl ldflags and cppflags slightly
* servers/slapd/Makefile.in: same here
* servers/slapd/back-perl/Makefile.in: added MODLIBS=$(MOD_PERL_LDFLAGS) for when we are
  using a dynamic module (problem, libtool wont allow linking static libs into a libtool
  lib, so unless perl's libs are compiled dynamic, then back-perl can't be a dynamic
  module. We need a test for this on perl and tcl).
* servers/slapd/back-tcl/Makefile.in: added MODLIBS=$(MOD_TCL_LIB)
* back-perl and back-tcl now compile