[Date Prev][Date Next]
Re: support for dynamically loadable modules (ITS#196)
Julio Sánchez Fernández wrote:
> Bastiaan Bakker wrote:
> > > Could libtool be of help here?
> > >
> > Probably. I still have to learn how to use libtool, that's why I did not make use of it.
> Well, anyway, it is dying in ldif2ldbm. No core. Nothing visible
> anywhere. Output is:
> >>>>> Starting test001-ldif2ldbm ...
> running defines.sh . ldbm
> Datadir is ./data
> Cleaning up in ./test-db...
> Running ldif2ldbm to build slapd database...
> ldif2ldbm failed!
> >>>>> ./scripts/test001-ldif2ldbm failed (exit 1)
> make: *** [test-ldbm] Error 1
> Seems like ldif2ldbm (and maybe other programs) needs to be fixed for
> this. Go to directory tests and do 'make test' there.
When I increased the debug level I saw your problem is caused by failure to load the
back-ldbm module. The module had some unresolved dependencies. A rebuilt of the module linked
with the proper libraries solved the problem in make case. I did:
gcc -g -O2 -shared -o libback-ldbm.so *.o -L/opt/testbin/openldap/lib -L../../../libraries/
-lavl -lldbm -lldif -lldap_r -llber -llutil
Your problem clearly indicates that the backend Makefiles need rules to build the modules,
doing it by hand is no fun.
Also it indicates failure to load a module may better be a fatal error. That way it will be
easier to spot problems like these. Maybe it should be configurable: e.g. failure to load a
module with non exportable crypto should not be fatal, whereas failure to load a backend
fn: Bastiaan Bakker
org: LifeLine Networks BV