On Wed, Mar 27, 2013 at 09:20:23PM +0100, a b wrote: > > The point is to use instantclient libs, not the server libs, as intended. > > Ah, so. I always avoid instant client libraries, because they were never > linked correctly, libtclntsh.so cannot find libnnz.so because Oracle does not > link with the $ORIGIN linker keyword. If they did that, instant client > libraries would be great. As they are, they require one to set > LD_LIBRARY_PATH, which is a nasty, nasty hack which we absolutely refuse to > do. > > So what I end up doing is installing the Oracle RDBMS in our software stack > prefix, which causes the installer to link libclntsh.so with libnnz.so > correctly. Then I patch the relevant Makefile from Oracle to include the > $ORIGIN keyword, and relink. > > Finally, I grab libclntsh.so and libnnz.so, package them up separately and > make the Oracle package depend on the client libraries' package. > > > > And at least on my devsystem, the instantclient-devel package installs > > under /usr/include. Ofcourse it might make some sense to add it to the list. > > How strange that they would do such a thing, since it violates all those > standards, one of which is their own! >
Full path is /usr/include/oracle/11.2/client64/ and libs go into /usr/lib/oracle/11.2/client64/lib not 100% sure if this is something caused by alien or the instantclient-devel rpm. Aki
signature.asc
Description: Digital signature
_______________________________________________ Pdns-users mailing list [email protected] http://mailman.powerdns.com/mailman/listinfo/pdns-users
