Dave Pooser <dave-pg@pooserville.com> writes:
> On 7/2/11 7:16 PM, "Tom Lane" <tgl@sss.pgh.pa.us> wrote:
>> The Oracle documentation avers that "-lldap" ought to be sufficient;
>> see for example ldap_simple_bind page in
>> http://download.oracle.com/docs/cd/E19963-01/pdf/821-1466.pdf
>> and even without that evidence it's hard to believe that they'd expect
>> calling applications to hardwire a library version number into their
>> build scripts.
> They don't; they cheat and symlink:
> root@testdb:/home/locadmin/postgresql-9.0.4# ls -l /usr/lib/libldap.so
> lrwxrwxrwx 1 root root 12 Jul 2 14:48 /usr/lib/libldap.so
> -> libldap.so.5
That's not a cheat; that's exactly the way everybody else does it.
That's how you get from a non-versioned library "-l" switch in a
makefile to a version-numbered reference in the resulting executable.
(The linker is supposed to dereference the symlink at link time.)
> But there's no such libldap_r.so symlink created by default.
Then you've got a broken package. Make the symlink by hand, and file a
bug report with Oracle.
>> If all else fails you could --disable-thread-safety.
> Is that safe to do on a server with a few dozen concurrent users?
The server doesn't care about this. It's only an issue for putting
libpq into multithreaded application programs.
regards, tom lane