Sounds like we are getting close. From strace:
stat64("/home/pramsey/pgtest/8.0/bin/lib/postgresql/plpgsql",
0xbfffe480) = -1 ENOENT (No such file or directory)
stat64("/home/pramsey/pgtest/8.0/bin/lib/postgresql/plpgsql.so",
0xbfffe480) = -1 ENOENT (No such file or directory)
stat64("$libdir/plpgsql", 0xbfffe540) = -1 ENOENT (No such file or
directory)
I started the database from "/home/pramsey/pgtest/8.0/bin" using
"./pg_ctl start -D /home/pramsey/pgtest/8.0/data"
Judging from the behavior I have posted to the list of "pg_config
--pkglibdir", if I started it from any *other* location at all then
things would have worked.
Paul
Tom Lane wrote:
> Paul Ramsey <pramsey@refractions.net> writes:
>
>>... It is the actual binaries that seem to not know
>>where $libdir is supposed to be.
>
>
> Where do they think it is?
>
> A useful way to check would be to strace the backend while you're
> executing "createlang" or another command that tries to load a
> dynamic library. The file names that it tries to stat() would be
> good evidence.
>
> Also, how are you starting the postmaster? I believe that current
> sources will try to locate $libdir via a relative path from where
> the postmaster executable is. We've fixed some bugs in that area
> recently, but maybe there are more.
>
> regards, tom lane