Jan Wieck wrote:
<snip>
> What I fail to see is the actual error message that aborts the 
> transaction. Seems the log output was a bit too much for your syslog to 
> swallow. Can you run the postmaster in the foreground and pipe 
> stdout+stderr through tee(1) or something like that?

No worries.  Just did that and captured the log file.  It shows the 
error properly I think.  Might need to turn up the verbosity level 
though.  Only included the end section where it has the error, as the 
bits before seem to be fine.  Can attach the whole log file if required.

***********
<snipped>
         CREATE OPERATOR <= (
                 PROCEDURE = "_test1"."xxidle",
                 LEFTARG = "_test1"."xxid",
                 RIGHTARG = "_test1"."xxid",
                 COMMUTATOR = >=, NEGATOR = >,
                 RESTRICT = scalarltsel, JOIN = scalarltjoinsel
         );
         CREATE OPERATOR >= (
                 PROCEDURE = "_test1"."xxidge",
                 LEFTARG = "_test1"."xxid",
                 RIGHTARG = "_test1"."xxid",
                 COMMUTATOR = <=, NEGATOR = <,
                 RESTRICT = scalargtsel, JOIN = scalargtjoinsel
         );


         --
         -- Finally the default operator class so that we can use our
         -- new data type in btree indexes.
         --
         CREATE OPERATOR CLASS "_test1"."xxid_ops"
                 DEFAULT FOR TYPE "_test1"."xxid" USING btree AS
                 OPERATOR 1 <  ("_test1"."xxid", "_test1"."xxid"),
                 OPERATOR 2 <= ("_test1"."xxid", "_test1"."xxid"),
                 OPERATOR 3 =  ("_test1"."xxid", "_test1"."xxid"),
                 OPERATOR 4 >= ("_test1"."xxid", "_test1"."xxid"),
                 OPERATOR 5 >  ("_test1"."xxid", "_test1"."xxid"),
                 FUNCTION 1 "_test1"."btxxidcmp" ("_test1"."xxid", 
"_test1"."xxid");



NOTICE:  type "_test1.xxid" is not yet defined
DETAIL:  Creating a shell type definition.
ERROR:  could not access file "$libdir/xxid": No such file or directory
LOG:  statement: select 1;
ERROR:  current transaction is aborted, commands ignored until end of 
transaction block
LOG:  statement: rollback transaction;
LOG:  received fast shutdown request
LOG:  shutting down
LOG:  recycled transaction log file "000000000000002E"
LOG:  database system is shut down
bash-2.05b$

***********

Does this just mean that a $libdir isn't getting translated correctly by 
sed or something?

Regards and best wishes,

Justin Clift