Forum OpenACS Development: PostgreSQL 7.2.1 ?bug?

Ok, I have tried this many times with PG-7.2.1 and I get the same results every time. I'm using PG-7.2.1 and tree_next_key(null,null) returns:

1000000000000000000000000000000

instead of:

00000000

So, I have looked more into this and to my surprise here's what I get:

# select 0<2^7;
 ?column?
----------
 f
(1 row)

# select 0<=2^7;
 ?column?
----------
 t
(1 row)

select 0<=(2^7-1);
 ?column?
----------
 t
(1 row)


# select 2^7 > 0;
 ?column?
----------
 t
(1 row)

 select 0::integer<(2^7)::integer;
 ?column?
----------
 t
(1 row) 

Could someone confirm this, please? To relate this with the tree_sortkey mechanism 0<2^7 is the condition used in the int_to_tree_key function. After personal communication with Don, I have changed acs-kernel/sql/postgresql/postgresql.sql from p_intkey < 2^7 to p_intkey < 128 so that we are sure.

I have tried this on a fresh pg-7.2.1 install. Maybe, it works for you due to the upgrade. Can someone else check it on a fresh install? David, have you restored your dumped data by using template1 or each database separately?

template1=# select version();
                           version                           
-------------------------------------------------------------
 PostgreSQL 7.2.1 on i686-pc-linux-gnu, compiled by GCC 2.96



template1=# select 0 < 2^7;
 ?column? 
----------
 f
(1 row)

From a clean install of 7.2.1:

demo=> select version();
                           version
-------------------------------------------------------------
 PostgreSQL 7.2.1 on i686-pc-linux-gnu, compiled by GCC 2.96
(1 row)

demo=> select 0 < 2^7;
 ?column?
----------
 t
(1 row)

David:

# select 0=2^7;
 ?column? 
----------
 t
(1 row)

Very strange. I'll have to check my system. No idea what could be wrong.

Then, something is seriously messed up with my system. Anyone has any ideas of what could be wrong?

# select 2^7;
 ?column? 
----------
      128
(1 row)

Same result in template1 and other databases. Here's how I configure and start PG, respectively:

Installing PostgreSQL 7.2.1

cd /usr/local/src
        tar -xzvf postgresql-7.2.1.tar.gz
        cd postgresql-7.2.1
        ./configure --enable-multibyte=UNICODE --with-tcl --with-perl --enable-syslog
        gmake
        gmake install
        gmake install-all-headers
        mkdir /usr/local/pgsql/data
        useradd postgres
        passwd postgres
        chown postgres /usr/local/pgsql/data
        su - postgres
        export LANG=C
        /usr/local/pgsql/bin/initdb  --encoding unicode  -D /usr/local/pgsql/data
        /usr/local/pgsql/bin/postmaster -D /usr/local/pgsql/data >logfile 2>&1 &        /usr/local/pgsql/bin/createdb test
        /usr/local/pgsql/bin/psql test
        /usr/local/pgsql/bin/dropdb test
        exit

Starting PG

/etc/rc.local

su -l postgres -c 'export LANG=C; /usr/local/pgsql/bin/postmaster -N 200 -B 1000 -o "-S 2000" -S -D /usr/local/pgsql/data '

I'm using RH-7.2. Only thing I can think of is that I messed something while trying to compile the new linux kernel with the grsecurity patches. I'm downloading RH-7.3, try to upgrade and see what comes out of it.

Same thing as your post (same result as well -- returns false). I'm still downloading RH-7.3. Hopefully, that would settle it.

I compiled and initdb'ed a Unicode postgres and got the right
answer.

I'd suggest running some of PG regression tests on that installation, here the full set of things above returns correct results:

select version();
                           version                           
-------------------------------------------------------------
 PostgreSQL 7.2.1 on i686-pc-linux-gnu, compiled by GCC 2.96
(1 row)

select tree_next_key(null,null);
 tree_next_key 
---------------
 00000000
(1 row)

select 0<2^7;
 ?column? 
----------
 t
(1 row)

select 0=2^7;
 ?column? 
----------
 f
(1 row)

select 2^7;
 ?column? 
----------
      128
(1 row)