Well, I dont know about the solution in Cayenne.
this is my env.
sun-jdk 1.5.0.06
postgresql 8.1.3
cayenne 1.2M11
jdbc driver 8.2dev-501 JDBC 3 // Do you use beta? or you mean 5.01??       Yours 8.2dev-510.jdbc3
http://jdbc.postgresql.org/download.html
Windows. // however this is java.
"It works with an empty project "
I have table and model wihtout relations in db. I added relations to db and reengineerd. I see now the relations in the Cayenne GUI.
I alos added completely new table with relationships with an old table and reengineerd the model. 
It works here.
I see the relations twice even now like
--------------------------------------
Old Table Artist
New Table test
After reengineer
Old Table Artist
        toArtist
        toArtist1   // I think this should be cleaned in Cayenne.
New Table test
        toArtist
--------------------------------------
 
If you mean this doesnt work for you, i would update to 1.2M11 first.
Sako.
  ----- Original Message ----- 
  From: Tomi NA 
  To: cayenne-use..bjectstyle.org 
  Sent: Saturday, March 11, 2006 6:25 PM
  Subject: Re: reverse engineering a postgresql database: no relationships detected?
  On 3/11/06, WONDER <mmmmmmmmm5..eb.de> wrote:
    Hello,
    I am not sure if i understand you.
    I am using PostgreSQL 8.1.3 with the jdbc 8.2dev-501 JDBC 3
    and already imported tables with thier relationships!
    Try to update the server/jdbc and if now maybe you ask again to send the other settings.
    Sako.
  I'm updating postgresql to 8.1.3 as I write this: I'll see what I come up with.
  As for other relevant information, here's my environment description:
  sun-jdk 1.5.0.06
  postgresql 8.1.2
  cayenne 1.2M10
  jdbc driver 8.2dev-510.jdbc3
  gentoo linux, custom 2.6.15 kernel
  Status update:
  I've just upgraded pgsql from 8.1.2 to 8.1.3 and now reengineering works. Sort of. It works with an empty project (=the first time it's called). If I add a test table referencing an existing table in the database and try to reengineer, I see the same behaviour as before. 
  I'd say this suggest that we should look for a solution in the cayenne code, instead of the pgsql-jdbc code.
  Tomislav
This archive was generated by hypermail 2.0.0 : Sat Mar 11 2006 - 14:27:11 EST