This patch accounts for the behaviour of the backend when a table with a
primary key constraint is created. The implicitly created index is named
based on the table name with _pkey (truncating the table name as necessary),
which may collide if the table name is not unique in the first
(maxnamelen-5) characters. Not sure if this should be considered a bug in
the jdbc driver or the server. For example, using psql:

test=# create table t_345678901234567890123456789_a (pk int not null primary key);
NOTICE:  CREATE TABLE / PRIMARY KEY will create implicit index 't_345678901234567890123456_pkey' for table 't_345678901234567890123456789_a'
CREATE
test=# create table t_345678901234567890123456789_b (pk int not null primary key);
NOTICE:  CREATE TABLE / PRIMARY KEY will create implicit index 't_345678901234567890123456_pkey' for table 't_345678901234567890123456789_b'
ERROR:  index named "t_345678901234567890123456_pkey" already exists
test=# 

I'm not convinced this patch is the "right thing" to do though.

Index: src/interfaces/jdbc/org/postgresql/jdbc1/AbstractJdbc1DatabaseMetaData.java
===================================================================
RCS file: /projects/cvsroot/pgsql-server/src/interfaces/jdbc/org/postgresql/jdbc1/AbstractJdbc1DatabaseMetaData.java,v
retrieving revision 1.13
diff -u -u -r1.13 AbstractJdbc1DatabaseMetaData.java
--- src/interfaces/jdbc/org/postgresql/jdbc1/AbstractJdbc1DatabaseMetaData.java	2002/12/11 21:02:58	1.13
+++ src/interfaces/jdbc/org/postgresql/jdbc1/AbstractJdbc1DatabaseMetaData.java	2002/12/17 03:28:17
@@ -1528,7 +1528,10 @@
 	 */
 	public int getMaxTableNameLength() throws SQLException
 	{
-		return getMaxNameLength();
+		// We must be unique in the first (namelength - 5) characters
+		// as the backend appends _pkey to generate an index name for primary
+		// key constraints when a table is created.
+		return getMaxNameLength() - 5;
 	}
 
 	/*