[ https://issues.jboss.org/browse/TEIID-3386?page=com.atlassian.jira.plugin... ] Steven Hawkins commented on TEIID-3386: --------------------------------------- This is an oracle bug. Our source query: SELECT g_1.INTKEY AS c_0, g_2.INTKEY AS c_1, g_0.INTKEY AS c_2 FROM LargeB g_0 LEFT OUTER JOIN (SmallA g_1 FULL OUTER JOIN MediumB g_2 ON g_1.INTKEY = g_2.INTKEY) ON g_2.INTKEY = g_0.INTKEY WHERE g_0.INTKEY < 1500 ORDER BY c_2, c_1, c_0 Will produce the correct results against a different database, such as SQL Server. The same is true even if we preserve the full outer join: SELECT g_1.INTKEY AS c_0, g_2.INTKEY AS c_1, g_0.INTKEY AS c_2 FROM LargeB g_0 FULL OUTER JOIN (SmallA g_1 FULL OUTER JOIN MediumB g_2 ON g_1.INTKEY = g_2.INTKEY) ON g_2.INTKEY = g_0.INTKEY WHERE g_0.INTKEY < 1500 ORDER BY c_2, c_1, c_0 Reversing the full outer join, does produce the correct results against Oracle 10: SELECT g_1.INTKEY AS c_0, g_2.INTKEY AS c_1, g_0.INTKEY AS c_2 FROM (SmallA g_1 FULL OUTER JOIN MediumB g_2 ON g_1.INTKEY = g_2.INTKEY) FULL OUTER JOIN LargeB g_0 ON g_2.INTKEY = g_0.INTKEY WHERE g_0.INTKEY < 1500 ORDER BY c_2, c_1, c_0 So that's obviously a bug as the results from a full outer join should be the same either way. I'm not finding good information on how to categorize this type of oracle bug other than seeing that there are a lot of known issues even through Oracle 10 with full outer joins. And this has been fixed in Oracle 11. So I'm not sure what we should do to resolve this at the moment. -- This message was sent by Atlassian JIRA (v6.3.11#6341)