Using Openedge version 11.2 & Progress Developer Studio.
I'm using several TEMP-TABLE definitions (each in a separate include file) to form a DataSet. Everything looks OK if I use only one pair RELATION-FIELDS for a single data relation. As soon as i add another RELATION-FIELDS pair to the definition and drop the include file on the form (or import the schema via "Import Schema from File" button) the DataSet shows a duplicate child table with no columns. For simplicity sake and testing, I've set up three test files :
tt1.i - First TEMP-TABLE:
/* Temp Table 1 */
DEFINE TEMP-TABLE tt1
FIELD tt_test AS CHARACTER
FIELD tt_rel_field_1 AS INTEGER
FIELD tt_rel_field_2 AS INTEGER
INDEX tt_idx tt_rel_field_1 tt_rel_field_2.
tt2.i - Second TEMP-TABLE:
/* Temp Table 2 */
DEFINE TEMP-TABLE tt2
FIELD tt_test2 AS CHARACTER
FIELD tt_rel_field_1 AS INTEGER
FIELD tt_rel_field_2 AS INTEGER
INDEX tt_idx tt_rel_field_1 tt_rel_field_2.
dsTest.i - Dataset definition:
/* Dataset Definition */
{tt1.i}
{tt2.i}
DEFINE DATASET dsTest FOR tt1 , tt2
DATA-RELATION drTest FOR tt1 , tt2
RELATION-FIELDS (
tt_rel_field_1,tt_rel_field_1,
tt_rel_field_2,tt_rel_field_2
).
Printscreen of what happens when i drop dsTest.i on the form :

If I remove the second pair, everything works fine - GUI wise. Am I missing something obvious here ? All the examples I've found so far all use a single RELATION-PAIR. Now I wonder why. According to Progress Knowledgebase article 000018088 there is no voodoo involved.