SYSTEM ERROR: (40) (14675): Repeatedly NEWing a class with a NO-UNDO TEMP-TABLE in a TRANSACTION

Repeatedly NEWing a class that defines a NO-UNDO TEMP-TABLE inside a TRANSACTION will eventually crash the session, produce protrace and procore files and generate error (40) (14675)

The issue occurs only when the code is executed against the r-code of the class.

The issue does not occur when the code is executed against the uncompiled source code of the class.

The no UNDO TEMP-TABLE definitions of each class instance NEWed inside the TRANASCTION are not being properly released as they should be.

When there is an active task, the object is linked to a pending reuse chain.
When the outermost transaction ends, objects on the chain are processed and moved to the reusable chain, in the event the temp-table is needed during undo processing necessary for UNDO temp-tables.
For NO-UNDO temp-tables, this results in a number of instances of the same object in memory, which eventually hits the limit on the maximum number of indexes.

The issue does not occur when the code is executed against the uncompiled source code of the class beacuse the AVM does not try to reuse an object unless it has been compiled.

Upgrade to OpenEdge 10.2B06 Service Pack or later, OpenEdge 11.1.0 or later.

• When there are only NO-UNDO temp-tables, the object is linked to the reusable chain.
• When the transaction ends, UNDO temp-tables are then linked to the reusable chain.

Use the -reusableObjects 0 session startup parameter in the ABL session startup command line.

Progress Articles: 

AppServer SYSTEM ERROR: Attempt to define too many indexes for area 6 database DBIxxxxaxxxxx. (40)   
How to troubleshoot "Attempt to define too many indexes" errors for OpenEdge clients ?