Thanks Army for the help.
Army wrote:
This is probably silly, but have you confirmed that the property is in
fact being set correctly? If you output the value of the "noTimeout"
field from the OptimizerImpl constructor, does it return "true"?
Yes, it seems to be set correctly and return true.
To see if memory is an issue for the hash join, maybe search for
"memoryUsageOK" in OptimizerImpl and print out the results of those
calls...?
Both when the test succeeds and when it fails it skips due to excess
memory usage twice leading up to the point of failure, so I am assuming
that is the same for both runs.
I'll try your other suggestions and get back. I wonder also if it is
worth trying the patch for DERBY-1905 to see if that resolves the out of
whack costs.
Thanks again for the help.
Kathey
