Hannes Dorbath <light(at)theendofthetunnel(dot)de> writes:
> Hannes Dorbath wrote:
>> Yes, I'll try to catch the row.
> I just slammed a BEFORE UPDATE FOR EACH ROW trigger with RAISE NOTICE on
> the table to catch the row ID where it fails.
> The row ID is never the same. Neither can I reproduce it when I use the
> last reported row ID.
So most likely it's triggered by a sequence of operations rather than
any individual row :-(. You'll have to think in terms of producing
a multi-table-row test case.
regards, tom lane