A check on conversion failure was incorrectly taking into account the result of a previous column conversion on the same row. The defect was without practical implications, since a second, impossible condition would have had to match to lead to an incorrect failure. However the incorrect first check lead to a failing assertion.
The IRD descriptor was not re-init'd on statement handle closing. This defect is fixed just for good house-keeping, since the descriptor would be re-init'd on attaching a new result set to it (which happens regularly when the result for a query is being paged / using a cursor). Along with this, some code duplication has been removed.
A few logging statements had the wrong format specifiers, mostly expecting signed, instead of the correct unsigned types of the integer variants.
This PR bundles a few minor fixes:
A check on conversion failure was incorrectly taking into account the result of a previous column conversion on the same row. The defect was without practical implications, since a second, impossible condition would have had to match to lead to an incorrect failure. However the incorrect first check lead to a failing assertion.
The IRD descriptor was not re-init'd on statement handle closing. This defect is fixed just for good house-keeping, since the descriptor would be re-init'd on attaching a new result set to it (which happens regularly when the result for a query is being paged / using a cursor). Along with this, some code duplication has been removed.
A few logging statements had the wrong format specifiers, mostly expecting signed, instead of the correct unsigned types of the integer variants.