Kathey Marsden <[email protected]> writes:

> On 12/15/2010 6:20 AM, Knut Anders Hatlen wrote:
>> Knut Anders Hatlen<[email protected]>  writes:
>>
>>> Kathey Marsden<[email protected]>  writes:
>>>
>>>> On 12/13/2010 11:58 AM, Rick Hillegas wrote:
>>>>> I see that Ole ran this test on 10.5.3 and 10.6.1. Ole no longer
>>>>> works on Derby. As you note, no-one has run that test on 10.6.2 or
>>>>> 10.7.1.
>>>>>
>>>> Is there someone that has access to these tests that can run them now
>>>> that Ole is gone or is it just one more bit of testing we don't do
>>>> anymore?  It would be nice to know we are compliant for  a release off
>>>> trunk.
>>> I found the instructions for running these tests and have started
>>> them. So if everything goes well, we'll have the results shortly.
>> It didn't go all that well... Here's what the report says:
>>
>> batchUpdate.out:     passed: 68
>> callStmt.out:        passed: 1,580; failed: 20
>> connection.out:      passed: 36
>> dateTime.out:        passed: 152
>> dbMeta.out:          passed: 940
>> escapeSyntax.out:    passed: 324
>> exception.out:       passed: 56
>> prepStmt.out:        passed: 1,072; failed: 12
>> resultSet.out:       passed: 452; failed: 4
>> rsMeta.out:          passed: 84
>> stmt.out:            passed: 132
>> Total time: 09:58:28
>>
>> I haven't looked at all the failures yet. Those I have looked at appear
>> to happen because our meta-data now says that we support BOOLEAN,
>> whereas the database schema created by the test itself contains
>> workarounds for Derby's lack of support for BOOLEAN (like using SMALLINT
>> columns instead of BOOLEAN columns). So I think there are some changes
>> required for the DDL and DML scripts that create the database schema in
>> order to make the test run successfully, but I haven't found anything
>> that appears to be a Derby problem yet.
>>
> Would a workaround to changes the DML scripts for the short run be to
> create the database with an old version and run against the soft
> upgraded database now that the metadata issue in soft upgrade is
> fixed.

That may possibly work. At least it seemed to me that the failures I
analyzed wouldn't have happened if getTypeInfo() hadn't returned
boolean.

-- 
Knut Anders

Reply via email to