Hi,
> In my opinion it's OK to check in a patched Torque in 1.0 branch since > it is not used by OJB in production, only for the tests. And if you do > so, we can then release OJB 1.0.2 with passing regression tests for > CLOB/BLOB (Oracle and sapDB inclusive). > > Unfortunately this will require re-testing 1.0 branch for all platforms > after checkin of updated test schema. > Possibly other platforms need > similar Torque patches to pass.
Agree. I checked my local modified torque jar against sapdb without side-effects. Are you willing to check in all discussed changes?
<changes>
- change profiles for sapdb.profile/maxdb.profile replace 'database' with 'database.torque' property.
- fix blob test table in ojbtest-schema.xml
- patch torque-3.0.2.jar/sql/base/sapdb/dp.props file, replace
CLOB = LONG UNICODE
with
CLOB = LONG VARCHAR
- oracle specific changes
</changes>
regards, Armin
Martin Kal�n wrote:
Armin Waibel wrote:
I made the changes you mentioned. The problem now is a wrong mapping for sapdb in torque-3.0.2.jar/sql/base/sapdb/dp.props file. As sapdb does not support BLOB/CLOB the mapping is
...
BLOB = LONG BYTE
CLOB = LONG UNICODE
[snip]
to make the creation of the DB table possible it needs ... BLOB = LONG BYTE CLOB = LONG VARCHAR because LONG UNICODE is not supported in sapdb.
If I patch the torque jar the blob test run without problems.
So I suggest to patch the torque jar file (and add a note in the release-notes) to enable best support for all used databases.
What do you think?
Agreed - and as Tom states; if we are moving away from Torque it's not too important anyway. (I have not tested this using commons-sql in 1.0 branch, neither updated my 1.1 project lately to test these issues against HEAD.)
In my opinion it's OK to check in a patched Torque in 1.0 branch since it is not used by OJB in production, only for the tests. And if you do so, we can then release OJB 1.0.2 with passing regression tests for CLOB/BLOB (Oracle and sapDB inclusive).
Unfortunately this will require re-testing 1.0 branch for all platforms after checkin of updated test schema. Possibly other platforms need similar Torque patches to pass.
If we start going down this path we can also resolv the Oracle issues with locking tests, since I could then introduce a new Torque platform "oracle9" (as opposed to the shipped "oracle")- similar to our own platform switching between pre-9 and 9+ version of Oracle. This would allow mapping of TIMESTAMP = TIMESTAMP and nanosecond lock precision in the tests.
I suspect that there are similar issues (TIMESTAMP=>DATE = only sec. precision) that make the ODMG tests fail on Oracle, but I have yet to debug this tonight.
Regards, Martin
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
