AW: [OJB-developers] Let's fix the testsuite!

[Mahler T. <tho...@it...>]

Hi Oleg,

Oleg wrote:
> Hi!
> 
> I decided to investigate why "build junit" output looks so bad, 
> since this makes me nervous :)

good idea!

> BTW "build junit2" output is even worse on the second run.
> Here's what I found:
> 1) Several messages
> ERROR: Ignoring unknown Element orderby
> that are not counted as errors, this is probably due to 
> changes in syntax of 
> repository.xml
> 

right! These messages will disappear after my next checkin of the new
repository grammar.

> 2) A bunch of exceptions like
> java.sql.SQLException: Violation of unique index in statement 
> [INSERT INTO 
> PERSON_PROJECT (PROJECT_ID,PERSON_ID) VALUES (1,1)]
> They are not counted as errors, but they makes impression 
> that something goes 
> wrong. Actually this is just a way how OJB stores M:N 
> relations: it tries to 
> insert all records into an indirection table and ignores SQL 
> exceptions 
> telling that the record exists, 
> see 
> PersistenceBrokerImpl.storeMtoNImplementor()
> But JdbcAccess.executeUpdateSQL() logs these exceptions at 
> ERROR level.
> Something should be changed here. 

In my first draft I simply suppressed the respective error messages. But
this can't be the final solution!

> Of course it would be better to insert only new records and 
> not to ignore SQL 
> exceptions. 

I agree!

> But maybe there are some problems with this?

I don't see any. in storeMtoNImplementor() we should check if an entry
already exists. If so do nothing, else insert a new entry.

> BTW I wonder if the current algorithm works when some link 
> between Persons 
> and Projects is removed. 
> 

Currently there is no support to remove m:n entries automatically. They have
to be deleted manually.

> 3) Several errors "Object has been modified by someone else"
> I am not yet familiar with locking stuff, so could somebody 
> look into this?
> 

(Jakob answered on this already)

> 4) Table not found: ELVIS 
> I found this name in OdmgExamples.testBrokerCrash()
> so I don't know if this error is expected, 

Right that's an expected error: in testBrokerCrash() I'm provoking a Crash
of the broker (by using an unknown table 'ELVIS') to check the ODMG
Transaction abort mechanisms.

> but there are 3 
> failures and 
> 1 error for ODMG tests, and I don't see other error messages.
> 

That's bad! The last I performed the ODMG Junit tests I had 0 errors and 0
failures. These things must be fixed!

> 5) AssertionFailedError in 
> SodaExamples.testWithFakedQueryPreEmptUnlimited() 
> Any comments?
> 

Mhh, I don't know of any bug in the SODA stuff. The last time I checked it
was OK. 
When I'm finished with my changes to the repository syntax, I'll be able to
run JUnit tests again.
I'll have a look at 4. and 5. then!

cheers,

Thomas


> Thanks in advance,
>  Oleg
> 
> 
> 
> _______________________________________________________________
> 
> Don't miss the 2002 Sprint PCS Application Developer's Conference
> August 25-28 in Las Vegas -- http://devcon.sprintpcs.com/adp/index.cfm
> 
> _______________________________________________
> Objectbridge-developers mailing list
> Obj...@li...
> https://lists.sourceforge.net/lists/listinfo/objectbridge-developers
>