Re: [eclipselink-dev] Updated Proposal for submitted DatabasePlatform in

[Dies Koper <diesk@xxxxxxxxxxxxxxxxxxx>]

Hi Markus,

I don't know the background of your troubles with the tests as I think I
joined this list after that, but I have some questions/comments.

> Sounds good, but in the particular case of SQL Anywhere, doesn't match the
> reality: While my contributed code at runtime is superior to the existing
> code, the existing code passes more tests. Sad but true, this is not a
> problem of the platform, but of the particular tests. As many of them are
> not well documented, it is nearly impossible to understand what they intend
> and what the reason is why they not pass. In the SQL Anyhwere scenario many

When I ran tests on the Symfoware platform I also had problems
understanding why tests did not pass. Sometimes the cleanup or error
reporting logic was buggy so the error message did not contain enough
info, caused an exception itself that masked the original exception, or
failures occurred because previous tests left the tables in a wrong state.

I also found it difficult to understand what tests intend to do, most
are not documented, but with most issues I encountered I don't think it
would have helped knowing what the tests were trying to do anyway.
It usually was my DB or driver that did not support a JDBC method or SQL
thrown at it, or put it in a state that would make the next operation
fail. As the test did not intend to put it in such a state, I don't know
what kind of documentation could have helped me.

> of the tests are written in the expectation that SQL Anywhere is derived
> from Sybase ASE, which it actually is not the case. So it is quite obvious

I can see the platform class extends Sybase's, which shouldn't be hard
to fix, but how are tests written to be expecting that?
I can't recall encountering such issues with Symfoware, so I'm very curious.

I did find a few locations were functionality was expected to be
supported by only Oracle (and e.g. MySQL) so I had to implicitly change
it to run on Symfoware as well. Except for not knowing an easy way to
find such tests (I found them by chance and might not have found all),
they were not causing any problems.

> that the tests will pass with a "wrong" platform much better than with a
> "correct" one. In the end, this cannot be solved without massive help of
> Oracle to either document or rewrite the tests.

I had to rewrite a number of tests (70 or so?), and even in core code
because behaviour was expected that was different on Symfoware. (In
these cases the code could be changed without affecting the intention of
the test, and without breaking it on other DBs). I have been receiving
enough support from Tom and Andrei to make most pass.

That said, if I had known (if my manager had known) the contribution of
our DB platform would to take as long as it's been taking (almost 6
months now, and still counting), we probably wouldn't have done it.

Hopefully this proposal (and the experience Tom and the team have gained
with including the Symfoware platform) will make the process go faster
for other future contributions.

Regards,
Dies


>> -----Original Message-----
>> From: eclipselink-dev-bounces@xxxxxxxxxxx [mailto:eclipselink-dev-
>> bounces@xxxxxxxxxxx] On Behalf Of Tom Ware
>> Sent: Donnerstag, 18. Februar 2010 16:34
>> To: dev mailing list for Eclipse Persistence Services
>> Subject: [eclipselink-dev] Updated Proposal for submitted
>> DatabasePlatform integration
>>
>> Hi All,
>>
>>     I have updated the proposal for lowering the barrier for integrating
>> contributer-submitted DatabasePlatforms into the product.  Please
>> provide
>> feedback either via this mailing list, or the Discussion page.
>>
>> http://wiki.eclipse.org/EclipseLink/Development/Incubator/Extensions/Da
>> tabasePlatformPromotion
>>
>> -Tom