By the way, the main hint was seeing some warnings about constant
redefinition (DB). Of course, if I had seen your previous message it would
have helped too :)
Em quinta-feira, 9 de janeiro de 2014 15h42min35s UTC-2, Rodrigo Rosenfeld
Rosas escreveu:
>
> Em quinta-feira, 9 de janeiro de 2014 14h12min08s UTC-2, Jeremy Evans
> escreveu:
>>
>> On Thursday, January 9, 2014 6:37:01 AM UTC-8, Rodrigo Rosenfeld Rosas
>> wrote:
>>>
>>> I'm struggling with an issue with one of my tests which not require the
>>> full Rails environment...
>>>
>>> When looking at the full statements log, I can only see a single BEGIN
>>> line and no COMMIT statement, only a ROLLBACK in the end. And there are
>>> no INSERT statements outside the BEGIN-ROLLBACK block. But after running
>>> a single test I can see that the database is no longer empty and I have
>>> to recreate it...
>>>
>>> If I replace spec_db_only_helper with spec_helper, which loads the full
>>> Rails environment, it works for some reason. I can't understand this.
>>>
>>> Another difference is that I get several SELECT NULL; statements when
>>> using spec_db_only_helper while they don't show up with spec_helper.rb.
>>>
>>> There's a single connection in the pool for the test environment.
>>>
>>> Also, I had to call save_changes with transaction: false in my
>>> factory-girl Model.save! implementation otherwise it would generate some
>>> BEGIN-COMMIT blocks inside a DB.transaction(savepoint: true, rollback:
>>> :always){} block.
>>>
>>> Any ideas what could be happening here?
>>>
>>
>> My guess would be problems with your environment. If BEGIN-COMMIT are
>> being issued inside a DB.transaction block, and you don't have sharding
>> enabled and aren't using threads, my only guess is that the model's
>> database doesn't match DB. The SELECT NULL statements only showing up with
>> spec_db_only_helper and not with spec_helper also indicates that two
>> different Database objects may be in play. Is Sequel::DATABASES the same
>> when using both helpers?
>>
>> Thanks,
>> Jeremy
>>
>
> Sorry, I missed your answer. I sent the e-mail from Thunderbird and found
> out it doesn't automatically mark the message to be followed by me so I
> didn't get the e-mail and just noticed it by chance when I visited this
> topic in the browser to answer my own questions.
>
> I'm not sure what was happening, but I know how to fix it. I was using
> "require 'myapp/db'" in some places and "require_dependency 'myapp/db'" in
> other places, which were causing the file to be load more than once,
> extending Sequel multiple times (DB.extension(:connection_validator), for
> instance).
>
> This was the cause for several "SELECT NULL" I was seeing, I guess... But
> it also fixed all other problems. But I still don't understand why the
> records were remaining when looking at the statements in the log.
>
> My guess is that Sequel was recreating the connection multiple times, and
> some of those connections were not inside the main transaction (which has
> been canceled by the first closed connection).
>
> I just find it weird that I didn't get any warnings or errors. I'd expect
> ROLLBACK to raise an error if there was no previous BEGIN, but it seems
> it's not the case with PG, so I didn't get any errors. Tried: "DB <<
> 'ROLLBACK'". In psql I just get a warning...
>
> That would explain why the records are kept...
>
> I'm only not sure why the connection would be recreated several times, but
> anyway, I shouln't be extending Sequel multiple times either :P
>
> Sorry for the distraction.
>
--
You received this message because you are subscribed to the Google Groups
"sequel-talk" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/sequel-talk.
For more options, visit https://groups.google.com/groups/opt_out.
