[ckan-dev] test failures

[David Read]

On 3 February 2011 22:35, David Raznick <kindly at gmail.com> wrote:

>
>
> On Thu, Feb 3, 2011 at 9:50 PM, David Read <david.read at okfn.org> wrote:
>
>> Hi all,
>>
>> I've changed buildbot to run ckan tests twice now: firstly with sqlite and
>> then again with postgres in the same buildbot script (I should split it into
>> two tasks, but that is for another day). I've fixed a few problems with the
>> tests under sqlite and they pass now, but there are still problems with
>> tests in postgres.
>>
>> Seb changed lots of 'rebuild_db' statements to 'clean_db', which then
>> requires 'init_db' at the start of the next test. So there are now 'init_db'
>> statements at the start of a great many tests. Was there any reason for
>> this?
>>
>
> This may not be necessary any more with the fix we did the other day on
> migrate.  However, the in memory sqlite database gets lost mysteriously
> sometimes and its hard to figure out where.  That is why I thought the
> init_dbs were there.
>

These problems are with postgres in this case, and there is no mystery about
losing the d.b. - we call clean_db!

Note to Seb: Things were probably working in the past because the
'faster_db_hacks' config option wasn't being read properly, which David and
I fixed last Friday, so the full init_db was happening even with postgres,
whereas it's not now. Also the init_db parameter 'conditional' (BTW what a
terrible name for a parameter...!) is not being used. I think basically this
all needs a bit of a review / tidy up to get it to make sense and work too.


> It seems a right faff to remember this - is there no better way? Indeed I
>> believe the current test failures are caused by a test not having a init_db.
>>
>> Note for everyone: Seb, James and I have agreed to set in test.ini to use
>> sqlite.
>>
>
> I disagree to this.  You should specify what database you want in your own
> development.ini.  The template development.ini could have sqlite as the
> default instead.  This is the only sane way to do it in my opinion.
>

Yes this might be better. Anyone else?


> We think it is a good idea to default tests to being run quickly. However,
>> I can't see a good way to then run tests under postgres - if you comment out
>> the sqlite in test.ini (as the buildbot currently does), 9/10 times you end
>> up checking it in by mistake. Maybe someone has a better idea, since passing
>> parameters to nose is a no no.
>>
>
> Why is this a no no?
>

>From what I read, nose doesn't believe in passing parameters onto tests -
they should be stateless. You can fudge it by hacking argv, but then you
lose flexibility of the many useful nose command-line params.

Dave


>
> David
>>
>> _______________________________________________
>> ckan-dev mailing list
>> ckan-dev at lists.okfn.org
>> http://lists.okfn.org/mailman/listinfo/ckan-dev
>>
>>
>
> _______________________________________________
> ckan-dev mailing list
> ckan-dev at lists.okfn.org
> http://lists.okfn.org/mailman/listinfo/ckan-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.okfn.org/pipermail/ckan-dev/attachments/20110204/0bdd568b/attachment-0001.html>