[prev in list] [next in list] [prev in thread] [next in thread] 

List:       markus-dev
Subject:    Re: Using buildbot for testing
From:       Mike Conley <mike.d.conley () gmail ! com>
Date:       2011-07-28 13:28:44
Message-ID: 4E31640C.7000005 () gmail ! com
[Download RAW message or body]

All:

I think it's totally worth trying.  Thumbs up from me.  :)

And yes, a huge thanks to Nelle and Benjamin, who've been 
single-handedly kicking butt on MarkUs this summer!

-Mike

On 11-07-28 8:54 AM, Nelle Varoquaux wrote:
>
>
> On 28 July 2011 13:51, Severin Gehwolf <severin.gehwolf@utoronto.ca
> <mailto:severin.gehwolf@utoronto.ca>> wrote:
>
>     Hi,
>
>         On 28 July 2011 08:56, Nelle Varoquaux
>         <nelle.varoquaux@gmail.com <mailto:nelle.varoquaux@gmail.com>>
>         wrote:
>
>             Hi
>
>             On 28 July 2011 01:27, Severin Gehwolf
>             <Severin.Gehwolf@utoronto.ca
>             <mailto:Severin.Gehwolf@utoronto.ca>>__wrote:
>
>                 Hi,
>
>                 I like the idea, we might hit resource constraints however.
>
>                 On Wed, 2011-07-27 at 17:41 +0100, Nelle Varoquaux wrote:
>                  > Benjamin and I have noticed recently that some tests
>                 seem to fail on
>                  > postgres and mysql but not sqlite.
>
>                 FWIW, I don't recommend to use sqlite for MarkUs
>                 development. I
>                 understand that it's convenient to set up, but we don't
>                 recommend it for
>                 production use at all and the more developers using
>                 sqlite the less
>                 coverage of mysql/pg versions we have. Personal opinion :)
>
>
>             Running tests on sqlite is much faster than to run them on
>             mysql or
>             postgresql. This is why I use sqlite for development.
>
>             In any case, having a buildbot running the tests on all
>             three environment
>             should insure that they all pass everywhere.
>
>
>     Cool.
>
>
>                  > Sqlite doesn't work quite the same as postgres and
>                 mysql when it comes
>                 to
>                  > the ordering of the results when doing a "SELECT" query.
>                  >
>                  > In order to have a better and more robust testing, I
>                 suggest that we
>                 replace
>                  > our automatic testing with something a bit more
>                 adapted: buildbot, in
>                 order
>                  > to test on the different database.
>                  > It will give us better and more reliable testing.
>                  >
>                  > I currently have an instance working on my computer,
>                 but no one apart
>                 from
>                  > me can access the reports. We can have an instance of
>                 buildbot running
>                 on
>                  > the server very quickly.
>
>                 If you volunteer to set it up on markusproject.org
>                 <http://markusproject.org>, I'm all for it.
>                 Having said that, I have my doubts that this box with
>                 its 512MB of RAM
>                 is beefy enough to run all the services we'd require for
>                 this to work.
>                 You can give it a shot, but we may have to roll-back if
>                 things go awry.
>                 At the moment we run tests using mysql as back-end on
>                 there once a day
>                 (branch master):
>                 http://www.markusproject.org/__dev/unit_tests_report.txt
>                 <http://www.markusproject.org/dev/unit_tests_report.txt>
>                 http://www.markusproject.org/__dev/functional_tests_report.__log
>                 <http://www.markusproject.org/dev/functional_tests_report.log>
>
>
>
>             It uses twisted, so no ressources at all. It will only use
>             ram when we
>             launch the tests (which we already to).
>             The advantage also is that we can set up the master on the
>             box, and slaves
>             elsewhere.
>
>
>     Hmm, are you sure? Where would the PostgreSQL DB engine run?
>     Wouldn't this kind of setup require a MySQL daemon _AND_ a
>     PostgreSQL daemon on markusproject.org <http://markusproject.org>?
>     That was what I was worried about. We're pretty much maxed out on
>     RAM already...
>
> We can run it on postgresql only. Right now, it only runs the tests on
> sqlite I think. In any case, I think it would be better to replace the
> cron job by something a bit more secure, even if we only run the tests
> on sqlite for now.
>
> I can also set up a slave on my work computer for mysql (if I manage to
> install mysql on my computer...).
>
>
>     But if you say it should work, I trust you :)
>
>     Cheers,
>     Severin
>
>
>


-- 
http://www.mikeconley.ca
[prev in list] [next in list] [prev in thread] [next in thread]