On Tue, 24 Sep 2013 12:31:20 -0300
Douglas Campos <qmx(a)qmx.me> wrote:
On Tue, Sep 24, 2013 at 04:19:10PM +0200, Karel Piwko wrote:
> Hi All,
>
> QE needs to setup jobs that would allow us to test more than what's
> available in Travis - Ubuntu hosts are pretty limiting plus they contains
> OpenJDK related bugs [1][2].
>
> The problem is that only instance we have access to is behind VPN, so it
> might be complicated/impossible to read the results/get better feedback.
>
> Following questions needs to be answered, some suggestions included in []:
>
> * Where to put result logs? [
http://hudson.jboss.org/hudson/]
This might be a good place, since it's already set up.
> * How to notify and who to notify? [email|push messages ;-)|some specific
> ML|commiter only]
We can always create a build status mailing list and funnel everything
there.
> ** IRC is unfortunately a no-go here
boo :(
> * When to notify? [failure only|unstable|always]
Failure only
Sorry, I was not explicit here:
* Failure: job didn't finish (aborted to to some error)
* Unstable: some tests failed
What about specific states, like still-failing or still-unstable?
> * How often to run tests? [nightly|per commit]
This really boils down to what kind of tests are being run, and how busy
is the build-slave queue.
Well, per commit strategy is unrealistic setup when trying to rerun everyting.
> * How to enable developers to easily rerun something? [email|irc]
irc would be easier, email safer?
I prefer IRC, however QE does not cover all timezones. In such cases email
makes more sense.
> * Something else?
how crowded is this env? do we have OSX slaves on it?
Pretty crowded but this does not mean we can't get some time there :-).
As for OS X slaves, they are especially crowded.