ubuntu-phone team mailing list archive

[Kevin Gunn <kevin.gunn@xxxxxxxxxxxxx>]

hey Didier/Leo - yes, this is precisely the kind of integration test that
I've been outlining that we should perform with AP using real components.
Something mimicking the manual test i am performing would cover quite a bit
of ground. It would cover some Dash, some click, some accounts, and some
window management.

Note, I would prefer real components rather than fake application. For
covering all that ground listed above, as well, our confidence will be
higher. For instance, i ran the AP suite for ubuntu-keyboard (osk) and it
passed 100%, I guess that makes me worry we might not catch reality trying
to mock it out.

Lastly, this isn't something we'd want to cram in for the next mir attempt
(i hope today/tomorrow) - but yes, it would be highly valuable if you could
prioritize this work as I think its a really useful integration test case,
as it covers so many component interactions of a very important ux case.
br,kg




On Fri, Mar 7, 2014 at 12:44 PM, Leo Arias <leo.arias@xxxxxxxxxxxxx> wrote:

> On Fri, Mar 7, 2014 at 1:07 AM, Didier Roche <didrocks@xxxxxxxxxx> wrote:
>
>> Le 07/03/2014 04:09, Kevin Gunn a écrit :
>>
>>  Hey Didier - on the Mir build topic.
>>> the build looked really good, AP tests were passing...including the
>>> keyboard AP test. however, through manual testing I found a regression we
>>> created somehow.
>>> the bug is here
>>> https://bugs.launchpad.net/unity-mir/+bug/1289058
>>>
>>> for now, we'll take mir out of silo and let Mr Voss back in :)
>>> we'll try again once we nail down a solution.
>>>
>>
>> Ok, thanks for trying and then freeing to let Thomas continue then :)
>>
>> Just curious, we didn't get any AP failure on that specific issue? Can we
>> think of a way to catch that automatically through an integration test?
>> Thanks again for the serious dogfooding and the feedback ;)
>>
>
> One important thing to notice, that I mentioned to Kevin and some other
> people, but I'm not sure if everybody knows.
>
> Currently, on the devices we are not using the OSK to test the apps. We
> are simulating keystroke events so things get written to the text fields
> without simulated fingers being involved. That is wrong, of course.
>
> We have a proposed solution for that, using the TextField custom proxy
> object [1]. For that to happen, we just need to agree on a way to start
> maliit with testability. Well, and update the apps that are not using the
> helper, which is easy, maybe just a little tedious.
>
> However, I suppose that's not a change we want to introduce before we
> stabilize the current release, right?
>
> Now back to this specific issue, it sounds like we can prevent it to
> regress with a simple autopilot test that opens a fake application with two
> text fields. If there's no way to test this at a lower level, let me know
> and I'll help you doing it on autopilot.
>
> pura vida.
>
> [1]
> http://bazaar.launchpad.net/~ubuntu-sdk-team/ubuntu-ui-toolkit/trunk/view/head:/tests/autopilot/ubuntuuitoolkit/emulators.py#L456
> --
> ¡paz y baile!
> http://one.ubuntu.com
>