ubuntu-phone team mailing list archive

[Olivier Tilloy <olivier.tilloy@xxxxxxxxxxxxx>]

On Wed, Nov 6, 2013 at 5:51 AM, Paul Larson <paul.larson@xxxxxxxxxxxxx>wrote:

> It's getting late here, but wanted to give a heads up on some observations
> so far for image 13.
>
> First, ui-toolkit seems to have added quite a few more testcases. This is
> always great to see!
>
> Gallery - new failures:
> gallery_app.tests.test_events_view.TestEventsView.test_delete_a_photo
> gallery_app.tests.test_photo_viewer.TestPhotoViewer.test_photo_delete_works
> gallery_app.tests.test_photos_view.TestPhotosView.test_delete_a_photo
>

Those failures are due to the new default behaviour of the toolbar (shown
by default, automatically hides after 5 seconds).
Previously it was hidden by default, so the tests would reveal it and then
interact with it. The new behaviour introduces a lot of potential race
conditions on desktop/slow configurations where autopilot sees the toolbar
up, starts moving the cursor towards one of its buttons, but before it
reaches it the toolbar has started automatically hiding and when the cursor
reaches the position where the button was it’s no longer there.
The solution is to first ensure that the toolbar is hidden, then reveal it,
and then click the button (when manually revealed, the toolbar won’t
autohide).
I fixed the most obvious ones already, but others will surface with time,
so we’ll have to fix them as they come.



> Webbrowser: - new failure:
> webbrowser_app.tests.test_title.TestWindowTitle.test_window_title
>

This one happens only on otto (desktop VM), and we can’t seem to reproduce
it locally.
For some weird reason, autopilot finds two instances of QQuickWindow (there
should always be only one) and thus raises an exception because we use
select_single() to retrieve the instance. The browser hasn’t changed since
the saucy release, so it might be a problem in autopilot itself?


Calendar - new failure:
> calendar_app.tests.test_calendar.TestMainView.test_new_event
>

> Rssreader:
> Seems to be failing all tests now on both devices, but was bouncing around
> quite a bit before too. We should try rerunning this and see if it's just
> the previous flakiness of the tests, or if it's really failing all of them
> every time.
>
>
> On Tue, Nov 5, 2013 at 11:20 AM, Didier Roche <didrocks@xxxxxxxxxx> wrote:
>
>> So, we kicked off image 11.
>>
>> Unfortunately, we can't release that image, seems that friends created a
>> lot of system idling issue impacting the image itself. We reverted it.
>> We have hope image 12 is then going to fix this (the imports from debian
>> and other distro uploads doesn't seem to have had any impact).
>>
>> At the same time, we tried to prepare the autopilot 1.4 rollout
>> mentionned yesterday, we released all components with 1.3 before the big
>> jump: all apps, unity8, keyboard…
>> Unfortunately, it seems that the ubuntu-ui-toolkit wasn't releasable this
>> morning (we have some AP tests failing). So everything is delayed (we have
>> a fix now but need to certify it).
>>
>> We have people monitoring this, then, they will flip on
>> autopilot/autopilot-gtk/autopilot-qt and xpathselect 1.4. Thomi and the
>> QA team will work on getting their AP 1.4 compatible branches in. Tomorrow,
>> we'll try to release everything (only containing those changes, so please
>> please, as already told, nobody merge anything on those branches) and be
>> done.
>>
>> Consequently, image #12 will be kicked in later on, as soon as the
>> toolkit makes it in, and will have a lot of refreshed apps as a consequence
>> :)
>>
>> Cheers,
>> Didier
>
>