ubuntu-phone team mailing list archive

[Kevin Gunn <kevin.gunn@xxxxxxxxxxxxx>]

>
>
> * The Mir update triggered 3 bugs (seems the first two are related to each
> other) (Kevin):
> - Settings app opens to a blank screen unless given enough time to render
> or the app is touched (https://bugs.launchpad.net/mir/+bug/1294053)
> - Apps are much slower to open (
> https://bugs.launchpad.net/mir/+bug/1294051)
> - Touch has a rotate issue (https://bugs.launchpad.net/mir/+bug/1294048)
> -> We can't revert Mir easily because of the ABI breakage and this can
> creates more problem that it will solve. We asked Kevin to provide a fix
> ASAP and to update their test checklist to ensure that those issues are
> explicitly tested to not regress anymore.
>
>
This has been fixed now. And sorry about that, these bugs were quite subtle
- with manually testing a touch would correct the problem. I've now added a
note in the manual testing to beware of the user temptation to touch if you
don't get a visual response.


>
>
> * random screen hanging up after a while (Kevin):
> https://bugs.launchpad.net/unity-mir/+bug/1290416
>
> So we invested quite a bit of time couple of people near a day and have
not been able to reproduce this. We are taking a "keep an eye out" approach
to this rather than dedicated repro attempts.
Actually, is there any consideration for soak testing? soak meaning keeping
one image on the device for multiple days with some sort of monkey runner
(think randomized never ending AP) ?
this would need to be a lab setup as engineers on teams can't afford to
keep devices unmodified for so long.

* The previous unity8 reliable AP test failure (Kevin)
> -> should be fixed in #240 -> unfortunately, wasn't fixed (still here on
> newer images). However, the real issue was found and a fix is in progress
> with an unity8 landing
> https://bugs.launchpad.net/unity8/+bug/1294121
>
> sorry if this is repeat info but notes from the bug..short story, it was a
CI infra bug.

So, it seems that the root cause was completely somewhere else. Thanks Leo
for finding this out!

The problem was in how the test has been ran on CI. It seems that there IS
a difference between how we run it and how it's ran in smoketesting -
smoketesting doesn't install unity8-autopilot but configure it as if it was
a click package. Therefore, since unity8-autopilot was not installed, all
the required dependencies were not pulled in. And the url-dispatcher test
requires url-dispatcher-tools to be installed.
We fixed this in the infra, so that now the dependency is installed as well.