← Back to team overview

ubuntu-phone team mailing list archive

Re: Landing team 25.02.14

 

Le 26/02/2014 07:17, Ricardo Salveti de Araujo a écrit :
On Tue, Feb 25, 2014 at 2:36 PM, Didier Roche <didrocks@xxxxxxxxxx> wrote:
Hey,

Not a lot to add compared to this morning status email, we still hold the
line on touch landings which are not targeting fixing the current image
issues.

we got image #207 containing:
* the music-app fix for the regression mentioned yesterday, overwriting thus
our revert
* some langpacks refresh
* some python and other platform changes inherited from the rest of the
archive like gcc and libc6

You will notice on http://ci.ubuntu.com/smokeng/trusty/touch/ that the
idling counting is now fixed thanks to Paul, getting us an easier way to
reflect reality.

On the current regressions, here are some refresh:

1. The most common one: StateNotFoundError issue (Julien)
Bill opened
https://bugs.launchpad.net/ubuntu/+source/autopilot/+bug/1284316. I tried to
revert and rerun with some of the most obvious components which can have an
impact without any success. I think we need the autopilot guys to help on
that one and see if they can dig the exact issue (weird that the android 4.4
switch has any side-effect on that one). Julien, can you help getting them
looking at the bug Bill opened (I added as well weather-app as a very easy
way to reproduce those issues)?

2. Unity8 (Michał/Ricardo)
It seems there is a crash, debug symbols on the android side are building,
Michał and Ricardo are looking into those. Seems it's maybe not Android-side
related but a gcc 4.7 vs 4.8 ABI mismatch. Ricardo is trying to disable tls
for more information.
We're still investigating the issue, but what we noticed yesterday:
1 - Missing android system properties hooks, added with latest hybris upload
2 - It's quite hard to reproduce the crash at the first time, but then
it gets way easier (e.g a crash at every 5 starts)
3 - Crash only happens when starting Unity8 (and you have to start it
at least 40 times to get the first crash)
4 - Also happens when building platform-api, location-service and
libdbus-cpp with 4.8
5 - Happens with android TLS disabled as well
6 - Happens with and without system-compositor
7 - It always crashes in the heap, which makes it quite hard to debug
(useless trace)

It's not that critical as it only crashes when starting unity8, and
not that frequently either, but it's good to get it fixed as well.

Yeah, part of the issue is that as we have 50 Unity8 AP tests, we would have at last one or more random crash while testing it.
Basically:
208: 1 crash,
207: 2 crashes,
206: 5 crashes.

(and the only way will be to everyday to launch this test only manually, the CI infra doesn't support just rerunning one test, nor autopilot relaunch for flaky tests when a test is failure)

But I do agree the priority is lower than the first item. Did anybody heard from the QA team on that one?
Cheers,
Didier


Follow ups

References