[TOPBLOCKER] Scopes fail to launch when the network stack is not up

[Sergio Schvezov]

I've seen this randomly from time to time on every other boot, but it happens reliably when flight mode is enabled

Steps to reproduce:
- Enable flight mode
- Reboot

Expected result:
- Scopes launch and fail network requests

Actual:
- Black scope screen

Notes:
- Launcher items work and launch fine
- Indicators work fine

In the logs I see something like:
unity8:
Pre-populating scope "unity-scope-nearby"
Pre-populating scope "com.canonical.scopes.photos_photos"
file:///usr/lib/arm-linux-gnueabihf/unity8/qml/Dash/createCardComponent:77:1: QM
L Label: Binding loop detected for property "height"
file:///usr/share/unity8/Dash/CardTool.qml:179:38: Unable to assign [undefined]
to bool
....
invalidateScopeResults: no such scope ' "musicaggregator" '
invalidateScopeResults: no such scope ' "mediascanner-music" '
invalidateScopeResults: no such scope ' "videoaggregator" '
invalidateScopeResults: no such scope ' "mediascanner-video" '

scope-registry:
...
RegistryObject::ScopeProcess::on_process_death(): Process for scope: "clickscope" exited
...

current build number: 129
device name: krillin
channel: ubuntu-touch/ubuntu-rtm/14.09-proposed
last update: 2014-10-26 15:25:59
version version: 129
version ubuntu: 20141026
version device: 20141015-32e0f27
version custom: 1413941794

[Sergio Schvezov (sergiusens)]

These are the scopes I'm running:
gsettings get 'com.canonical.Unity.Dash' 'favorite-scopes'
['scope://com.canonical.scopes.dashboard_dashboard', 'scope://clickscope', 'scope://unity-scope-nearby', 'scope://com.canonical.scopes.photos_photos', 'scope://com.canonical.scopes.fitbit_fitbit']

[David Barth (dbarth)]

On mako, I don't observe the crash, but see only a white / blank scope.

channel: ubuntu-touch/ubuntu-rtm/14.09
last update: 2014-10-17 16:08:28
version version: 5
version ubuntu: 20141017.2
version device: 20141017.2
version custom: mako-1.1

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Paweł Stołowski (stolowski)]

The problem of only a white/blank scope is https://bugs.launchpad.net/ubuntu/+source/unity8/+bug/1382039

[Paweł Stołowski (stolowski)]

I still can't reproduce this exact issue where dash app is missing as per screenshot (not be confused with blank scopes problem that I commented on above). I've tried several times (with flight mode enabled), also reflashed my krillin with image 135 twice (first with my existing old settings and changing my favorites as per Sergio's comment; then also reflashed with --wipe), but no luck.
If you can find any other factor than flight mode that may have impact on it, please let me know.

[Olli Ries (ories)]

unity 8 crash file on #135

[Olli Ries (ories)]

app-launch-desktop crash on #135 w/ silo 13 installed

[Michał Sawicz (saviq)]

The only reason why the app would not *exist* at all in the app stack would be when it's hanging around from a previous session, meaning unity8 crashed (which is currently not difficult to achieve...) when dash was suspended, meaning that it doesn't die when unity8 dies.

[Michał Sawicz (saviq)]

Ah, and this will not happen at all, soon, as the dash won't be suspended, as the scopes infrastructure can't deal with that (items get lost due to UI not running).

[Pete Woods (pete-woods)]

To clarify, the scopes infrastructure does handle being suspended. But if you leave it suspended for over a minute, the queries will time out.

The problem is that the scopes front-end was being suspended at the same time queries were being run for initial surfacing. I don't see how we can handle running queries at startup at the same time as being suspended. :p

[Pete Woods (pete-woods)]

Olli, scopes don't run inside unity8, they run inside unity8-dash, so if the dash is crashing, we'd need that crash file instead.

[Michał Sawicz (saviq)]

Pete, didn't want to put blame (I don't think there is any to put anywhere, in fact), just explain the reasons.

[Michał Sawicz (saviq)]

Olli just confirmed that this is happening for him after a fresh boot, with no unity8 crash in sight, so the dash just does not start.

[Michał Sawicz (saviq)]

↑ that's with silo 13, preventing a crash.

[Pete Woods (pete-woods)]

Sure, I understand about not trying to blame. Just wanted to inject all the information I have on the issue (which is obvs not much).

I personally have chased the rabbit hole of looking at unity8 logs instead of unity8-dash logs a few times ...

[Sergio Schvezov (sergiusens)]

On Wed, Oct 29, 2014 at 1:44 PM, Pawel Stolowski <<email address hidden>
> wrote:

> I still can't reproduce this exact issue where dash app is missing as per
> screenshot (not be confused with blank scopes problem that I commented on
> above). I've tried several times (with flight mode enabled), also reflashed
> my krillin with image 135 twice (first with my existing old settings and
> changing my favorites as per Sergio's comment; then also reflashed with
> --wipe), but no luck.
> If you can find any other factor than flight mode that may have impact on
> it, please let me know.
>

Pawel, I have a bit more setup than that, probably should update the steps
to reproduce:

online-accounts setup:
- u1 (for clicks)
- fitbit
- facebook
- 2 google ones

The fitbit and dashboard scope have access to my fitbit account
Any scope that requires it has access to one of my google accounts.
The photos scope has access to my facebook account.

Additionally, I have 10GB of music on my SD Card (I tried booting with it
removed and still failed)
I tried booting with only the click scope favorited and it still happens.

[Sergio Schvezov (sergiusens)]

On Wed, Oct 29, 2014 at 2:40 PM, Michał Sawicz <email address hidden>
wrote:

> Olli just confirmed that this is happening for him after a fresh boot,
> with no unity8 crash in sight, so the dash just does not start.
>

Yeah, that's what I think I said in the bug; FM doesn't necessarily need to
be on for it to happen, it just proved to happen 100% of the time here; but
all the dogfooders I've run into saw this at some point during the devices
sprint.

[Pat McGowan (pat-mcgowan)]

assigning pete for now

[Olli Ries (ories)]

while I have been able to reproduce this issue reliably on images up to #135 I am not able to reproduce anymore on #139

[Michi Henning (michihenning)]

I'm not seeing it either (on mako, build 293). The logs all look normal. With flight mode on, the scopes that try to access things on the smart scopes server log a bunch of exceptions, saying that they can't get the proxy for those scopes. That's as expected, seeing that the smartscopesproxy can't talk to the SSS to retrieve the list of available scopes.

After turning flight mode back on, things recover nicely and scopes work corretly (requires manual refresh though).

[Sergio Schvezov (sergiusens)]

On Friday, October 31, 2014, Olli Ries <email address hidden> wrote:
> while I have been able to reproduce this issue reliably on images up to
> #135 I am not able to reproduce anymore on #139

Same here on krillin, can't reproduce on r139.

[Pete Woods (pete-woods)]

I've tried rebooting my krillin on image 139 with airplane mode turned on around 30 times this morning, and haven't seen this happen a single time.

[Olli Ries (ories)]

I have not been able to reproduce anymore, krillin #139

[Sergio Schvezov (sergiusens)]

Ideally instead of marking invalid, it should be marked as "Fixed Released"
with a link to the project branch and/or package that fixed it.

On Fri, Oct 31, 2014 at 2:20 PM, Thomas Strehl <email address hidden>
wrote:

> ** Changed in: unity8 (Ubuntu)
> Status: Confirmed => Invalid
>
> ** Changed in: unity-scope-click (Ubuntu)
> Status: Confirmed => Invalid
>
> ** Changed in: unity-scopes-api (Ubuntu)
> Status: New => Invalid
>
> ** Changed in: unity-scope-scopes (Ubuntu)
> Status: Confirmed => Invalid
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1386653
>
> Title:
> [TOPBLOCKER] Scopes fail to launch when the network stack is not up
>
> To manage notifications about this bug go to:
>
> https://bugs.launchpad.net/ubuntu/+source/unity-scope-click/+bug/1386653/+subscriptions
>

[Albert Astals Cid (aacid)]

Ideally yes, but does anyone know really what fixed it?

[Michi Henning (michihenning)]

Doesn't look like it :-)

There could be a million reasons. In the absence of any other information, I don't think we have a choice but to wait and see whether it comes back. I could stare at our code for years without seeing the bug (assuming there is one).

I guess the most important point here is that we need to beef up our integration testing (in unity-scopes-api and elsewhere). We are working on that at the moment…

[kevin gunn (kgunn72)]

the only thing that's gone in recently was the dbus update to clean up dead client obj's