Canonical System Image

When an application is suspended or closed it causes a large UI stutter

Bug #1518935 reported by Andrew Hayzen on 2015-11-23

This bug affects 17 people

Affects		Status	Importance	Assigned to
Canonical System Image	Fix Released	High	Michał Sawicz	Canonical System Image 14 "OTA14"
qtmir (Ubuntu)	Invalid	High	Unassigned
ubuntu-app-launch (Ubuntu)	Fix Released	High	Ted Gould

Bug Description

When an application is suspended or closed it causes a large UI stutter.

What happened:
1) Open multiple applications (eg Browser and Music)
2) Switch from Browser to Music
3) Scroll up/down a listview
4) Around 2 seconds after switching from Browser to Music the process is suspended and it can be observed that a UI stutter occurs

What I expected to happen:
At step 4) for there to be no stutter in the UI

Note that this also occurs if you close an application, and happens when switching from an application to the scopes.

$ system-image-cli -i
current build number: 169
device name: mako
channel: ubuntu-touch/rc-proposed/bq-aquaris.en
last update: 2015-11-23 01:08:21
version version: 169
version ubuntu: 20151120.1
version device: 20150911
version custom: mako-1.1

Tags: performance

Related branches

lp:~ted/ubuntu-app-launch/non-blocking-resume

👁 Image
Merged into lp:ubuntu-app-launch at revision 266

Charles Kerr (community): Approve on 2016-11-10

unity-api-1-bot: Approve (continuous-integration) on 2016-11-10

Michał Sawicz: Needs Fixing on 2016-11-10

Diff: 470 lines (+158/-126)

6 files modified

libubuntu-app-launch/application-impl-base.cpp (+97/-75)
libubuntu-app-launch/application-impl-base.h (+15/-6)
libubuntu-app-launch/application-impl-libertine.cpp (+1/-2)
libubuntu-app-launch/registry-impl.cpp (+19/-26)
libubuntu-app-launch/snapd-info.cpp (+1/-1)
tests/libual-cpp-test.cc (+25/-16)

api

👁 Image
Merged into lp:ubuntu-app-launch/touch-vivid at revision 267

Michał Sawicz: Approve on 2016-11-11

Diff: 470 lines (+158/-126)

6 files modified

api

Revision history for this message 👁 Image

Launchpad Janitor (janitor) wrote on 2015-11-23:

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

Changed in unity8 (Ubuntu):
status:	New → Confirmed

Revision history for this message 👁 Image

Michael Zanetti (mzanetti) wrote on 2015-11-23:

This is the relevant debug output while the stuttering occurs.

qtmir.surfaces: MirSurface["webbrowser-app"]::stopFrameDropper()
qtmir.applications: Application::setInternalState - appId= "webbrowser-app" state= SuspendingWaitProcess
qtmir.applications: TaskController::suspend - appId= "webbrowser-app"
qtmir.applications: Application::setInternalState - appId= "webbrowser-app" state= Suspended
qtmir.applications: ApplicationManager::onAppDataChanged: Received "state" update "webbrowser-app"

affects:

unity8 (Ubuntu) → qtmir (Ubuntu)

Michał Sawicz (saviq) on 2015-11-23

Changed in qtmir (Ubuntu):
status:	Confirmed → Triaged
importance:	Undecided → High

Pat McGowan (pat-mcgowan) on 2015-11-23

Changed in canonical-devices-system-image:
assignee:	nobody → Michał Sawicz (saviq)
importance:	Undecided → High
status:	New → Confirmed
milestone:	none → backlog

Revision history for this message 👁 Image

Andrea Bernabei (faenil) wrote on 2016-04-07:

I'm currently seeing many-seconds delay on rc-proposed r308 on Krillin...

Not sure how I got there (yet) but when I switched from one app to others, it was freezing for a few seconds every time...pretty annoying :)

Revision history for this message 👁 Image

Andrea Bernabei (faenil) wrote on 2016-04-07:

In my case, I was continuously switching between Telegram and Twitter apps (post a tweet, then switch, chat a bit, then switch, etc)...could it have something to do with the browser? (since it's also part of the usecase presented by the bug author)

Revision history for this message 👁 Image

Andrew Hayzen (ahayzen) wrote on 2016-04-07:

Agreed, from playing about with this for a while now, anything that is using oxide (webbrowser or webapp) seems to cause a very large and noticeable delay. Although it should be confirmed that other apps are not susceptible to this.

Revision history for this message 👁 Image

Albert Astals Cid (aacid) wrote on 2016-06-22:

Cut the rope (which is also oxied based) is also a big reproducer of this bug when switching from it to somewhere else (in case you need help with reproducing)

Revision history for this message 👁 Image

Michał Sawicz (saviq) wrote on 2016-11-08:

<ted> So I think that the answer is, yeah, we do that syncronously.
We do a dbus call to cgmanager to get the list of pids and try to write their proc value.
If we fail, then we need a setuid helper for things like oxide.
So it could be that spawning that helper is part of what is causing a delay
I'm not sure how to do it another way though.
We're not blocking on it running, the overhead is just starting it.

Changed in qtmir (Ubuntu):
status:	Triaged → Opinion

Daniel van Vugt (vanvugt) on 2016-11-09

tags:

added: performance

Revision history for this message 👁 Image

Launchpad Janitor (janitor) wrote on 2016-11-09:

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

Changed in ubuntu-app-launch (Ubuntu):
status:	New → Confirmed

Michał Sawicz (saviq) on 2016-11-11

Changed in ubuntu-app-launch (Ubuntu):
status:	Confirmed → In Progress
Changed in qtmir (Ubuntu):
status:	Opinion → Invalid
Changed in canonical-devices-system-image:
milestone:	backlog → 14
status:	Confirmed → In Progress
Changed in ubuntu-app-launch (Ubuntu):
assignee:	nobody → Ted Gould (ted)
importance:	Undecided → High