1024893 - [dolphin][v1.4]Entering settings app takes more than 2.2s

Status: RESOLVED FIXED

(Firefox OS Graveyard :: Gaia::Settings, defect, P2)

Description

[Xiaohong.Fan]

Steps to reproduce，
1. open the settings app from homescreen,start recording time
2. when the setting‘s main interface is loaded，stop recording time

Actual result，
Entering  settings app takes more than  2.2s，a little longer than android phone‘s.

Comment 1

[ying.xu]

not 1.3t branch

Comment 2

[James Zhang (Spreadtrum)]

The same as Flame v1.4.
The performance is worse than dolphin android 4.4(1.4s).

Comment 3

[Arthur Chen [:arthurcc][inactive after 6/30]]

James, may I know which language setting are you using? I tried on my Flame with v1.4 and settings app was launched within 1s. Please refer to the video (https://www.youtube.com/watch?v=rIDnafI5-RY)

Comment 4

[James Zhang (Spreadtrum)]

Loop Xiaohong.Fan,
Please attach your analysis here.

Comment 5

[Xiaohong.Fan]

Time-consuming analysis:
1) the app launches from system to settings,it takes an average of 0.9s.
2) from completing preInit  to window.addEventListener('load', function loadSettings() {...})
   it takes an average of 1.1s.
3) LazyLoader.load  takes an average of 0.37s.

Comment 6

[Wayne Chang [:wchang]]

Can we have buri 1.4 numbers here?

Comment 7

[Peipei Cheng (needinfo if you need my action)]

The avg cold load time on Buri v1.4 is about 1.5s. Please see the test results here: 
https://datazilla.mozilla.org/b2g/?branch=v1.4&device=hamachi&range=7&test=cold_load_time&app_list=marketplace,settings&app=marketplace&gaia_rev=c36c0a721ce3d509&gecko_rev=780a128fef4e&plot=median&err_bars=1

I also tested on dolphin, but it seems the performance is better on my device. It's also above 1.5s in average.

Xiaohong, could you please let us know whether you are using user build or not. Did you enable APZ for app?

Comment 8

[Xiaohong.Fan]

I use the user build version, 
it takes about 2.2s to launch the settings app,
I'm sorry,what does APZ mean?

Comment 9

[ying.xu]

(In reply to Xiaohong.Fan from comment #8)

> I'm sorry,what does APZ mean?

async panning and zooming. an optimization for user input response.

on tarako, you can see it long press power button.

But it this case, there was no user input after launching the app. so I think apz won't help here.

Comment 10

[Arthur Chen [:arthurcc][inactive after 6/30]]

Per comment 5 it seems you were referring to the time of loading all scripts. However, in settings app we lazy loaded things to have better perceived performance, which means the screen shows up before loading all script. I was managed to get a dolphin and had tests, the screen showed up within 2s after I clicked on the icon. My build is as the following:

Gaia      15da4fde475db99626c8a47964b12715775cd5c1
Gecko     31e829082914e3ad1833ba81eb46329f5fcb672c
BuildID   20140513061255
Version   30.0
ro.build.version.incremental=89
ro.build.date=Tue May 13 06:09:31 CST 2014

Comment 11

[Sergi Mansilla [:sergi] (Telenor) [PTO]]

Attachment: Github PR

This patch reduces the Settings startup time by more than 1.5 second on Flame by doing the following things:

- Load root list inline during build time, by removing the dependency on the JavaScript chain that does it dynamically. This way we don't have to wait for JavaScript to do that, and the settings list is ready right away. In order to accomplish this, a new attribute data-inline-template has been introduced.

- Removes link elements that are meant for tablet only during the build process, because it spent a lot of time (more than 200ms) executing the media query code in there. In order to accomplish that, a new attribute data-device-type has been introduced.

- Removed unnecessary background-color setting on the html and body elements (see https://bugzilla.mozilla.org/show_bug.cgi?id=1036446). Color is already set on the children elements anyway.

The difference can be seen in the following video: https://www.youtube.com/watch?v=IVHp0w7wW1Q

Comment 12

[Arthur Chen [:arthurcc][inactive after 6/30]]

Comment on attachment 8453623 [details] [review]
Github PR

Thanks for the patch, Sergi. I think we could try to land it first before we have the proper fix as you suggested. As one of the comments I left in github, I would suggest to move the elements to index.html so that we don't need to introduce a new build script configuration. Which also makes the app less rely on the build script.

And as the patch is involved with build script so I would like the owner of the build script have a look on the patch.

Comment 13

[Yuren [:yurenju] (AFK ~ Oct. 12) (volunteer mode)]

Comment on attachment 8453623 [details] [review]
Github PR

r=yurenju if removed inlineTemplate part in webapp-optimize.js according to Arthur's comment on github.

Comment 14

[Sergi Mansilla [:sergi] (Telenor) [PTO]]

Merged at 6ecc793750a99c3fc4477aaa7a601dc1d9b67b86 for master

Comment 15

[Anthony Ricaud (:rik)]

Sergi: Can you open a bug to remove this build time optimisation for media queries? With metrics on how much it affects settings and probably a branch that Gecko engineers can experiment with. Thanks!

Comment 16

[Xiaohong.Fan]

Thanks for the patch

Comment 18

[Ryan VanderMeulen [:RyanVM]]

Needs rebasing for v1.4/v2.0 uplift.

Comment 19

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

Please don't remove the blocking 1.4+ flag.

Comment 20

[Ivan Tsay (:ITsay)]

Hi Sergi,

Could you help to provide rebased patch for v1.4? We need to land this 1.4 blocker. Thanks.

Comment 21

[James Zhang (Spreadtrum)]

Yang, land this WIP patch and verify.

Comment 22

[yang.zhao]

hi,Sergi
  Could you help to merge it to v1.4? Thank you!^_^

Comment 23

[Jason.Liu]

Dear Sergi,
   Can you merge the patch to  V1.4  ?
   
   Happy holidays!

Comment 24

[Sergi Mansilla [:sergi] (Telenor) [PTO]]

Attachment: Version of the patch for 1.4

Patch for uplifting to 1.4

Comment 25

[Ryan VanderMeulen [:RyanVM]]

Please create a v2.0 patch as well? :)

Comment 26

[Arthur Chen [:arthurcc][inactive after 6/30]]

Comment on attachment 8464659 [details] [review]
Version of the patch for 1.4

r=me for the settings part. Flagging Yuren for the build script.

Note that people are encouraged to use master on their tablet. I'm fine with removing settings_large.css if adding the support of multiple devices to the build script is not trivial.

Comment 27

[Yuren [:yurenju] (AFK ~ Oct. 12) (volunteer mode)]

Comment on attachment 8464659 [details] [review]
Version of the patch for 1.4

looks good. r=yurenju

Comment 28

[Sergi Mansilla [:sergi] (Telenor) [PTO]]

Thanks! Merged at d328c9aaf43d955dd45a207e8df4cf781982b3dd

Comment 29

[Ryan VanderMeulen [:RyanVM]]

v2.0: https://github.com/mozilla-b2g/gaia/commit/231d9fc927d3689ca75eee29001faa410a5a83fe

Comment 30

[renfeng.mei]

looks good for me

Comment 31

[Brogan Zumwalt [Inactive]]

New test case needs to be written to address bug.

Comment 32

[KTucker [:KTucker][Inactive 3/4/2016]]

This is done with automation for accurate results. No test case added in moztrap.

Comment 33

[James Zhang (Spreadtrum)]

The time of Firefox v2.0 dolphin is about 3s.

Comment 34

[Jan Jongboom [:janjongboom] (inactive)]

James, this is not what I see actually on the Bangladesh Fly device on v1.4. Time is (manually timed) less than 2s.

Comment 35

[James Zhang (Spreadtrum)]

Hi Jan,

   1.4 is OK, 2.0 is slower than 1.4.