1041963 - [tarako][v1.3t]sometimes homescreen showed a full-black wallpaper after switching camera to background

Status: RESOLVED FIXED

(Firefox OS Graveyard :: Gaia::Homescreen, defect)

Description

[ying.xu]

STR:
1.keep as many apps at background at possible, then launching camera ;
2. take some photos, then press home button to get back homescreen;
3. sometimes there showed a full-black wallpaper.

and it can get recovered by long pressing home button

Comment 1

[Keven Kuo [:kkuo]]

Hi! Evelyn,

Could you take a look first?
Thnks

--
Keven

Comment 2

[Danny Liang [:dliang]]

I can reproduce it by switch camera and homescreen under background music, but it have fail rate, not 100% reproduced. When blackscreen happened, the memory pressure event was triggered, we suspected it related to bug 1022586 and there might be some racing conditions. 

Re-launch homescreen can sometimes recover the problem but sometimes cannot.

Comment 3

[James Zhang (Spreadtrum)]

Hi Danny, I can see this issue on dolphin 1.4 too.

Comment 4

[Danny Liang [:dliang]]

After revert below commit, I cannot reproduce it.

commit 4ab1432ed292b306cf52d005a2b312ae5bc95d66
Author: ying.xu <[e-mail]>
Date:   Mon Jun 16 17:10:19 2014 +0800

    Bug#1022586 kill the homescreen app when memory low
    
    [self test   ] kill and restore ok

Comment 5

[Danny Liang [:dliang]]

Attachment: WIP-homescreen-wont-be-killed-by-memory-pressure-event.patch

WIP patch for homescreen won't be killed due to memory pressure event

Comment 6

[Wayne Chang [:wchang]]

Blocking device launch 

James, please arrange your QA to test in parallel. Our testing resources are limited.

Comment 7

[ying.xu]

Comment on attachment 8460834 [details] [diff] [review]
WIP-homescreen-wont-be-killed-by-memory-pressure-event.patch

Review of attachment 8460834 [details] [diff] [review]:
-----------------------------------------------------------------

There may be still some issues of displaying abnormal of homescreen.

we found them after monkey-test.

If it could not be killed forever,  this could be a problem.

Comment 8

[(inactive after 6/18) Alive Kuo [:alive]]

Comment on attachment 8460834 [details] [diff] [review]
WIP-homescreen-wont-be-killed-by-memory-pressure-event.patch

Review of attachment 8460834 [details] [diff] [review]:
-----------------------------------------------------------------

Well, it seems we are back and forth many times. Please note someday there might be another issue: homescreen is always residing in the memory.

Comment 9

[Danny Liang [:dliang]]

(In reply to ying.xu from comment #7)
> Comment on attachment 8460834 [details] [diff] [review]
> WIP-homescreen-wont-be-killed-by-memory-pressure-event.patch
> 
> Review of attachment 8460834 [details] [diff] [review]:
> -----------------------------------------------------------------
> 
> There may be still some issues of displaying abnormal of homescreen.
> 
> we found them after monkey-test.
> 
> If it could not be killed forever,  this could be a problem.

In my understanding, homescreen oop will be killed when cache is lower than 16MB by old setting.
To improve launch time of homescreen, we let homescreen in process.
If homescreen is killed by memory pressure event ( will trigger at 14MB ), why don't we just set LMK parameter as 14MB? It's easily and there should be more safe.

In my opinion, I recommend to revert the commit of bug 1022586 then someone can help to find the root cause, why does it show black screen/wallpaper when homescreen was killed?

Hi Evelyn, could you help to check why does it show black screen/wallpaper when homescreen was killed?

Comment 10

[James Zhang (Spreadtrum)]

monkey test running now.

Comment 11

[Danny Liang [:dliang]]

(In reply to James Zhang (Spreadtrum) from comment #10)
> monkey test running now.

James, please notice the black screen issue is easily reproduced within 10 minutes by switching camera and homescreen under background music playback.

Comment 12

[James Zhang (Spreadtrum)]

revert Bug#1022586 change.
https://github.com/mozilla-b2g/gaia/commit/ab87f1ced984ea95f4fde1b02b4b942cb033b8dc

Comment 13

[Lawrence Mandel [:lmandel] (use needinfo)]

Corrected the blocking flag to 1.3T+ as per Ryan's request.

Comment 14

[Ryan VanderMeulen [:RyanVM]]

So comment 3 says this affects 1.4 as well. What about other branches? Per our standard branch landing policy [1], these questions should have been answered prior to landing on a release branch.

Assuming this is a fix we want on 1.4, please at least request approval for that as well.

[1] https://wiki.mozilla.org/Release_Management/B2G_Landing#Landing_Procedure_2

Comment 15

[Danny Liang [:dliang]]

(In reply to Ryan VanderMeulen [:RyanVM UTC-4] from comment #14)
> So comment 3 says this affects 1.4 as well. What about other branches? Per
> our standard branch landing policy [1], these questions should have been
> answered prior to landing on a release branch.
> 
> Assuming this is a fix we want on 1.4, please at least request approval for
> that as well.
> 
> [1]
> https://wiki.mozilla.org/Release_Management/B2G_Landing#Landing_Procedure_2

The solution only works on 1.3T. Homescreen is in-process on 1.3T and we let homescreen won't be killed by memory pressure event. On other branches, homescreen is oop and will be killed by LMK, so that's different.
I will create another bug for tracking issue on master.