983772 - [Tarako] We should not kill an app just by sending it to the background

Status: RESOLVED DUPLICATE of bug 1006374

(Firefox OS Graveyard :: Runtime, defect, P2)

Description

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

1. load 50 songs onto an SDcard
2. launch music app
3. while the songs are loading, long tap on the home button
4. select the music app

Expected: the Music app is still loading songs
Actual: the Music app is killed.

Gaia      ab75c02ae390da583476053a53951684265d5684
BuildID   20140314165458
Version   28.0
ro.build.version.incremental=110
ro.build.date=Fri Mar 14 16:57:21 CST 2014
Tarako

Comment 1

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

Note: Music app is killed, there's still a preview in the Task Manager that does nothing when you select it.

Comment 2

[Jason Smith [:jsmith]]

Why is this nominated for 1.3? Shouldn't this be nomed for 1.3T? The bug is filed for Tarako.

Comment 3

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

Good question.  I thought I did have it on the 1.3T?  Apparantly, I did not.

Comment 4

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

Check to see if it's device specific or not; also get logcat.

Comment 5

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

Attachment: logcat.txt

Logcat attached; does not occur on 1.3 buri 20140317004001

Comment 6

[[:fabrice] Fabrice Desré]

It may not be a bug because KillBackgroundApps() that appears in that log is from a patch that has not landed on the 1.3t branch. It's unfortunate to let QA spend time with such a frankenbuild :(

Comment 7

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

Attachment: logcat.txt

Gaia      f6a7afd0d716f7a38dfc1808ee0aab3e06ba2f1d
Gecko     fb7149d9b320d939d716011f865ec34661ebc353
BuildID   20140326061134
Version   28.1
ro.build.version.incremental=68
ro.build.date=Wed Mar 26 06:17:50 CST 2014

I can still reproduce this on today's build; it takes a little bit more.
1. Launch music app
2. while it's loading long tap home
3. select music app

Expected: music app comes up
Actual: music app died.

If it doesn't occur the first time you can do step 2 and 3 again and it should die
From the logcat : 03-26 20:12:17.546: I/log(534): <4>0[ 60.656861] select 459 (Music), adj 10, size 8837, to kill

Comment 9

[Joe Cheng [:jcheng] (please needinfo)]

triage: since it's more difficult to reproduce now and tarako is indeed a very limited memory device, not blocking release with this

Comment 10

[Jason Smith [:jsmith]]

(In reply to Joe Cheng [:jcheng] from comment #9)
> triage: since it's more difficult to reproduce now and tarako is indeed a
> very limited memory device, not blocking release with this

I haven't seen any data justifying this. Where is there proof of this analysis?

I'm flagging qawanted here to get a picture of how often this problem happens across the phone.

Comment 11

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

When the device is under memory pressure it can happen for the following apps:
music app
gallery app
Calendar app
Usage app
email app
Video app
Settings app
Contacts app
SMS app

Tested with : 
Gaia      718a06816327fcb6a18095f677cfff4b86adc292
Gecko     https://hg.mozilla.org/releases/mozilla-b2g28_v1_3t/rev/9ef12c19ddc9
BuildID   20140416004007
Version   28.1
ro.build.version.incremental=eng.cltbld.20140416.123359
ro.build.date=Wed Apr 16 12:34:05 EDT 2014
Tarako

Comment 12

[Joe Cheng [:jcheng] (please needinfo)]

let's take a look at this after bug 989572 is done

Comment 13

[Joe Cheng [:jcheng] (please needinfo)]

Naoki, do you mind testing again given that bug 989572 had landed? thanks

Comment 14

[Jason Smith [:jsmith]]

QA Wanted for comment 13

Comment 15

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

This still can occur:
APPLICATION      USER     PID   PPID  VSIZE  RSS     WCHAN    PC         NAME
b2g              root      84    1     143716 41328 ffffffff 400d1518 S /system/b2g/b2g
(Nuwa)           root      332   84    50532  7544  ffffffff 400b2518 S /system/b2g/plugin-container
Homescreen       app_351   351   332   67440  19312 ffffffff 400b2518 S /system/b2g/plugin-container
(Preallocated a  root      511   332   58716  15528 ffffffff 400b2518 S /system/b2g/plugin-container


         NAME PID PPID CPU(s) NICE  USS  PSS  RSS VSIZE OOM_ADJ USER   
            b2g  84    1   35.0    0 30.3 33.8 40.3 144.3       0 root   
         (Nuwa) 332   84    0.9    0  0.4  2.5  7.4  49.3       0 root   
     Homescreen 351  332    4.3    1  7.4 11.2 18.9  65.9       2 app_351
(Preallocated a 511  332    1.0   18  6.2  8.9 15.2  58.3       1 root   

System memory info:

            Total 98.9 MB
     Used - cache 64.3 MB
  B2G procs (PSS) 56.3 MB
    Non-B2G procs  8.0 MB
     Free + cache 34.6 MB
             Free  6.5 MB
            Cache 28.1 MB

Low-memory killer parameters:

  notify_trigger 14336 KB

  oom_adj min_free
        0  4096 KB
        1  5120 KB
        2  6144 KB
        6  7168 KB
        8 16384 KB
       10 20480 KB

04-29 10:40:30.290: I/log(524): <4>0[ 313.409502] lowmem_shrink select 491 (Music), adj 10, size 7090, to kill 
04-29 10:40:30.340: I/Gecko(84): [Parent 84] WARNING: pipe error (122): Connection reset by peer: file ../../../gecko/ipc/chromium/src/chrome/common/ipc_channel_posix.cc, line 446
04-29 10:40:30.360: I/Gecko(84): ###!!! [Parent][MessageChannel] Error: Channel error: cannot send/recv
04-29 10:40:30.410: I/Gecko(84): [Parent 84] WARNING: waitpid failed pid:491 errno:10: file ../../../gecko/ipc/chromium/src/base/process_util_posix.cc, line 254
04-29 10:40:30.740: I/Gecko(511): ###################################### forms.js loaded
04-29 10:40:30.760: I/Gecko(511): ############################### browserElementPanning.js loaded
04-29 10:40:30.790: D/Sensors(84): SensorTMD2771::readEvents()======data_fd=51,count=16
04-29 10:40:30.790: D/Sensors(84): SensorTMD2771::processEvent()=====code=40::value=186 
04-29 10:40:30.790: D/Sensors(84): SensorTMD2771: mPendingEvents[Light].light = 186.000000
04-29 10:40:30.800: I/Gecko(511): ######################## BrowserElementChildPreload.js loaded
04-29 10:40:32.410: I/Gecko(84): [Parent 84] WARNING: waitpid failed pid:491 errno:10: file ../../../gecko/ipc/chromium/src/base/process_util_posix.cc, line 254
04-29 10:40:32.410: I/Gecko(84): [Parent 84] WARNING: Failed to deliver SIGKILL to 491!(3).: file ../../../gecko/ipc/chromium/src/chrome/common/process_watcher_posix_sigchld.cc, line 118

Comment 16

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

Forgot to post which version I'm working with:

Gaia      b5adc5a943d3abbd6ab070a47c847f2c24891cc5
Gecko     https://hg.mozilla.org/releases/mozilla-b2g28_v1_3t/rev/e9890f5d4709
BuildID   20140429014002
Version   28.1
ro.build.version.incremental=eng.cltbld.20140428.052137
ro.build.date=Mon Apr 28 05:21:44 EDT 2014
Tarako

Comment 17

[Joe Cheng [:jcheng] (please needinfo)]

ni? Thinker, any idea on what's next? thanks

Comment 18

[Thinker Li [:sinker]]

This is a relative of bug 944457.  For recently setting, lmk is too aggressive.  We are studying new setting for lmk.

Comment 20

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

This is not a dup as this still occurs.