1017827 - [Tarako] Continually trying to launch e.me apps will cause the device to reboot by itself

Status: RESOLVED WONTFIX

(Firefox OS Graveyard :: Gaia::Everything.me, defect)

Description

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

Attachment: e_me_reboot.log

1. go to homescreen -> games
2. tap on Cupcakes vs Veggies
3. when the app closes, launch it again

Expected: the system doesn't reboot
Actual: the system reboots

Note:
1. most of the e.me games aren't launching
2. Cupcakes vs Veggies might be different in that it seems to try to connect to a server.  Having said that, I can cause a reboot to occur with other games.  It takes longer.

Gaia      94b7ca3030b4b924f0ffe97910bb383fef9e160a
Gecko     https://hg.mozilla.org/releases/mozilla-b2g28_v1_3t/rev/20ff18dbee48
BuildID   20140529014005
Version   28.1
ro.build.version.incremental=eng.cltbld.20140529.053252
ro.build.date=Thu May 29 05:33:02 EDT 2014
Tarako

Comment 1

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

noming because the system itself should not reboot without the user requesting it.

Comment 2

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

Video shows another attempt after a fresh reboot.  It took more than two tries.  You may have to relaunch several more times : http://www.youtube.com/watch?v=3Z8FTSlnblM

Comment 3

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

1.3 seems unaffected as 1.3 seems to be able to launch the app.

Gaia      0ce948e378cab7ed3db20231281dd7ca2eb99779
Gecko     https://hg.mozilla.org/releases/mozilla-b2g28_v1_3/rev/ce0dd450bffb
BuildID   20140528024003
Version   28.0
ro.build.version.incremental=324
ro.build.date=Thu Dec 19 14:04:55 CST 2013
Buri

Comment 4

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

Ran, is it something you can take a quick look? Thanks

Comment 5

[Ran Ben Aharon [:ranbena] EverythingMe]

The apps are launched basically with the simple window.open method.

Seeing that the app closes while webapp attempts to initialize, it looks like an OOM situation caused by the app (which seems quite logical in Tarako devices, limited in RAM).

Perhaps the perf team should be involved.

Comment 6

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

ni? Thinker, someone on your team can take a quick look? Thanks

Comment 7

[Thinker Li [:sinker]]

I don't find relative messages for rebooting in the log, and can not access the video at comment #2.  Could you provide more information? Or give me a hint?

Comment 8

[Ran Ben Aharon [:ranbena] EverythingMe]

In the video, Naoki clicks the Games Collection icon and then some webapp in the results.
The app opens in a browser wrapper, background image for a few seconds, loading indication at the top, then closes abruptly, returning to Games Collection results.
The process is repeated 3 more times with the same result, only at the 4th attempt, the system reboots and the firefox splash screen (animated fox) appears.

Comment 9

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

could we capture another log for thinker?
how easy is it to reproduce this? is it 100% reproducible if the STR is performed repeatedly? Just a matter of time? thanks

Comment 10

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

I will retest with the latest build.  It was 100 % repeatable with the build I had used.

Comment 11

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

Attachment: 20140612220152.zip

I was able to replicate on the userdebug build from today and grabbed the slog off the sdcard.

I rebooted to move the slog into the lastlog after repeating the bug.

Video of a different attempt can be seen here: https://drive.google.com/a/mozilla.com/#folders/0B_0LdM1CVycISTRoZnlMTHdlM2M

Comment 12

[Gregor Wagner [:gwagner]]

Joe, can you do triage here?

Comment 13

[ying.xu]

(In reply to Ran Ben Aharon (Everything.me) from comment #5)
> The apps are launched basically with the simple window.open method.
> 
> Seeing that the app closes while webapp attempts to initialize, it looks
> like an OOM situation caused by the app (which seems quite logical in Tarako
> devices, limited in RAM).
> 
> Perhaps the perf team should be involved.

yes. it's an OOM issue.
From the log, I found the PID of b2g process has changed.

Comment 14

[ying.xu]

I'm afraid most of apps in everything.me can not run properly with tarako.

Can you simply remove them?

Comment 15

[Ran Ben Aharon [:ranbena] EverythingMe]

It's possible (client side).
This should be taken up with Mozilla product team.

Comment 16

[ying.xu]

Any update here?
Or tell us who should we talk to?

Comment 17

[Peter Dolanjski [:pdol]]

Naoki, is this problem limited to games or does it affect the other Smart Collections?

Comment 18

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

It's a general issue not limited to games.  The game Cupcake vs Veggies takes a lot of memory up so it's the easiest to reproduce this on.  Someone may need to go through the list of apps to see which e.me apps would be best suited for the low end device.

Comment 19

[Peter Dolanjski [:pdol]]

Fabrice, any input on how this OOM issue can be tackled?  Disabling key product functionality to work around this issue doesn't sound like a great path forward.

Comment 20

[Marvin Khoo [:Marvin_Khoo]]

it's hard for every OS to prevent OOM especially on such low memory product.
To be frank, the same situation on competitor OS as well (mostly provide a FC dialog to close app or when there is no responding).
 
I do not think we should remove any services provide on Tarako now, but I think we should try to prevent not to let system reboot.

for app that not able to run, current behavior will simply make it return to the Home screen.


BR,
Marvin

Comment 21

[ying.xu]

(In reply to Marvin Khoo [:Marvin_Khoo] from comment #20)
> I do not think we should remove any services provide on Tarako now, but I
> think we should try to prevent not to let system reboot.

we have fixed the issue that b2g was killed and verified that by monkey test.

Comment 22

[[:fabrice] Fabrice Desré]

(In reply to ying.xu from comment #21)
> (In reply to Marvin Khoo [:Marvin_Khoo] from comment #20)
> > I do not think we should remove any services provide on Tarako now, but I
> > think we should try to prevent not to let system reboot.
> 
> we have fixed the issue that b2g was killed and verified that by monkey test.

Was that a fix on you side? Where did that land?

Comment 23

[ying.xu]

(In reply to Fabrice Desré [:fabrice] from comment #22)
> (In reply to ying.xu from comment #21)
> > (In reply to Marvin Khoo [:Marvin_Khoo] from comment #20)
> > > I do not think we should remove any services provide on Tarako now, but I
> > > think we should try to prevent not to let system reboot.
> > 
> > we have fixed the issue that b2g was killed and verified that by monkey test.
> 
> Was that a fix on you side? Where did that land?

yes,in kernel.
kill foreground process first 
and then waiting foreground process exit completely before killing b2g process.

Comment 24

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

seem like partner already fixed it on their side, let me put this to backlog for now and see if there are any fix on gecko side

Comment 25

[Kevin Grandon :kgrandon]

Mass update: Resolve/Wontfix all existing collections bugs as this feature is now removed.

Please re-open or file a new bug if you feel that this bug still exists in master.