938556 - [B2G][Helix][Homescreen][Hancheng]Homescreen lose some pages

Status: VERIFIED WORKSFORME

(Firefox OS Graveyard :: Gaia::Homescreen, defect, P1)

Description

[lecky]

User Agent: Mozilla/4.0 (compatible; MSIE 8.0; Windows NT 6.1; Trident/4.0; SLCC2; .NET CLR 2.0.50727; .NET CLR 3.5.30729; .NET CLR 3.0.30729; Media Center PC 6.0; aff-kingsoft-ciba; .NET4.0C; .NET4.0E)

Steps to reproduce:

* Description:
  We normally have 3 pages in homescreen. One is searching page, and the other two pages are for lay apps.
  When the issue occurs, the homescreen shows only two pages, the searching page and one apps page. And we can't enter any apps in homescreen. But we can enter settings app from system notifications.

* Precondition:
  N/A

* Reproduction steps:
  1. Flash a new software.
  2. Boot the device.

* Expected result:
  Homescreen is normal.

* Actual result:
  1. Homescreen lose one page. 
  2. When I touch any apps icon in homescreen, the apps icon show highlight, but it doesn't run.
  3. When I touch the settings icon in system notification, the device enter settings app. But settings app can't be opened from the icon in homescreen.
  4. If I hold the apps icon for seconds, homescreen doesn't show the uninstall page.
  5. If I reboot the device, the homescreen resume normal. 

* Reproduction build:( V1.1.0 hd/2013-10-18)

* Analyze
  1. I want to know how does the homescreen create the layout when first boot. Could someone tell me where is the related code?

  2. When I type b2g-info, it shows:
shell@android:/ $ b2g-info
b2g-info
Fatal error: B2G main process not found.

  3. I typed ps, all the processes are abnormal.
root      1     0     472    332   ffffffff 00000000 S /init
root      2     0     0      0     ffffffff 00000000 S kthreadd
root      3     2     0      0     ffffffff 00000000 S ksoftirqd/0
root      6     2     0      0     ffffffff 00000000 S migration/0
root      10    2     0      0     ffffffff 00000000 S khelper
root      11    2     0      0     ffffffff 00000000 S suspend_sys_syn
root      12    2     0      0     ffffffff 00000000 S suspend
root      13    2     0      0     ffffffff 00000000 S sync_supers
root      14    2     0      0     ffffffff 00000000 S bdi-default
root      15    2     0      0     ffffffff 00000000 S kblockd
root      16    2     0      0     ffffffff 00000000 S khubd
root      17    2     0      0     ffffffff 00000000 S l2cap
root      18    2     0      0     ffffffff 00000000 S a2mp
root      19    2     0      0     ffffffff 00000000 S modem_notifier
system    168   1     884    232   ffffffff 00000000 S /system/bin/servicemanager
root      169   1     6532   792   ffffffff 00000000 S /system/bin/vold
root      171   1     5808   816   ffffffff 00000000 S /system/bin/test_diag
root      172   1     8872   748   ffffffff 00000000 S /system/bin/oem_rpc_svc
nobody    173   1     5996   256   ffffffff 00000000 S /system/bin/rmt_storage
root      174   1     3420   480   ffffffff 00000000 S /system/bin/callife
root      175   1     872    256   ffffffff 00000000 S /system/bin/filebackup
root      179   1     216744 67764 ffffffff 00000000 S /system/b2g/b2g

Comment 1

[lecky]

Attachment: 2013-11-04-12-12-28.png

Comment 2

[lecky]

We found the indexedDB of homescreen(1694510209hnoemeersc.sqlite) is abnormal. When this bug happan, "file" in "main"->"Tables" is empty, and the "object_data" has some issues.

object_data when this issue happen:
id    object_store_id    key_value    file_ids    data
1     1                  01 80        null        D0 0F 60 00 00 00 00 08 00 FF FF 05 00 00 00 04  
2     1                  01 BF F0     null        B0 A0 04 06 00 00 00 00 08 00 FF FF 05 00 00 00
3     1                  01 C0        null        F0 8F 01 60 00 00 00 00 08 00 FF FF 05 00 00 00

object_data in normal device:
id    object_store_id    key_value    file_ids                              data
1     1                  01 80        1 2 3 4 5                             80 1B 60 00 00 00 00 08 00 FF FF 05 00 00 00 04 
2     1                  01 BF F0     6 7 8 9 10 11 12 13 14 15 16 17 18    88 B1 07 60 00 00 00 00 08 00 FF FF 05 00 00 00
3     1                  01 C0        19 20 21 22 23 24                     88 9E 01 60 00 00 00 00 08 00 FF FF 05 00 00 00

Comment 3

[lecky]

Attachment: log-for-bug938556.zip

Comment 4

[lecky]

I have two questions, please help answer.
1. When first power on the phone, how does b2g start Homescreen?
2. When I input "kill ****(pid of Homescreen)" to kill Homescreen, how dows b2g automaticly start Homescreen? Is there any difference between this way and step 1?

We can reproduce this issue by just flash a new software once in 100 times.
But we can't reproduce this issue by step 2 even for 1000 times.

Comment 5

[Wayne Chang [:wchang]]

(In reply to lecky from comment #4)
> I have two questions, please help answer.
> 1. When first power on the phone, how does b2g start Homescreen?
> 2. When I input "kill ****(pid of Homescreen)" to kill Homescreen, how dows
> b2g automaticly start Homescreen? Is there any difference between this way
> and step 1?
> 
> We can reproduce this issue by just flash a new software once in 100 times.
> But we can't reproduce this issue by step 2 even for 1000 times.

1. This information has been provided multiple times over email, please trace starting with init.b2g.rc.

2. Not a blocker on this as low occurrence rate and recovers with device reboot.

3. It seems the problem occurred when you're generating the database for the first time, that's probably the difference between your 2 different steps.

Comment 6

[lecky]

How does Homescreen generating the database? I found configurator.js opened js/init.json. Does the database generated in configurator.js?

Comment 7

[lecky]

Attachment: failed_log.rar

log when failed in homescreen.

Comment 8

[lecky]

Hi all

    I have attached the log when issue happened.

    Please get the source code where I add the log output from Shi Jinde

Br.
Loren

Comment 9

[lecky]

Comment on attachment 8337224 [details]
failed_log.rar

wrong log file

Comment 10

[lecky]

Attachment: failed_log.rar

log when issue happen

Comment 11

[lecky]

Hi All

    Please see the second log package upload at 2013-11-23 17:37 CST, lecky. (One of the log file in the first one is wrong.)

Br.
Loren

Comment 12

[lecky]

Hi Wayne

    I have checked the code when the homescreen init the pages, there are two steps. 1>. It will send message by "xhr.send(null)" in IconRetriever.retrieve for every one of the App's icon in the pages, and then return. 2>.The process that will actually load the image and implement the homescreen pages will be reactived after the xhr response (by xhr.onload) for every App's icon retrieve. Am I right?

    By adding more logs, I can see that the step 1> may stoped somewhere without regularity. And in the step 2>, only some of the pages' element will be implemented, these element was init before the error happen in the step 1>. 

    So I am wonder, is there anything can cause the fuction stop in the middle, in the step 1>, without any error message?

    Do you have any suggestion?

Br.
Loren

Comment 13

[Luke Chang [:lchang] (inactive)]

Hi Lecky,

Per offline discussion, please disable FTU and see if this issue can be fixed first.

Comment 14

[lecky]

Hi All

    After disabling the FTU, the issue could not reproduced during 70 tests. Comparatively, with the build enable FTU base on the same source code, the issue can be reporduced in 5 times. 

Br.
Loren Ni

Comment 15

[Luke Chang [:lchang] (inactive)]

Attachment: 0001-Bug-938556.patch

According to comment 14, this workaround patch tries to prevent FTU and homescreen from launching at the same time.

Lecky, could you please apply this patch and see if this bug can be fixed?

Comment 16

[lecky]

Hi Luke,

Thanks a lot. I test with the patch for 113 times and can't reproduce this issue. It seems this patch works.
I will submit this change to our baseline.

Comment 17

[Luke Chang [:lchang] (inactive)]

William,

Could you please help verify this bug on our build?
I'll close this bug if it can't be reproduced.

Comment 18

[William Hsu [:whsu]]

Hi, Luke,

No problem. I will verify it later.

But, could I confirm a information?
Has this patch been merged into Gaia repository?
Master or V1.1.0hd?
Thanks.

Comment 19

[Luke Chang [:lchang] (inactive)]

This patch is just a workaround and is currently for local patch only since I can't reproduce this bug on the mozilla build. Thus, we won't merge this patch or make further investigation until it is confirmed.

Thanks for your help.

Comment 20

[William Hsu [:whsu]]

Hi, Luke,

I would like to verify it but it is a local build.
As our agreement, we need to prevent to verify a local build unless it is high risk or hotfix patch because it will easy to encounter other problems and hard to identify whether the bug is side effect or not.

I suggest that we can request our partner to verify it because it only happened on partner build.

Needinfo partner.

Sorry for any inconvenience!


------------------------------------------------------------------------

Hi, Lecky,

Could I have your help to verify the patch?
Thanks!

Comment 21

[lecky]

Hi William,

We have verified the patch for 113 times and can't reproduce this issue.
Without the patch, we have reproduced 3 times in 40 tests.

Comment 22

[Luke Chang [:lchang] (inactive)]

Hi William,

Sorry to cause misunderstanding. Let's check whether the original problem will appear on our build or not (without my patch.)

Take your time, thanks.

Comment 23

[William Hsu [:whsu]]

Hi, Luke,

Thanks for the reminder.

I did the test (20 times) on the latest Mozilla V1.1.0hd build but I cannot reproduce it.
I would like to close this bug.

* Test Build: (V1.1.0hd)
 - Gaia:     2d79d8c461262b4d819fba17a0e8cbc03b36b6b1
 - Gecko:    http://hg.mozilla.org/releases/mozilla-b2g18_v1_1_0_hd/rev/94c719e506b1
 - BuildID   20131212042202
 - Version   18.0