1155091 - [Flame][Notifications]If we pull notification, but stop moving before it expands, and keep the finger touching the screen, the whole notification view may disappear.

Status: VERIFIED FIXED

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

Description

[Boon]

Attachment: logcat_1107.txt

[1.Description]:
[Flame][v2.2&v3.0][Notifications] When playing a music, If we pull notification, but stop moving before it expands, and keep the finger touching the screen, the whole notification view may disappear.
Attachments: logcat_1107.txt and 1107.mp4
Happen time: 11:07 AM

[2.Testing Steps]: 
1. Launch Music
2. Play a song, and the music controller will appear in Notification column
3. Pull notification, but stop moving before it expands, and keep the finger touching the screen.

[3.Expected Result]: 
The notification view should not disappear.

[4.Actual Result]: 
The whole notification may disappear. 
(Continue to move, the notification appear.)

[5.Reproduction build]: 
Flame 2.2 version(Affected):
Build ID               20150415162504
Gaia Revision          89a94a8b9c9087da916751697a61ba15bc0688c3
Gaia Date              2015-04-15 20:05:38
Gecko Revision         https://hg.mozilla.org/releases/mozilla-b2g37_v2_2/rev/20e92b626530
Gecko Version          37.0
Device Name            flame
Firmware(Release)      4.4.2
Firmware(Incremental)  eng.cltbld.20150415.201243
Firmware Date          Wed Apr 15 20:12:52 EDT 2015
Bootloader             L1TC000118D0

Flame 3.0 version(Affected):
Build ID               20150415160205
Gaia Revision          777d01f4a2c7b41c4b02e3cf87715714ccc0590b
Gaia Date              2015-04-15 17:20:09
Gecko Revision         https://hg.mozilla.org/mozilla-central/rev/53ceefb0e1c8
Gecko Version          40.0a1
Device Name            flame
Firmware(Release)      4.4.2
Firmware(Incremental)  eng.cltbld.20150415.192055
Firmware Date          Wed Apr 15 19:21:04 EDT 2015
Bootloader             L1TC000118D0

[6.Reproduction Frequency]: 
occasionally Recurrence,3/10

[7.TCID]: 
Free Test

Comment 1

[Boon]

Attachment: Bug video: 1107.mp4

Comment 2

[Gregor Wagner [:gwagner]]

Milan, any ideas?

Comment 3

[Milan Sreckovic [:milan] (needinfo for best results)]

Could be related to audio, but in case it's APZ or Silk, lets see what Kats & Mason think.  There are a few E/HWComposer(  208): Non-uniform vsync interval: 10728908850 type messages, not sure if they are expected, and the E/AudioFlinger(  204): no wake lock to update! ones look a bit iffy as well.

Comment 4

[Kartikaya Gupta (email:[e-mail])]

I can look into this, looks like compositor or tiling related. It's easy enough to reproduce with the STR, particularly if the music is paused and you only drag the notification tray down a little bit (into the top third of the screen).

Comment 5

[Kartikaya Gupta (email:[e-mail])]

So actually this looks like a layout issue. The display list being generated is missing some stuff although I'm not entirely sure why. I'll attach stuff in a sec.

Comment 6

[Kartikaya Gupta (email:[e-mail])]

Attachment: Good version

Here is a frame/displaylist/layers dump of the "good" case, where the notification tray is dragged about halfway down the screen and is visible. Note how inside the display-list item 0xaac45c10 there is a bunch of stuff including BackgroundColor and Text items and so on. This is the content of the notification tray.

Comment 7

[Kartikaya Gupta (email:[e-mail])]

Attachment: Bad version

And here is a bad version, where the notification tray is dragged maybe a fifth of the way down the screen, and ends up invisible. Note how the display item at 0xaac43d30 contains nothing but LayerEventRegions and ScrollInfoLayer items, and no actual items that paint anything. What's also odd about this is that between the "before optimization" and "after optimization" display-list dumps, we see the subdocument get painted twice again using basic layers. I'm not really sure what that's all about, or where that ends up in the final layer tree, if anywhere.

Comment 9

[Chris Lord [:cwiiis]]

A duplicate bug (bug 1158926) was granted blocking status, requesting on this one.

kats, would you like me to attempt to find a work-around for this, or are you going to work on the root cause?

Comment 10

[Kartikaya Gupta (email:[e-mail])]

If you can find a workaround that would be great. I wasn't planning on looking into it further since it's more on the layout side of things. If it's blocking and there's nobody on layout who can look into it and there's no workaround then I can probably investigate more but it will be slow going.

Comment 11

[Chris Lord [:cwiiis]]

Ok, I'm going to take this to investigate a work-around.

n?jet to see if there's anyone on layout who can also take a look at the root cause. Even if we find a work-around, there's the possibility that this is causing other similar issues that we haven't filed/discovered yet.

Comment 12

[Wesley Huang [:wesley_huang] (EPM) (NI me)]

triage:
The dup has blocking status, because it is a regression.
https://bugzilla.mozilla.org/show_bug.cgi?id=1158926#c7

Comment 13

[Jet Villegas (inactive)]

I can't reproduce this bug using the STR described (I'm on a trunk build.) 

I assume the regression range from the duped bug still applies?
http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=6ed69c4d2c15&tochange=07479758ab68

Comment 14

[Botond Ballo [:botond]]

I can reproduce the bug intermittently on a trunk build, with Kats' amendments to the STR in comment 4. In particular, the disappearing tends to only happen for me when the notification bar has only been pulled down by a small amount, 100-200 pixels from the top at most.

As for the regression range, I'm not sure. At this point too many things that depend on bug 1127066 have landed to back it out locally and test if that resolves the problem. I could flash the nightly builds back from when it landed, though, and see what the behaviour is for those.

Comment 15

[Kartikaya Gupta (email:[e-mail])]

(In reply to Botond Ballo [:botond] from comment #14)
> As for the regression range, I'm not sure. At this point too many things
> that depend on bug 1127066 have landed to back it out locally and test if
> that resolves the problem. I could flash the nightly builds back from when
> it landed, though, and see what the behaviour is for those.

I don't think that's going to be worth the effort. If there's nobody on the layout side who can look into this and Chris can't come up with a workaround on the Gaia side then I can dig into it more.

Comment 16

[Chris Lord [:cwiiis]]

Working on a work-around... It seems it's the play controls that cause this, which use a custom icon font. If you remove those, everything is fine. If you expand the tray quick enough, you can see them load in as well, it seems the disappearance of the tray coincides with the loading of this font.

Comment 17

[GitHub Autolander Bot]

Attachment: [gaia] Cwiiis:bug1155091-fix-utility-tray-flicker > mozilla-b2g:master

Comment 18

[Chris Lord [:cwiiis]]

Comment on attachment 8601457 [details] [review]
[gaia] Cwiiis:bug1155091-fix-utility-tray-flicker > mozilla-b2g:master

Well, that was pleasantly easy to fix - single-line CSS change.

Comment 19

[Kevin Grandon :kgrandon]

Comment on attachment 8601457 [details] [review]
[gaia] Cwiiis:bug1155091-fix-utility-tray-flicker > mozilla-b2g:master

I don't know this code too well, but this looks simple enough. Thanks!

Comment 20

[GitHub Autolander Bot]

Pull request has landed in master: https://github.com/mozilla-b2g/gaia/commit/494e46c05db3d2af40b31cee58fe89deb5c306be

Comment 21

[Chris Lord [:cwiiis]]

Comment on attachment 8601457 [details] [review]
[gaia] Cwiiis:bug1155091-fix-utility-tray-flicker > mozilla-b2g:master

[Approval Request Comment]
[Bug caused by] (feature/regressing bug #): This bug
[User impact] if declined: Utility tray disappears and reappears while you open it if the music app is loaded
[Testing completed]: Manual testing completed
[Risk to taking this patch] (and alternatives if risky): Low risk, small CSS change - easy to back out if need be.
[String changes made]: None.

Comment 22

[helen（leave from Mozilla）]

Attachment: verified_v3.0.mp4

This bug has been verified as pass on latest build of Flame v3.0 by the STR in Comment 0.

Results: If we stop sliding before nofitifcation view expands, but keep finger touching the screen, the notification view will not disappear.

See attachment: verified_v3.0.mp4
Reproduce rate: 0/5


Device: Flame 3.0 build(Pass)
Build ID               20150506160205
Gaia Revision          426fe6450ab8da92bb473fef12ccb39c6c920dd0
Gaia Date              2015-05-06 08:40:16
Gecko Revision         https://hg.mozilla.org/mozilla-central/rev/5593ac626826
Gecko Version          40.0a1
Device Name            flame
Firmware(Release)      4.4.2
Firmware(Incremental)  eng.cltbld.20150506.193508
Firmware Date          Wed May  6 19:35:21 EDT 2015
Bootloader             L1TC000118D0

Comment 23

[Ryan VanderMeulen [:RyanVM]]

v2.2: https://github.com/mozilla-b2g/gaia/commit/91d8140ced9616f94d7b9d808a1878aea484b257

Comment 24

[helen（leave from Mozilla）]

Attachment: logcat2347.txt

This bug has been failed verified on latest Nightly Flame v2.2 by the STR in Comment 0.

Result:
Pull notification, and stop moving before it expands, then keep another finger touching the screen, sometimes the whole notification will disappear. 
(Continue to move, the notification appear.)

See attachments: v2.2.mp4 and logcat2347.txt
Reproduce rate: 3/10

Device: Flame 2.2 build(Fail)
Build ID               20150507002500
Gaia Revision          83a63e0e6fcc22c6a74b06ef77b88d5049719cad
Gaia Date              2015-05-07 00:17:17
Gecko Revision         https://hg.mozilla.org/releases/mozilla-b2g37_v2_2/rev/118ddfc76b60
Gecko Version          37.0
Device Name            flame
Firmware(Release)      4.4.2
Firmware(Incremental)  eng.cltbld.20150507.041847
Firmware Date          Thu May  7 04:18:59 EDT 2015
Bootloader             L1TC000118D0

Comment 25

[helen（leave from Mozilla）]

Attachment: v2.2.mp4

Comment 26

[Norry.L.F(Leave from Mozilla)]

Hi Helen,

Please make sure patch has landed your test build. could you check your test build again? and check the result again?

Comment 27

[helen（leave from Mozilla）]

Attachment: Last Verified v2.2.mp4

Yes, I have checked the latest build that is newer than comment 24, and the test result is verified to success. 

This bug has been verified as pass on latest build of Flame v2.2 by the STR in Comment 0.

Results:
The notification view will not disappear.

See attachment: verified v2.2.mp4
Reproduce rate: 0/20

Device: Flame 2.2 build(Pass)
Build ID               20150508002501
Gaia Revision          88d3ac2721a5484495c2ed60e4a068945f0de5aa
Gaia Date              2015-05-08 03:36:54
Gecko Revision         https://hg.mozilla.org/releases/mozilla-b2g37_v2_2/rev/8ad16ebe659d
Gecko Version          37.0
Device Name            flame
Firmware(Release)      4.4.2
Firmware(Incremental)  eng.cltbld.20150508.035130
Firmware Date          Fri May  8 03:51:40 EDT 2015
Bootloader             L1TC000118D0