981278 - Chrome process paints unnecessarily while launching an application

Status: RESOLVED WONTFIX

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

Description

[Ting-Yu Chou [:ting] (away)]

When I worked on bug 974308 with b2g v1.3t, I noticed chrome process enters nsViewManager::ProcessPendingUpdatesForView() several times while launching an application even status bar has nothing changed. This makes application launching longer especially when it is low on memory.

Can see status bar flashing after enable "Flash repainted area".

Comment 1

[Ting-Yu Chou [:ting] (away)]

Attachment: InvalidateFrameSubtree_backtraces

I captured 2 backtraces from chrome process which hit nsIFrame::InvalidateFrameSubtree() while launching application.

Comment 2

[[:fabrice] Fabrice Desré]

Great finding Ting-Yu! We enter DoProcessRestyles() that may due to some style change in the system app.

Please cc dbaron if you need help on the css side.

Comment 3

[Ting-Yu Chou [:ting] (away)]

Tapping an application icon on Tarako:

  GeckoInputDispatcher::notifyMotion action=10(HOVER_EXIT) flags=0 count=1
  GeckoInputDispatcher::notifyMotion action=0(DOWN) flags=0 count=1
  nsAppShell::sendMouseEvent() msg=1128645728 x=0 y=0 forwardToChildren=1
  nsAppShell::sendMouseEvent() msg=302 x=293 y=66 forwardToChildren=0
  GeckoInputDispatcher::notifyMotion() action=2(MOVE) flags=0 count=1
  GeckoInputDispatcher::notifyMotion() action=2(MOVE) flags=0 count=1
  ...
  ...
  GeckoInputDispatcher::notifyMotion() action=1(UP) flags=0 count=1
  GeckoInputDispatcher::notifyMotion() action=9(HOVER_ENTER) flags=0 count=1
  GeckoInputDispatcher::notifyMotion() action=7(HOVER_MOVE) flags=0 count=1
  nsAppShell::sendMouseEvent() msg=301 x=294 y=66 forwardToChildren=0
  nsAppShell::sendMouseEvent() msg=1128645728 x=0 y=0 forwardToChildren=1
  nsAppShell::sendMouseEvent() msg=300 x=0 y=0 forwardToChildren=1

Am trying to understand why not forward DOWN/UP to children (Homescreen).

Comment 4

[Ting-Yu Chou [:ting] (away)]

(In reply to Fabrice Desré [:fabrice] from comment #2)
> Great finding Ting-Yu! We enter DoProcessRestyles() that may due to some
> style change in the system app.
> 
> Please cc dbaron if you need help on the css side.

Ok, thanks for reminding :)

Comment 5

[Ting-Yu Chou [:ting] (away)]

(In reply to Ting-Yu Chou [:ting] from comment #3)
>   GeckoInputDispatcher::notifyMotion action=0(DOWN) flags=0 count=1
>   nsAppShell::sendMouseEvent() msg=302 x=293 y=66 forwardToChildren=0

forwardToChildren == 0 since sendTouchEvent() passes the touch event to Homescreen and returns nsEventStatus_eConsumeNoDefault before calling sendMouseEvent(), this seems related to spec: http://goo.gl/3B5Wbq.

Not sure why the motion event is processed by both Homescreen and chrome process.

Comment 6

[Ting-Yu Chou [:ting] (away)]

Attachment: b2g_motion_events_handling

Trying to understand how the motion events are handled by PresShell::HandleEventInternal() in b2g process. Use ">" and "<" to indicate entering/exiting a function.

Comment 7

[Ting-Yu Chou [:ting] (away)]

I saw the same thing while phone ringing, b2g process paints several times prior Communications.

Comment 8

[BMO Automation]

Firefox OS is not being worked on