728026 - MAPLE: geolocation shows little map fragments while the map is being loaded

Status: VERIFIED FIXED

(Core :: Graphics, defect, P2)

Description

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

1. go to http://paulrouget.com/mwc-demos/apis/geolocation.html
2. click on the icon
3. watch it load

Expected: no graphic defect while loading
Actual: you can see little map fragments while the map is being loaded

Note:
1. MAPLE 02/16/2012; Nexus Galaxy ICS/LG Revolution 2.3
2. Does not occur on nightly 02/16/2012
3. video : http://www.youtube.com/watch?v=8wvCU7IbcmU&feature=youtube_gdata_player

Comment 1

[Erin Lancaster [:elan]]

We are not showing this on Maple; only on nightly

Comment 2

[Paul Rouget [:paul]]

The APIs demo are located here: http://paulrouget.com/mwc-demos/apis/
To show the map, click on the globe.

Comment 3

[Paul Rouget [:paul]]

The original demo is here: http://jsbin.com/isujul

Comment 4

[Joe Drew (not getting mail)]

Oh wait, I misunderstood Erin's comment.

Comment 5

[Joe Drew (not getting mail)]

This doesn't show up on the main demo site, just on http://jsbin.com/isujul

Comment 6

[Ali Juma [:ajuma]]

(In reply to Paul Rouget [:paul] from comment #3)
> The original demo is here: http://jsbin.com/isujul

It doesn't seem to be anymore. Can the original demo be found anywhere else?

Comment 7

[Joe Drew (not getting mail)]

This is a very, very bad bug that indicates we're not clipping layers properly, and it's not restricted to this page (I've seen it elsewhere). I'd like it to be re-evaluated.

Meanwhile, I've put this demo up on my personal server. Mats, can you see what's causing this?

Comment 8

[Brad Lassey [:blassey] (use needinfo?)]

solid justification to block on it

Comment 9

[Jet Villegas (inactive)]

Mats, can you comment on your latest findings here? Thx.

Comment 10

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

This issue still occurs on the Galaxy Nexus (Nightly 3/30/2012) going to http://paulrouget.com/mwc-demos/apis/ and clicking on the "view on a map" button.  If you do not see the issue on load, try panning in the map area.

Comment 11

[Erin Lancaster [:elan]]

Fennec Native needs to be at zero beta blockers by 04/16. Who should this be assigned to?

Comment 12

[Brad Lassey [:blassey] (use needinfo?)]

sorry,  moving the product and component apparently unassigned the bug. Mats, has this been on your radar?

Comment 13

[Mats Palmgren (inactive)]

I looked at this a couple of weeks ago, and I could reproduce it at that time
but I can't say I understood what the problem is.  I've been working on other
stuff since then (security bugs mostly).  I'll take another look later tonight
and see what I can find this time...

Comment 14

[JP Rosevear [:jpr]]

Roc, do you agree with #7.

Comment 15

[Robert O'Callahan (:roc) (email my personal email if necessary)]

I don't know. Nightly doesn't really work on my phone recently. (Maybe since I enabled Sync?)

I don't have time to work on this anyway. I'm working on display-list performance instead.

Comment 16

[Jet Villegas (inactive)]

Back to Mats: Please update with your latest findings.

Comment 17

[Mats Palmgren (inactive)]

I have nothing new to add really, the bug still occurs and I don't know what
the problem is.

Comment 18

[Mats Palmgren (inactive)]

The Google maps content is rather complex, it's many nested levels of abs.pos.
blocks with different z-index and the whole DOM is built dynamically.

The root cause of this bug appears to come from a transition of 'opacity':
<div style="width: 256px; height: 256px; position: absolute; left: -48px; top: 23px; opacity: 1; -moz-transition: opacity 200ms ease-out 0s;">

As a workaround for this particular demo site, adding this cures the flicker:
  <style> div { -moz-transition: none !important; } </style>

(FYI, bug 728983 is fixed in this tree, but it didn't help)

Comment 19

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Transitioning opacity creates a layer for the element with opacity. Seems like a reduced testcase is needed...

Comment 20

[Mats Palmgren (inactive)]

Attachment: Testcase #1

Comment 21

[Damon Sicore (:damons)]

Do we have a path forward here or are we stuck?

Mats, these Firefox on Android bugs are top priority and require immediate and persistent attention until resolution.

Comment 22

[Mats Palmgren (inactive)]

The bug doesn't occur when useIntermediateSurface is true in this method:
http://mxr.mozilla.org/mozilla-central/source/gfx/layers/Layers.cpp#457

The bug doesn't occur with this change:
-  if (opacity != 1.0f && HasMultipleChildren()) {
+  if (opacity != 1.0f) {


=> GFX

Comment 23

[Joe Drew (not getting mail)]

I'm going to look at this now.

Comment 24

[Joe Drew (not getting mail)]

It seems that this is being caused by the scissor bit being disabled on us.

Comment 25

[Joe Drew (not getting mail)]

Attachment: don't disable the scissor rect from java

This fixes the bug that we haven't been clipping at all inside the layers system. Unfortunately, this exposes latent bugs in our layers/layout system that must be fixed before we can land this.

Comment 26

[Jeff Muizelaar [:jrmuizel]]

Comment on attachment 618823 [details] [diff] [review]
don't disable the scissor rect from java

Perhaps add a comment about the expected scissor state in the different situations. e.g. at drawBackground/drawForeground

Comment 27

[Joe Drew (not getting mail)]

Attachment: unbitrotted

This is just an unbitrotted version of the previous patch. Carrying forward r+.

Comment 28

[Joe Drew (not getting mail)]

Attachment: unbitrotted again

Re-unbitrotted. Carrying forward r+.

Comment 29

[Joe Drew (not getting mail)]

Attachment: unbitrotted again

Comment 34

[Joe Drew (not getting mail)]

https://hg.mozilla.org/mozilla-central/rev/47b65f49d01b

Comment 35

[Tony Chung [:tchung]]

Tested against a build on ftp://ftp.mozilla.org/pub/mobile/tinderbox-builds/oak-android/1336516588/.   I still see the map fragments from the testcase URL quickly resize and redraw, but it doesnt do that outside of the box perimeter like how aurora does.

Comment 37

[Joe Drew (not getting mail)]

The quickly resize part is bug 750334.

Comment 38

[Mark Finkle (:mfinkle) (use needinfo?)]

https://hg.mozilla.org/releases/mozilla-aurora/rev/5fe40d328f31

Comment 39

[Catalin Suciu [:csuciu]]

Retested on Nightly 15.0a1 (2012-05-22)
            Aurora 14.0a2 (2012-05-22)
Samsung Galaxy SII (2.3.4)

The map fragments from the testcase are not longer redrawn outside the box perimeter.

Verifying.