832222 - Unexpected white content after page is loaded

Status: RESOLVED DUPLICATE of bug 830256

(Core :: Graphics: Layers, defect)

Description

[Julien Wajsberg [:julienw]]

Attachment: screenshot 1

STR:
* launch a new instance of Browser. It it was previously launched, please kill it first.
* type about:config

Expected:
* the Book of Mozilla excerpt should show it all its glory

Actual:
* part of it is missing (see screenshots, it's sometimes different)
* locking the phone and then unlocking doesn't make it appearing
* it appears as soon as I try to scroll (even if the page is not scrollable)

Comment 1

[Julien Wajsberg [:julienw]]

Attachment: screenshot 2

same STR, different result

Comment 2

[Julien Wajsberg [:julienw]]

Note that I see this in other websites as well, not only on the Boof of Mozilla, so this is not Mammon's revenge.

It's probably not a Browser issue so I put this in Layers instead.

Comment 4

[Julien Wajsberg [:julienw]]

Really I believe this is not the same bug.

In this bug, we see white space _without any interaction_
in Bug 831973 we see white space _with user interaction_.

So I'm reopening, sorry. I'd be happy if a fix to Bug 831973 would fix this bug too but in the mean time I prefer to keep separate bugs.

Comment 5

[Jason Smith [:jsmith]]

(In reply to Julien Wajsberg [:julienw] from comment #4)
> Really I believe this is not the same bug.
> 
> In this bug, we see white space _without any interaction_
> in Bug 831973 we see white space _with user interaction_.
> 
> So I'm reopening, sorry. I'd be happy if a fix to Bug 831973 would fix this
> bug too but in the mean time I prefer to keep separate bugs.

It's literally exactly the same cause. There's a bunch of symptoms all deriving from the same problem and same bug that caused the regression.

Comment 6

[Andrew Overholt [:overholt]]

Dupe is tef+ so doing the same here to ensure we don't miss this if the duplication is broken.

Comment 7

[Julien Wajsberg [:julienw]]

I reopen this now that Bug 831973 is fixed and the bug is still there. As expected.

Comment 8

[Jason Smith [:jsmith]]

(In reply to Julien Wajsberg [:julienw] from comment #7)
> I reopen this now that Bug 831973 is fixed and the bug is still there. As
> expected.

Did you make sure to confirm that bug was fixed with some of the other cases as well? Like marketplace and persona? That would confirm this bug is different.

Comment 9

[Julien Wajsberg [:julienw]]

Yep, I just tried the other use cases and I don't reproduce them.

This bug is different because it doesn't involve scrolling _at all_. That's why I didn't want to dupe it at first.

Actually, scolling _is a workaround_ to this bug.

Comment 10

[Julien Wajsberg [:julienw]]

What's strange is that I can't see this on any real-world page, although I occasionally see this while some are loading.

My theory is that we'll see this bug on pages that are small enough that everything is displayed at the first reflow, which doesn't happen much in real world now.

Comment 11

[Julien Wajsberg [:julienw]]

Another thing that I see (on about:mozilla) is that when the about:mozilla page is already displayed correctly (eg: load it, try to scroll it, and voilà it's displayed correctly), with this STR:

* click on the address bar
* clock on the cross to cancel
=> the page is displayed, then the keyboard disappear, and then the white block appears, and the height of white block + address bar is exactly the same as the height of the keyboard (I verified this with a rule made of paper ;) )

Interestingly enough I seem to reproduce only this now, and not the attached screenshots, so Bug 831973 might have actually changed something on this.

Comment 12

[Julien Wajsberg [:julienw]]

Ok, just reproduced screenshot 2 after all.


I created to remote testcases as I know that everybody does not have the new about page and this made the problem clearer.

STR:
* go to http://everlong.org/mozilla/testcase-white/
* click on "gradient version"
=> everything seems ok
* click on address bar
* click on the cross to cancel
=> screenshot 3

What's interesting here is that with the "no gradient" version, everything is fine. The only difference is that the "gradient version" has a radial gradient background.

Comment 13

[Julien Wajsberg [:julienw]]

Attachment: screenshot 3

Comment 14

[Andrew Overholt [:overholt]]

Triage discussed this today and decided that since this goes away if you touch the screen at all we won't block on it.

Jet, can someone please investigate and re-nom if it's a serious problem that should block?  Thanks.

Comment 15

[Julien Wajsberg [:julienw]]

Actually this doesn't go away if you merely touch the screen, it goes away if you try to scroll.

But I agree that it doesn't block because I can't seem to reproduce this a lot in real-life websites.

Comment 16

[Julien Wajsberg [:julienw]]

Found this on a IRL website: http://play.esviji.com.

This website also uses a radial gradient as background.

So I renominate.

As in comment 12, this appears when you click on address bar and then click on the cross.

The strangest is that we clearly see that the website is first well displayed, and only then the white region appears.

Comment 17

[Andrew Overholt [:overholt]]

Jet, can you find someone to investigate more here?  It would be great if we could figure out what's causing it and how common of an issue it will be.

Comment 18

[Andrew Overholt [:overholt]]

FWIW, play.esviji.com seems to work fine for me on the 2013-01-28 nightly on my unagi.

Comment 19

[Julien Wajsberg [:julienw]]

Andrew> even when you click on the awesome bar, and then click the cross ?

I still see this with a custom build from today, on the same website.

Comment 20

[Andrew Overholt [:overholt]]

(In reply to Julien Wajsberg [:julienw] from comment #19)
> Andrew> even when you click on the awesome bar, and then click the cross ?

I don't know what you mean.  I enter the URL and I'm given the game name with the coloured background circles.  There are 3 buttons:  "Start", "Tutorial", "Install".  I press "Start" and the game works as I'd expect it to (having never played it before).

Comment 21

[Julien Wajsberg [:julienw]]

Andrew> when you arrive on the page, everything is ok.
The bug happens if you click on the address bar of the browser, and then click the cross to escape the screen (still in the browser, this has nothing to do with the app).

This is the same with the testcase http://everlong.org/mozilla/testcase-white/gradient.html

Comment 22

[Anthony Jones (:ajones, :kentuckyfriedtakahe, :k17e)]

The problem is related to the resizing that happens when the keyboard goes away. The event ordering out of sequence. I've got a fix but I want to investigate further to make sure I'm not trading one race condition for another.

Comment 24

[Lukas Blakk [:lsblakk] use ?needinfo]

Batch edit: Bugs still affected on b2g18 after 2/13 merge to v1.0.1 branch are affected on v1.0.1 branch.