815593 - large image within div with overflow-y: scroll disappears for a few seconds when scrolling

Status: VERIFIED FIXED

(Core :: Graphics, defect)

Description

[zac]

User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:16.0) Gecko/20100101 Firefox/16.0
Build ID: 20121024073032

Steps to reproduce:

I put an image within a div and made the div scrollable using overflow-y: scroll
You can see an example here: http://www.serlo.org/math/exams/show/deutschland/bayern/mittlere-reife/2011/mathematik-i-haupttermin
or here: http://www.serlo.org/math/exams/show/deutschland/bayern/mittlere-reife/2011/mathematik-ii-haupttermin



Actual results:

When scrolling down inside the div, the image disappears for about 2-5 seconds.


Expected results:

The image should not disappear (like here: http://www.serlo.org/math/exams/show/deutschland/bayern/qualifizierender-hauptschulabschluss/2009/teil-b-aufgabengruppe-ii )

Comment 1

[Alice0775 White]

Attachment: reduced html (red image)

I can reproduce with HWA on in Firefox15.0.1 and later.
Deiabled HWA helps.

Regression window(m-c)
Good:
http://hg.mozilla.org/mozilla-central/rev/2db9df42823d
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:15.0) Gecko/15.0 Firefox/15.0a1 ID:20120503142649
Bad:
http://hg.mozilla.org/mozilla-central/rev/e1a40027dc7e
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:15.0) Gecko/15.0 Firefox/15.0a1 ID:20120504014349
Pushlog:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=2db9df42823d&tochange=e1a40027dc7e

Regression window(m-i)
Good:
http://hg.mozilla.org/integration/mozilla-inbound/rev/c608de1b6a53
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:15.0) Gecko/15.0 Firefox/15.0a1 ID:20120502221102
Bad:
http://hg.mozilla.org/integration/mozilla-inbound/rev/77bf50b33a05
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:15.0) Gecko/15.0 Firefox/15.0a1 ID:20120502223302
Pushlog:
http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=c608de1b6a53&tochange=77bf50b33a05
Regressed by: Bug 716439

Comment 2

[Benoit Girard (:BenWa)]

(In reply to zac from comment #0)
> Actual results:
> 
> When scrolling down inside the div, the image disappears for about 2-5
> seconds.
> 

Those few seconds are while the layer is active, afterwards it gets flatten to the background.

Comment 3

[Nick Cameron [:nrc]]

I am surprised that this would be regressed by bug 716439 - it should only affect elements with a border-radius. If it is an active layers problem, you can force active layers for testing by setting the pref layers.force-active = true in recent nightlies.

I'm happy to add this to my list, but I probably won't get to it for a while.

Comment 4

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

I think the problem is that the image is extremely tall so doesn't fit in a texture when we try to make an ImageLayer for it.

Comment 5

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

Attachment: fix

Comment 6

[Lukas Blakk [:lsblakk] use ?needinfo]

Has UUID change in patch, won't be taking this on the 17.0.1 respin.

Comment 7

[Alex Keybl [:akeybl]]

roc - is there any way we can prevent the UUID change? Or are we confident this will not have add-on compatibility impact?

Comment 8

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

I can write a different hack patch that doesn't require a UUID change for Aurora/Beta.

Comment 9

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

Attachment: Hack patch that doesn't require UUID changes

Comment 10

[Matt Woodrow (:mattwoodrow)]

Comment on attachment 686417 [details] [diff] [review]
Hack patch that doesn't require UUID changes

Review of attachment 686417 [details] [diff] [review]:
-----------------------------------------------------------------

::: image/src/RasterImage.cpp
@@ +980,5 @@
>  NS_IMETHODIMP
>  RasterImage::GetImageContainer(ImageContainer **_retval)
>  {
> +  // Limit size to avoid overflowing texture bounds.
> +  if (mSize.width > 2000 || mSize.height > 2000) {

2048 would make me feel happier, though it doesn't really matter at all.

::: layout/reftests/bugs/reftest.list
@@ +1721,5 @@
>  == 776265-2d.html 776265-2-ref.html
>  == 787947-1.html 787947-1-ref.html
>  fuzzy(40,800) == 797797-1.html 797797-1-ref.html # 'opacity:N' and rgba(,,,N) text don't match precisely
>  fuzzy(40,800) == 797797-2.html 797797-2-ref.html # 'opacity:N' and rgba(,,,N) text don't match precisely
> +== 815593-1.html 815593-1-ref.html

Looks like the actual files are missing from this patch.

Comment 11

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

https://hg.mozilla.org/integration/mozilla-inbound/rev/a4f64be02d63

Comment 12

[Matt Woodrow (:mattwoodrow)]

Pushed a follow-up to remove an unused variable since this failed to build on some platforms.

https://hg.mozilla.org/integration/mozilla-inbound/rev/445dc074e609

Comment 13

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/a4f64be02d63
https://hg.mozilla.org/mozilla-central/rev/445dc074e609

Comment 14

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/a4f64be02d63
https://hg.mozilla.org/mozilla-central/rev/445dc074e609

Comment 15

[Alex Keybl [:akeybl]]

Given the fact that this is reproducible since Firefox 15, we'll approve based upon a risk evaluation - not clear that this is any worse in FF17/18 so no need to track.

Comment 16

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

Given there are no dups and the age of this bug, I think it's not worth trying to uplift this to beta (where there is already a ton of risk). Let's just land the original UUID-changing patch on Aurora.

Comment 17

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

Comment on attachment 685928 [details] [diff] [review]
fix

[Approval Request Comment]
Bug caused by (feature/regressing bug #): unsure
User impact if declined: Image missing in Web page in some (apparently rare) cases
Testing completed (on m-c, etc.): a few days on m-c
Risk to taking this patch (and alternatives if risky): this is pretty low risk. Ultimately we're just disabling an optimization in some cases.
String or UUID changes made by this patch: UUID change.

Comment 18

[Lukas Blakk [:lsblakk] use ?needinfo]

Comment on attachment 685928 [details] [diff] [review]
fix

low impact, let's go ahead with Aurora and at least get a few more eyes on it as it rides up the trains.

Comment 19

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-aurora/rev/0d4beedf0c58
https://hg.mozilla.org/releases/mozilla-aurora/rev/c6df5983f54a

Comment 21

[feinzeig]

I'm seeing an issue that seems like exactly this bug, but when trying the beta and aurora builds I'm not seeing any difference (from this ticket, it seemed like a patch was applied to at least aurora).

If you look at:

http://killinglincolnconspiracy.com

Scroll down and click the arrow to open the right side panel.  The background image and other images turn black for a couple of seconds and then redraw.

Same issue?  Different issue?

Thanks!

Comment 22

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

I suspect that's a different issue, related to image decoding. Please file a bug for it!

Comment 23

[feinzeig]

I did a little more playing with the issue, and I do think it's related.  If I remove the javascript that adds or removes a noscroll class, which sets overflow-y: none or auto, when expanding that side panel and other things, the issue goes away.  Definitely seems like the same issue to me.

Comment 24

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

(In reply to feinzeig from comment #23)
> I did a little more playing with the issue, and I do think it's related.  If
> I remove the javascript that adds or removes a noscroll class, which sets
> overflow-y: none or auto, when expanding that side panel and other things,
> the issue goes away.  Definitely seems like the same issue to me.

You mean overflow-y:visible? That still sounds like an image decoding issue due to frame reconstruction, to me.

Comment 25

[Bogdan Maris, Desktop Test Engineering]

Mozilla/5.0 (Windows NT 6.1; WOW64; rv:19.0) Gecko/20100101 Firefox/19.0

Verified as fixed on Firefox 19 beta 6 (buildID: 20130212082553), latest Nightly (buildID: 20130213031137) and latest Aurora (buildID: 20130212042017) using the provided image from comment 1.

Comment 26

[Bogdan Maris, Desktop Test Engineering]

Mozilla/5.0 (Windows NT 6.1; WOW64; rv:20.0) Gecko/20100101 Firefox/20.0

Verified as fixed on Firefox 20 beta 5 (buildID: 20130313170052).