1459441 - Firefox 61 breaks Fancybox

Status: VERIFIED FIXED

(Core :: Web Painting, defect, P2)

Description

[Felix Eckhofer]

Attachment: Video showing the correct behavior in FF59 and the regression in FF61

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:61.0) Gecko/20100101 Firefox/61.0
Build ID: 20180503152818

Steps to reproduce:

1. Upgrade Firefox 61.0b1 (happens on 61.0b2 and Nightly 61.0a1 as well)
2. Open https://fancyapps.com/fancybox/3/ and click on a lightbox-enabled image


Actual results:

The lightbox jumps between foreground and background (see video)


Expected results:

The container for the lightbox should stay on top (it has a z-index of 99992 set in the CSS). This used to work up to including FF 60.

Comment 1

[Felix Eckhofer]

Attachment: Same problem on netflix

This is not a Fancybox-exclusive problem as well. I'm seeing the same behaviour on Netflix where things that should be in the background suddenly are visible (the "Dany Boon" poster in the screenshot).

Comment 2

[Gingerbread Man]

Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:61.0) Gecko/20100101 Firefox/61.0
20180505220904

The regression range from mozregression-gui doesn't make much sense to me. I don't see how a backout of developer tools code could cause this. Yet I double-checked and got the same result.

https://hg.mozilla.org/integration/autoland/json-pushes?changeset=d112cf7b2b60e6244099dc3b599a2444ba0d1da3&full=1

Comment 3

[Gingerbread Man]

And now I've triple-checked. I manually downloaded the "first with" and "last without" from https://hg.mozilla.org/mozilla-central/rev/d112cf7b2b60 and tested each in a brand new profile.

https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=00bdc9451be6557ccce1492b9b966d4435615380&tochange=9a2eac450781b026b42c44ca8f0f92bb0846b6e2

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

(In reply to Gingerbread Man from comment #3)
> And now I've triple-checked. I manually downloaded the "first with" and
> "last without" from https://hg.mozilla.org/mozilla-central/rev/d112cf7b2b60
> and tested each in a brand new profile.

What were the URLs and changesets for those two builds?

Comment 5

[Andreea Pavel [:apavel]]

(In reply to David Baron :dbaron: ⌚️UTC-7 from comment #4)
> (In reply to Gingerbread Man from comment #3)
> > And now I've triple-checked. I manually downloaded the "first with" and
> > "last without" from https://hg.mozilla.org/mozilla-central/rev/d112cf7b2b60
> > and tested each in a brand new profile.
> 
> What were the URLs and changesets for those two builds?

Is this what you are looking for https://bugzilla.mozilla.org/show_bug.cgi?id=1450360#c16 ?

Also, please read https://bugzilla.mozilla.org/show_bug.cgi?id=1450360#c22 i guess it might be helpful.
Thanks.

Comment 6

[Gingerbread Man]

(In reply to David Baron :dbaron: ⌚️UTC-7 from comment #4)
> (In reply to Gingerbread Man from comment #3)
> > And now I've triple-checked. I manually downloaded the "first with" and
> > "last without" from https://hg.mozilla.org/mozilla-central/rev/d112cf7b2b60
> > and tested each in a brand new profile.
> 
> What were the URLs and changesets for those two builds?

The changesets are already listed at the link you quoted, aren't they? It says 00bdc9451be6 last without, ff0efa4132f0 first with. If that's not it, where should I look? Pushlog at comment 3 is based on the about:buildconfig links.

Last without
https://archive.mozilla.org/pub/firefox/nightly/2018/04/2018-04-03-22-00-40-mozilla-central/firefox-61.0a1.en-US.win64.zip

First with
https://archive.mozilla.org/pub/firefox/nightly/2018/04/2018-04-04-10-01-27-mozilla-central/firefox-61.0a1.en-US.win64.zip

Comment 7

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

OK, I was confused by comment 2, which seems to say that you know that *exactly* https://hg.mozilla.org/mozilla-central/rev/d112cf7b2b60 is responsible ("I don't see how a backout of developer tools code could cause this.").

The ranges in the last line of comment 3 and the range implied by the first line of comment 3 and by comment 6 are still different, though.  The range implied by the first line of comment 3, or by comment 6 range is narrower than the last line of comment 3's and is just:
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=00bdc9451be6557ccce1492b9b966d4435615380&tochange=ff0efa4132f0efd78af0910762aec7dcc1a8de66

Comment 8

[Emilio Cobos Álvarez (:emilio)]

Narrowed it a bit more to: https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=7c2423c6e9d0d22b1db406db1f81fbab2a0b02ef&tochange=dd31fb345c11732432fa61a7d8a1b727d41b4a1c

Given that and that layout.display-list.retain = false fixes this, I'm going to guess bug 1443027.

Comment 9

[Gingerbread Man]

(In reply to David Baron :dbaron: ⌚️UTC-7 from comment #7)
> OK, I was confused by comment 2, which seems to say that you know that
> *exactly* https://hg.mozilla.org/mozilla-central/rev/d112cf7b2b60 is
> responsible ("I don't see how a backout of developer tools code could cause
> this.").

* It's not that I know, it's that testing with mozregression-gui 3 times yielded the same pushlog at comment 2. I found that dubious, and comment 8 confirms this. I must've hit either bug 1254812 or bug 1256379 (or maybe the latter is a dupe of the former).

* If I use the full log to manually piece together a push log, this is the result:
https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=6415ccbf739f36f71548540bd92c3b86ab9e1529&tochange=d112cf7b2b60e6244099dc3b599a2444ba0d1da3

* Yes, https://hg.mozilla.org/mozilla-central/rev/d112cf7b2b60 lists different changesets than the about:buildconfig page in those builds it links to, with the pushlogs based on the former being narrower than the latter. I don't know why.

Comment 10

[Matt Woodrow (:mattwoodrow)]

Attachment: Bug 1459441 - Make sure we build the full display list when we have blend containers in order to get the correct sorting for them.

Review commit: https://reviewboard.mozilla.org/r/242924/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/242924/

Comment 11

[Markus Stange [:mstange]]

Comment on attachment 8974577 [details]
Bug 1459441 - Make sure we build the full display list when we have blend containers in order to get the correct sorting for them.

https://reviewboard.mozilla.org/r/242924/#review249360

Comment 12

[Marat Tanalin]

An example of a large real-world site using fancybox is iXBT.com, a russian-language hi-tech news and articles site. E.g. click the first image here to see the glitches caused by the Firefox 61+ bug:

https://www.ixbt.com/news/2018/05/13/boston-dynamics-spotmini.html

Comment 13

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/9a2af4dad811
Make sure we build the full display list when we have blend containers in order to get the correct sorting for them. r=mstange

Comment 14

[Natalia Csoregi [:nataliaCs]]

https://hg.mozilla.org/mozilla-central/rev/9a2af4dad811

Comment 15

[Matt Woodrow (:mattwoodrow)]

Comment on attachment 8974577 [details]
Bug 1459441 - Make sure we build the full display list when we have blend containers in order to get the correct sorting for them.

Approval Request Comment
[Feature/Bug causing the regression]: Retained-dl
[User impact if declined]: Incorrect z-order sorting for modifications made with mix-blend-mode content.
[Is this code covered by automated tests?]: Yes, new reftest added.
[Has the fix been verified in Nightly?]: Reftest passes.
[Needs manual test from QE? If yes, steps to reproduce]: No
[List of other uplifts needed for the feature/fix]: Bug 1461812
[Is the change risky?]: No.
[Why is the change risky/not risky?]: Abandons retained-dl in the presence of mix-blend-mode, falls back to the non-retained code.
[String changes made/needed]: None.

Comment 16

[Ryan VanderMeulen [:RyanVM]]

Comment on attachment 8974577 [details]
Bug 1459441 - Make sure we build the full display list when we have blend containers in order to get the correct sorting for them.

Retained display list fix needed for the feature to ship in Fx61. Approved for 61.0b6.

Comment 17

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-beta/rev/a5b962b8bb3d

Comment 18

[Hani Yacoub, Desktop QA]

Build ID: 20180516220130
User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:62.0) Gecko/20100101 Firefox/62.0

Verified as fixed on Firefox Nightly 62.0a1 on Windows 10 x 64, Windows 7 x32, Mac OS X 10.12 and Ubuntu 16.04 x64.

Comment 19

[Hani Yacoub, Desktop QA]

Build ID: 20180517141400
User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:61.0) Gecko/20100101 Firefox/61.0

Verified as fixed on 61.0b6 Firefox on Windows 10 x 64, Windows 7 x32, Mac OS X 10.12 and Ubuntu 16.04 x64.