1244464 - WebkitCSSMatrix support breaks realtor.ca photo gallery (RoyalSlider)

Status: RESOLVED DUPLICATE of bug 1241021

(Core :: DOM: CSS Object Model, defect)

Description

[Shane Caraveo (:mixedpuppy)]

dev edition is my regular browser, this just started since the update to 46

STR
- visit realtor.ca
- search then select a property (that has multiple images)
- click on the image, an in-page dialog appears to view images at a larger size
- click on left or right arrows to view next image (may be a couple times till it stops working)

expected:
- you see the next image

failure:
- you dont see the next image, console shows:

TypeError: content is null
gGestureSupport.rotate()
 browser-gestureSupport.js:445
oncommand()
 browser.xul:1
GS__doCommand()
 browser-gestureSupport.js:338
GS__doAction()
 browser-gestureSupport.js:284
GS__doUpdate()
 browser-gestureSupport.js:148
GS__setupGesture()
 browser-gestureSupport.js:160
GS_handleEvent()
 browser-gestureSupport.js:91
 browser-gestureSupport.js

Comment 1

[Shane Caraveo (:mixedpuppy)]

[Tracking Requested - why for this release]: major breakage probably affects many sites

Comment 2

[:Gijs (he/him)]

I don't see the error message in the console, I expect that's profile dependent. But I can reproduce the issue on a clean profile on Windows. This is the regression window I found with mozregression:


https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=bb6bfd172d6e40b5d6a87d8118faf860c02f8545&tochange=10d4e8736dbbb4f79320273906d148c448cb8e0b

Couple of candidates in there because windows builds were busted in the middle. I suspect the asyncopen or the webkitcssmatrix work, but I've not looked at the site code yet, so I could be way off. Shane, any chance you can narrow this down further with mozregression (assuming you're using OS X?)? This should work:

mozregression --good bb6bfd172d6e40b5d6a87d8118faf860c02f8545 --bad 10d4e8736dbbb4f79320273906d148c448cb8e0b

Comment 3

[Shane Caraveo (:mixedpuppy)]

webkitcssmatrix is the winner!

Comment 4

[Shane Caraveo (:mixedpuppy)]

verified still broken in 02-02 aurora.

Comment 5

[Mike Taylor [:miketaylr]]

Note, this won't get past Nightly or Aurora for now. Taking a closer look...

Comment 6

[Mike Taylor [:miketaylr]]

This probably doesn't belong in the TE component, especially if it's our implementation of WebKitCSSMatrix that's breaking the site.

Comment 7

[Mike Taylor [:miketaylr]]

Shane, are you on Windows?

Comment 8

[Mike Taylor [:miketaylr]]

My windows question was related to this touch event sniffing, which would actually mean the bug was Bug 1241021.

if ("ontouchstart" in window) {
    if (!this.settings.disableTranslate3d) {
        if (("WebKitCSSMatrix" in window && "m11" in new WebKitCSSMatrix()) ) {
            this._ev.css({
                "-webkit-transform-origin": "0 0",
                "-webkit-transform": "translateZ(0)"
            });
            this._ir = true
        }
    }
    this.hasTouch = true;
    this._az = "touchstart.rs";
    this._by = "touchmove.rs";
    this._cx = "touchend.rs"
}

But, this is on Aurora (and I can reproduce on Mac as well).

There's also some code here that relies on webkitTransitionEnd listeners support, which just landed in Bug 236979 (uplift has been requested on that).

So a few things to explore here. Likely WebKitCSSMatrix is taking us down a codepath that is revealing a bug, rather than a bug in the implementation of WebKitCSSMatrix itself.

Comment 9

[Mike Taylor [:miketaylr]]

> My windows question was related to this touch event sniffing, which would actually mean the bug was Bug 1241021.

To be more clear, jQuery.fn.css('-webkit-thing', 'foo') fails due to Bug 1241021. This site is using RoyalSlider (see Bug 1240611), so it likely *is* that bug. They're just using WebKitCSSMatrix support as a flag to go down that codepath, apparently.

Comment 10

[Shane Caraveo (:mixedpuppy)]

I am on osx.

Comment 11

[Mike Taylor [:miketaylr]]

This is working now in Nightly. I'm gonna dupe against Bug 1241021.

Comment 12

[Liz Henry (:lizzard) (relman/hg->git project)]

This is tracked in the duplicated bug.