886232 - Firefox renders css 3d transform very flashing in high frequently loop

Status: RESOLVED INVALID

(Core :: Web Painting, defect)

Description

[Tetsuharu OHZEKI [:tetsuharu] (UTC+9)]

[Environment]
* http://hg.mozilla.org/mozilla-central/rev/4c4f75c20e9b

[Summary]
Firefox has the rendering problem with updating css 3d transform in high frequently loop.

[Step to reproduce]
Please open this page (The author of this page is not me). This page uses css 3d transfrom and updates it in high frequently loop.
http://ics-web.jp/lab-data/130527_html5_sp/translate3d/translate3d.html

[Result]
* Firefox: objects are very flashing.
* Chromium: objects animates.
* Opera (Presto): objects are shown. But they don't animate.

[Remark]
Of course, this test page may be bad codes from viewpoint of cross browser. However, I think the behavior of Firefox is very strange. Firefox may have some bugs in implementation of css 3d transform.

Comment 1

[Jet Villegas (inactive)]

Attachment: Test case showing the issue.

Seems to be a bug with the 3rd parameter to scale3D. Bug appears when passing in 0 as the Z factor. Test case works when using a non-zero scale factor. See line 51 in translate3D.js

Comment 2

[Matt Woodrow (:mattwoodrow)]

Scaling the z axis to 0 is a singular transform (at least in the traditional sense), so we don't draw these items.

We bail out here: http://mxr.mozilla.org/mozilla-central/source/layout/base/FrameLayerBuilder.cpp#2690

Since we're transforming 2d planes, rather than 3d objects, maybe it makes sense to have scaleZ(0) still be visible. Looking at the spec now to see if it covers this.

Comment 3

[Matt Woodrow (:mattwoodrow)]

"If a transform function causes the current transformation matrix (CTM) of an object to be non-invertible, the object and its content do not get displayed."

http://www.w3.org/TR/css3-transforms/#transform-function-lists

So, I'd say our behaviour here follows the spec and WebKit doesn't.

Comment 4

[Tetsuharu OHZEKI [:tetsuharu] (UTC+9)]

Thank you for considering. Should I close this bug?

Comment 5

[Jet Villegas (inactive)]

(In reply to Tetsuharu OHZEKI [:saneyuki_s] from comment #4)
> Thank you for considering. Should I close this bug?

Yes, please. Matt is correct. We shouldn't render this matrix scale factor. I logged a comment in this webkit bug:
https://bugs.webkit.org/show_bug.cgi?id=53695

Comment 6

[Tetsuharu OHZEKI [:tetsuharu] (UTC+9)]

I have a proposal. I think that it's useful for develop if Gecko display some error or waring about this "non-invertible" case in console.

I'll file the bug about it. What do you think about it?

Comment 7

[Jet Villegas (inactive)]

+1

Comment 8

[Tetsuharu OHZEKI [:tetsuharu] (UTC+9)]

Thank you, jet.

I filed bug 887142.