747381 - [regression] Incorrect rendering of images when perspective and scale is applied

Status: RESOLVED DUPLICATE of bug 735373

(Core :: Web Painting, defect)

Description

[Louis-Rémi BABE]

Attachment: Without a background color

The latest nightly (14.0a1 2012-04-20) introduced a bug in image rendering. If a `transform: perspective(500px) scale(1.2);` is applied to an image, the image is rendered approx. 100px on top and left of where it should be. 

The problem does not appear:
- if a background color is applied to the image,
- if the perspective is removed,
- if the scale is removed or changed to another 2d transform function,
- if the transform is applied to text.

An example is available in this jsbin: http://jsbin.com/3/ugipen/5/edit?html,css,live (make sure css and output tabs are open)
Remove the background color to see the difference.

The problem happens with latest nightly and has been verified both on Ubuntu and Mac (different hardware as well). I've attached two screenshots of what I'm seeing.

Comment 1

[Louis-Rémi BABE]

Attachment: With a background color

Comment 2

[Louis-Rémi BABE]

Actually, the image is rendered misplaced or not rendered at all, as in the example above.

Comment 3

[Boris Zbarsky [:bzbarsky]]

Chances are, a bug somewhere in display item or layer code....

Louis-Rémi, would you be willing to hunt down a regression range?

Comment 4

[Louis-Rémi BABE]

I've never done that but I'd be willing to.

Comment 5

[Louis-Rémi BABE]

Oops, Looks like I inadvertently changed the component.

Comment 6

[Alex Keybl [:akeybl]]

Difficult to say how many sites would be affected by this regression, so tracking for FF14 to resolve prior to release.

Comment 7

[Alice0775 White]

id is duplicated in the test case.

Comment 8

[Alice0775 White]

and perspective(500) in css. Missing unit?

Comment 9

[Lukas Blakk [:lsblakk] use ?needinfo]

Assigning to Louis-Rémi for now, if you find the regression range and are able to determine who was the cause, reassign as needed to keep an owner on the hook for a fix here.

Comment 10

[Lukas Blakk [:lsblakk] use ?needinfo]

[Triage Comment]

Any chance of finding the regression range soon? We're about to merge 14 to beta channel tomorrow so more users will be seeing this issue.

Comment 11

[Alex Keybl [:akeybl]]

(In reply to Lukas Blakk [:lsblakk] from comment #10)
> [Triage Comment]
> 
> Any chance of finding the regression range soon? We're about to merge 14 to
> beta channel tomorrow so more users will be seeing this issue.

Adding qawanted to make sure this gets on their radar.

Comment 12

[Mihaela Velimiroviciu (:mihaelav)]

Last good nightly: 26-02-2012 (rev cd4853b0b94a)
First bad nightly: 27-02-2012 (rev d1b2fd680235)

Pushlog: http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=cd4853b0b94a&tochange=d1b2fd680235

The CSS property which causes the issue seems to be -moz-transform.

Comment 13

[Alex Keybl [:akeybl]]

(In reply to Mihaela Velimiroviciu [QA] from comment #12)
> Last good nightly: 26-02-2012 (rev cd4853b0b94a)
> First bad nightly: 27-02-2012 (rev d1b2fd680235)
> 
> Pushlog:
> http://hg.mozilla.org/mozilla-central/
> pushloghtml?fromchange=cd4853b0b94a&tochange=d1b2fd680235
> 
> The CSS property which causes the issue seems to be -moz-transform.

Hey bz - do you suspect any of the changes in this list in particular?

Comment 14

[Alice0775 White]

Regression range:
Good:
http://hg.mozilla.org/integration/mozilla-inbound/rev/c928dc7c9dff
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:13.0) Gecko/20120225 Firefox/13.0a1 ID:20120225011949
Bad:
http://hg.mozilla.org/integration/mozilla-inbound/rev/7dcff354aeae
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:13.0) Gecko/20120225 Firefox/13.0a1 ID:20120225014049
Pushlog:
http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=c928dc7c9dff&tochange=7dcff354aeae

However, It works in Firefox14 and later.
Fixed range:
Bad:
http://hg.mozilla.org/integration/mozilla-inbound/rev/b69d000b726a
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:15.0) Gecko/15.0 Firefox/15.0a1 ID:20120514231326
Fixed:
http://hg.mozilla.org/integration/mozilla-inbound/rev/dd3c873674fc
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:15.0) Gecko/15.0 Firefox/15.0a1 ID:20120515030522
Fixed pushlog
http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=b69d000b726a&tochange=dd3c873674fc

This seems fixed by Bug 750334

Comment 15

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

Given those ranges, presumably caused by bug 724614 and fixed by bug 750334, yes.

Comment 16

[Boris Zbarsky [:bzbarsky]]

Yeah, agreed.  Not sure what we should do for 14 here...

Comment 17

[Aryeh Gregor (:ayg) (no longer with Mozilla)]

FWIW, this was known since March 13 but slipped between the cracks.  I'm guessing a straight backout of bug 724614 for 14 would be easy.