765107 - Filter output is mislocated with CSS transforms

Status: RESOLVED FIXED

(Core :: SVG, defect)

Description

[Brian Birtles (:birtles, Feb 26~ parental leave)]

Attachment: Test case

When an SVG filter is used inside content that is scaled using a CSS transform, the position of the filter output is sometimes incorrect.

In the attached test case there is a filter applied to a shape. The shape is then reproduced with just a stroke. The stroke should sit on top of the blurred shape.

If you look at the SVG file alone, or set the scale to 1, the result is as expected.

With the scale set to 1.2 the two shapes fail to align. Interestingly the horizontal position seems fine.

Comment 1

[Jonathan Watt [:jwatt]]

Hmm. I'm not seeing this problem, either in current m-c, or if F13.

Comment 2

[Brian Birtles (:birtles, Feb 26~ parental leave)]

(In reply to Jonathan Watt [:jwatt] from comment #1)
> Hmm. I'm not seeing this problem, either in current m-c, or if F13.

I see it in trunk and Aurora. Might be Windows specific. What platform are you on?

Comment 3

[Jonathan Watt [:jwatt]]

Mac. What if you turn on/off hardware acceleration, or any other gfx stuff that might affect this?

Comment 4

[Jonathan Watt [:jwatt]]

CC'ing roc, since he may be the only person with a Window's machine who knows the CSS-transforms code well enough to track this down.

Comment 5

[Brian Birtles (:birtles, Feb 26~ parental leave)]

(In reply to Jonathan Watt [:jwatt] from comment #3)
> Mac. What if you turn on/off hardware acceleration, or any other gfx stuff
> that might affect this?

Yeah, if I set gfx.direct2d.disabled to true it renders correctly.

Comment 6

[Matt Woodrow (:mattwoodrow)]

Can you check the layer state too please (in about:support)

Comment 7

[Brian Birtles (:birtles, Feb 26~ parental leave)]

(In reply to Matt Woodrow (:mattwoodrow) from comment #6)
> Can you check the layer state too please (in about:support)

I get:

GPU Accelerated Windows: 1/1 Direct3D 10

Comment 8

[Brian Birtles (:birtles, Feb 26~ parental leave)]

Oh, and AzureBackend: direct2d

Comment 9

[Jonathan Watt [:jwatt]]

Bas, see comment 5.

Comment 10

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

Most likely an Azure-drawn content bug.

Comment 11

[Brian Birtles (:birtles, Feb 26~ parental leave)]

I'm not sure if this is related, but I'm also seeing incorrect results with non-scaling stroke when direct2d is enabled for the following test:

http://schepers.cc/svg/vectoreffects/nonscalingstroke.svg

Specifically, the stroke appears way off to the left (i.e. doesn't even line up with the fill).

It's fine with direct2d disabled.

Comment 12

[Brian Birtles (:birtles, Feb 26~ parental leave)]

This appears to be affected by regular SVG transforms too. I'm seeing it, or something very similar, in situations where animateTransform is applied.

Comment 13

[Brian Birtles (:birtles, Feb 26~ parental leave)]

Attachment: Proposed patch v1a

Comment 14

[Brian Birtles (:birtles, Feb 26~ parental leave)]

There should probably be a reftest for this too. Will add tomorrow.

Comment 15

[Brian Birtles (:birtles, Feb 26~ parental leave)]

Attachment: Reftest v1a

Comment 16

[Brian Birtles (:birtles, Feb 26~ parental leave)]

Regression range:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=36e938e51481&tochange=f43e8d300f21

Which points to:
Bug 715768: Enable Azure-Thebes wrapper by default for D2D

Comment 17

[Bas Schouten (:bas.schouten)]

Comment on attachment 643370 [details] [diff] [review]
Reftest v1a

Review of attachment 643370 [details] [diff] [review]:
-----------------------------------------------------------------

I'm pretty sure you didn't mean to disable all those other test categories! Don't forget to re-enable before you push!

Comment 18

[Brian Birtles (:birtles, Feb 26~ parental leave)]

(In reply to Bas Schouten (:bas) from comment #17)
> I'm pretty sure you didn't mean to disable all those other test categories!
> Don't forget to re-enable before you push!

Ooh yeah. Thanks for that!

Comment 19

[Brian Birtles (:birtles, Feb 26~ parental leave)]

Pushed to m-i:
https://hg.mozilla.org/integration/mozilla-inbound/rev/9f9d8eb30fb7
https://hg.mozilla.org/integration/mozilla-inbound/rev/5510df5fbe30

Comment 20

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/9f9d8eb30fb7
https://hg.mozilla.org/mozilla-central/rev/5510df5fbe30

Comment 21

[Brian Birtles (:birtles, Feb 26~ parental leave)]

Comment on attachment 643150 [details] [diff] [review]
Proposed patch v1a

Requesting approval to land on Aurora and Beta

[Approval Request Comment]
Bug caused by (feature/regressing bug #): 715768
User impact if declined: Filter effects in some cases render in the wrong location
Testing completed (on m-c, etc.): Baked on m-c
Risk to taking this patch (and alternatives if risky): None
String or UUID changes made by this patch: None

Comment 22

[Lukas Blakk [:lsblakk] use ?needinfo]

Comment on attachment 643150 [details] [diff] [review]
Proposed patch v1a

Go ahead and land to branches, no risk and baked on m-c looks good.

Comment 23

[Brian Birtles (:birtles, Feb 26~ parental leave)]

Pushed to beta and aurora:
https://hg.mozilla.org/releases/mozilla-beta/rev/7ad53adfa8c1
https://hg.mozilla.org/releases/mozilla-aurora/rev/f62ab8356d23

Comment 24

[Paul Silaghi, QA [:pauly]]

Verified fixed on FF 15b3 loading the test case in comment 0. Able to see the issue on FF 15b2.