828157 - SVG / radialGradient - focal point of radial gradient not rendered correctly

Status: RESOLVED FIXED

(Core :: Graphics, defect)

Description

[kivoin]

Attachment: firebug.svg

User Agent: Mozilla/5.0 (Windows NT 6.2; WOW64; rv:18.0) Gecko/20100101 Firefox/18.0
Build ID: 20130104151925

Steps to reproduce:

I'm trying to get SVG image look same as with other browsers and SVG editors.


Actual results:

Firefox won't show SVG radialGradient's correctly.


Expected results:

Firefox should show SVG radialGradient's same as other browsers and SVG editors do.

Comment 1

[kivoin]

Attachment: Differences

Comment 2

[Mardeg]

Attachment: Differences on Windows XP

Screenshot of Firefox 18, Opera 12.12, and Chromium 25 with the testcase, on Windows XP with no hardware acceleration.

Comment 3

[Robert Longson [:longsonr]]

So this only seems to happen with HWA. On my unaccelerated Vista machine it's OK.

kivoin is this something that was OK in Firefox 17?

Comment 4

[Mardeg]

kivoin: to confirm if the testcase works with HWA off, go to Firefox -> Options, "Advanced" section, "General" tab, untick "Use hardware acceleration when available".

Comment 5

[Robert Longson [:longsonr]]

Setting/unsetting HWA on a Windows 7 machine that supports HWA breaks/fixes the drawing for me.

Comment 6

[kivoin]

(In reply to Robert Longson from comment #3)
> So this only seems to happen with HWA. On my unaccelerated Vista machine
> it's OK.
> 
> kivoin is this something that was OK in Firefox 17?

Problem appears with 17.0.1 also

Comment 7

[kivoin]

(In reply to Mardeg from comment #4)
> kivoin: to confirm if the testcase works with HWA off, go to Firefox ->
> Options, "Advanced" section, "General" tab, untick "Use hardware
> acceleration when available".

After turning HWA off, it shows SVG correctly. But if I turn it back on (and after restart) problem comes back.

Comment 8

[Loic]

I don't think it's a SVG bug, more graphics about Azure.

Regression range with HWA enabled (Win 7):

m-c
good=2012-05-23
bad=2012-05-24
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=36e938e51481&tochange=f43e8d300f21

m-i
good=2012-05-21
bad=2012-05-22
http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=de36859eb332&tochange=219dc3f47568

Suspected bug:
Bas Schouten — Bug 715768: Enable Azure-Thebes wrapper by default for D2D. r=jrmuizel

In addition, there is a similar bug, see bug 820988.
Workaround: gfx.content.azure.enabled = false fixes the issue.

Comment 9

[Bas Schouten (:bas.schouten)]

Definitely looks like Azure-Thebes. Investigating.

Comment 10

[Bas Schouten (:bas.schouten)]

Attachment: Properly offset gradient origin

Comment 11

[Jeff Muizelaar [:jrmuizel]]

Comment on attachment 700485 [details] [diff] [review]
Properly offset gradient origin

Review of attachment 700485 [details] [diff] [review]:
-----------------------------------------------------------------

This should have a test.

Comment 12

[Bas Schouten (:bas.schouten)]

(In reply to Jeff Muizelaar [:jrmuizel] from comment #11)
> Comment on attachment 700485 [details] [diff] [review]
> Properly offset gradient origin
> 
> Review of attachment 700485 [details] [diff] [review]:
> -----------------------------------------------------------------
> 
> This should have a test.

That does sound lovely, but I'm unsure of how we could create a similar rendering without using the same features.

Comment 13

[kivoin]

This issue exist in version 19.
Is there estimated time frame for fixes?

Comment 14

[Loic]

Plsea, don't play with the flags if you don't know them, thanks.

Comment 15

[Jonathan Watt [:jwatt]]

Attachment: patch

(In reply to Bas Schouten (:bas.schouten) from comment #12)
> That does sound lovely, but I'm unsure of how we could create a similar
> rendering without using the same features.

Here's a test that will do that, and hopefully unblock Jeff's review. I've checked that it fails on Windows 7, but if you want to double check and r+, Bas, that'd be great.

Comment 16

[Jeff Muizelaar [:jrmuizel]]

Comment on attachment 700485 [details] [diff] [review]
Properly offset gradient origin

r+ with the test

Comment 17

[Robert Longson [:longsonr]]

Bas, can you land this please?

Comment 19

[Jonathan Watt [:jwatt]]

Both reviewed patches need to be checked in.

Comment 20

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/0e664105158d
https://hg.mozilla.org/integration/mozilla-inbound/rev/9c8454e1506e

Comment 21

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/0e664105158d
https://hg.mozilla.org/mozilla-central/rev/9c8454e1506e

Comment 22

[Patrick Storz]

Why won't this be fixed in Firefox 20 and 21?

Comment 23

[Ed Morley [:emorley]]

I'#m not the patch author, you are best off asking Bas/jwatt.

That said, changes first land on mozilla-central (which is at version 22) and if deemed low enough risk/high enough reward, can be nominated for uplift to the aurora (v21) and beta (v20) branches.

Comment 24

[Patrick Storz]

So, could this also be landed on aurora and beta Bas?

Comment 25

[Patrick Storz]

Finally resolved with the just released Firefox 22.