Bug 4211 – REGRESSION: Most SVG's render black

• Component: SVG
• Status: RESOLVED FIXED

[Eric Seidel (no email)]

Most SVG's render black

There seems to have been a serious regression in color handling during the process of merging SVG 
support into TOT WebCore.  Possibly the regression is in KWQColor... more likely I'm doing something else 
wrong.

Either way, most SVG's fail to lookup their colors properly and thus all colors are black.

[Ben La Monica]

Created attachment 3181 [details]
A simple SVG to show problems with drawing colors

It turns out that those SVGs that render properly don't use any solid fills.
However, if you use a RadialGradient (or I assume, any gradient) it applies the
color correctly.

This file has 2 ellipses in it, the smaller ellipse should be red. The larger
ellipse is blue.

[Eric Seidel (no email)]

Created attachment 3184 [details]
An even simpler reduction

[Eric Seidel (no email)]

pojo2 found the problem.

Basically, this happens because we're not linking in WebKit... which means that some bridge functions 
don't get defined.

One example is this one:

- (CGColorSpaceRef)createRGBColorSpace
{
    return 0;
}

Thus we're getting back a NULL color space.  Now... why this means that gradients are still working, I 
don't know.  That sounds like a bug.

[Eric Seidel (no email)]

Created attachment 3189 [details]
Removes WebCore's usage of WebKitGraphicsBridge, fixing bug.

[Darin Adler]

Comment on attachment 3189 [details]
Removes WebCore's usage of WebKitGraphicsBridge, fixing bug.

Looks great, but doesn't go far enough. Should also remove the unused methods
from the WebCoreGraphicsBridge interface, and from WebKit.

(And later we should remove the underlying calls from WebKitSystemInterface as
well.)

[Eric Seidel (no email)]

Created attachment 3191 [details]
Removes WebKit portion of WebKitGraphicsBridge

[Eric Seidel (no email)]

According to autovicki, this change produced no speed regressions.  Possibly a speed improvement:

Before:

TestSuite: PageLoadTest
TestType: cvs-base
OSVersion: 8C46
SafariVersion: (v412+)
Machine Model: PowerBook G4 12"
Number Of CPUs: 1
CPU Speed: 1.33 GHz
Memory: 1.25 GB
Date: Mon Aug  1 02:06:45 PDT 2005
==================================

cvs-base Cold:      	0.416
cvs-base Uncached:      0.226
cvs-base AvgCached:     0.159
cvs-base Cached 1:  0.159
cvs-base Cached 2:  0.159
cvs-base Cached 3:  0.158
cvs-base Cached 4:  0.161
cvs-base Cached 5:  0.158


After:

TestSuite: PageLoadTest
TestType: cvs-base
OSVersion: 8C46
SafariVersion: (v412+)
Machine Model: PowerBook G4 12"
Number Of CPUs: 1
CPU Speed: 1.33 GHz
Memory: 1.25 GB
Date: Mon Aug  1 01:41:51 PDT 2005
==================================

cvs-base Cold:      	0.418
cvs-base Uncached:      0.226
cvs-base AvgCached:     0.158
cvs-base Cached 1:  0.159
cvs-base Cached 2:  0.158
cvs-base Cached 3:  0.157
cvs-base Cached 4:  0.158
cvs-base Cached 5:  0.159

[Eric Seidel (no email)]

Commited WebCore fixes, waiting for WebKit fix approval.

[Darin Adler]

Comment on attachment 3191 [details]
Removes WebKit portion of WebKitGraphicsBridge

Looks fine. Lets be sure to remove the WebCore side too.

[Eric Seidel (no email)]

Sent mail to Darin w/ proposed SPI removal.  The public part of this bug is fixed.