Bug 40362 – [Qt] canvas/philip/tests/2d.path.stroke.overlap.html fails with Qt-4.6.2 or higher

• Status: RESOLVED
• Resolution: FIXED
• Component: Tools / Tests

[Csaba Osztrogonác]

canvas/philip/tests/2d.path.stroke.overlap.html passes 
with Qt-4.6.0, but fails with Qt-4.6.2 and Qt-4.7.0.

--- /home/webkitbuildbot/slaves/release32bit-4.7.0/buildslave/qt-linux-32-release-qt470/build/layout-test-results/canvas/philip/tests/2d.path.stroke.overlap-expected.txt	2010-06-09 03:27:09.907399661 -0700
+++ /home/webkitbuildbot/slaves/release32bit-4.7.0/buildslave/qt-linux-32-release-qt470/build/layout-test-results/canvas/philip/tests/2d.path.stroke.overlap-actual.txt	2010-06-09 03:27:09.903399630 -0700
@@ -5,4 +5,4 @@
 Actual output:
 Expected output:
 
-Passed
+Failed assertion: got pixel [0,191,0,255] at (50,25), expected [0,127,0,255] +/- 1

[Csaba Osztrogonác]

Skipped by http://trac.webkit.org/changeset/60890 until fix
to make buildbots happy after update to Qt-4.6.2.

[WebKit Review Bot]

http://trac.webkit.org/changeset/60890 might have broken GTK Linux 32-bit Debug
The following changes are on the blame list:
http://trac.webkit.org/changeset/60890
http://trac.webkit.org/changeset/60891
http://trac.webkit.org/changeset/60892

[Csaba Osztrogonác]

(In reply to comment #2)
> http://trac.webkit.org/changeset/60890 might have broken GTK Linux 32-bit Debug
> The following changes are on the blame list:
> http://trac.webkit.org/changeset/60890
> http://trac.webkit.org/changeset/60891
> http://trac.webkit.org/changeset/60892

no, we are innocent now. :)

[Csaba Osztrogonác]

Still valid bug with Qt 4.8 and Qt 5 WK1 and WK2.

[Rafael Brandao]

I'll investigate it.

[Rafael Brandao]

(In reply to comment #5)
> I'll investigate it.

According to spec: "As a result of how the algorithm to trace a path is defined, overlapping parts of the paths in one stroke operation are treated as if _their union_ was what was painted."

We are not handling like that, but we are painting the intersection darker, thus the failure on it. It came out that the problem is how the function QPainter::strokePath works, I'm still trying to figure out a great solution for it.

[Rafael Brandao]

Created attachment 143148 [details]
Patch

Hey No'am, could you take a look?

[Noam Rosenthal]

Comment on attachment 143148 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=143148&action=review

> Source/WebCore/platform/graphics/qt/GraphicsContextQt.cpp:575
> +            QPainterPath stroke = pathStroker.createStroke(platformPath);
> +            p->fillPath(stroke, shadowPen.brush());

You repeat this sequence about 5 times, please put it in some static function.

[Rafael Brandao]

Created attachment 143318 [details]
Patch

[Noam Rosenthal]

Comment on attachment 143318 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=143318&action=review

> Source/WebCore/platform/graphics/qt/GraphicsContextQt.cpp:531
> +inline static void fillPathStroke(QPainter* painter, QPainterPathStroker& pathStroker, QPainterPath& platformPath, const QBrush& brush)

shouldn't platformPath be a const reference?

[Rafael Brandao]

Created attachment 143326 [details]
Patch

[WebKit Review Bot]

Comment on attachment 143326 [details]
Patch

Clearing flags on attachment: 143326

Committed r118020: <http://trac.webkit.org/changeset/118020>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.