[webkit-reviews] review canceled: [Bug 47467] SVG <path> DRT dumps have rounding problems across platforms : [Attachment 114942] Patch v2

[bugzilla-daemon at webkit.org]

Nikolas Zimmermann <zimmermann at kde.org> has canceled Nikolas Zimmermann
<zimmermann at kde.org>'s request for review:
Bug 47467: SVG <path> DRT dumps have rounding problems across platforms
https://bugs.webkit.org/show_bug.cgi?id=47467

Attachment 114942: Patch v2
https://bugs.webkit.org/attachment.cgi?id=114942&action=review

------- Additional Comments from Nikolas Zimmermann <zimmermann at kde.org>
Patch v3 won't build on win yet, the String::number() signature changed, and I
need to wait for the win build to show me the right symbol name that I can add
to the JSC.def file.
I inspected the layout test results that Philippe sent me and found out that
the most common source of rounding errors between gtk/mac is the dumping of the
renderer positions, see here:

-	 RenderSVGPath {path} at (94,162) size 220x37 [stroke={[type=SOLID]
[color=#000000] [stroke width=10.00] [miter limit=17.00]}] [data="M 20 110 L
200 120 L 20 130"]
+	 RenderSVGPath {path} at (94,162) size 220x36 [stroke={[type=SOLID]
[color=#000000] [stroke width=10.00] [miter limit=17.00]}] [data="M 20 110 L
200 120 L 20 130"]

These values come from the absoluteClippedOverflowRect() - which maps the
renderers local repaint rect through a bunch of mapRect() calls up to the root
renderer.
mapRect() is using FloatQuad::boundingBox(), operating on rounded float points.
It returns FloatRect(left, top, right - left, bottom - top), where all values
are floats.
We should rather use doubles then calculate the subtraction, then round to
float. Fixing that changes several test results on mac.

Lets see how this affects Gtk. Philippe could you build it again and produce a
baseline comparison against mac?
Quick'n'dirty: mv LayoutTests/platform/gtk/svg gtk-svg-old; mv
LayoutTests/platform/mac/svg/ platform/gtk/svg/