The characters Mario Kart game at <http://www.nihilogic.dk/labs/mariokart/> render weird when they turn - it's like the character is looping around to the other side of where it's supposed to be drawing. See the attached screenshot for more clarity. Firefox 3rc1 on Leopard doesn't have the problem.
Created attachment 21279 [details]
Safari 3.1.1 and ToT both have this problem.
Created attachment 21289 [details]
Qt port glitches
Created attachment 21290 [details]
Gtk port glitches 1
Created attachment 21291 [details]
Gtk Port glitches 2
On the Qt port (r32870), the game area is white except for the karts.
On the Gtk port (r33044), the track can partially be seen if selecting the "Per line canvas", otherwise the game area is black. The trees in the background are drawn though along with karts.
In both cases, the rendering glitch in the first screenshot can be seen.
(In reply to comment #5)
> Created an attachment (id=21291) 
> Gtk Port glitches 2
> On the Qt port (r32870), the game area is white except for the karts.
> On the Gtk port (r33044), the track can partially be seen if selecting the "Per
> line canvas", otherwise the game area is black. The trees in the background
> are drawn though along with karts.
> In both cases, the rendering glitch in the first screenshot can be seen.
It's probably worth splitting those off into separate bugs, unless Oliver thinks those are related to the Mario/Luigi/Princess weirdness.
Also confirmed this happens on Windows
I believe the Qt issue is that they don't support patterns in canvas iirc
Gtk ones also look to be unrelated
Bug #19190 (Gtk) and Bug #19191 (Qt) have been raised separately as it looks like a regression on Gtk and missing code on Qt.
Never the less, the characters weirdness still applies to these 2 ports.
Pierre-Luc? the character also get truncated/offset incorrectly at times? (sorry for delay, didn't get cc mail for some reason)
The game has one image with ca. 33 pictures per driver. Every picture another perspective of the driver. It seems that drawImage(image, sx, sy, sw, sh, dx, dy, dw, dh) gets the wrong parameters.
Either a problem of calculating in webkit or a failure in the script.
I tendent to say it's a failure in the script, because opera has the same glitches like https://bugs.webkit.org/attachment.cgi?id=21279
Hmmm, as far as I can tell the character sprites are drawn with divs on top of the canvas, not using canvas APIs at all.
Hyatt any thoughts?
This is still an issue with Safari 5.
I see no glitches with modern WebKitGTK+.
I can still reproduce with Safari 7.0.3 on Mavericks.