[whatwg] Opera canvas bug?

[Ian Hickson]

On Sun, 1 Oct 2006, Benjamin Joffe wrote:
>
> I have set up a few pages to demonstrate bugs with Opera's canvas 
> implementation. You can see it here: 
> http://abrahamjoffe.com.au/ben/opera_canvas_bugs/index.htm , I am sure 
> that the first 3 are genuine Opera bugs but the 4th one might not be 
> because I do not think that the WhatWG spec is specific enough with 
> nested clipping paths, none-the-less I feel that the way Firefox handles 
> the script is more correct than how Opera does. Would you agree?

The first test is:

   http://abrahamjoffe.com.au/ben/opera_canvas_bugs/bug_scale.htm

Definitely an Opera bug. I've added a test based on this to my test suite:

   http://www.hixie.ch/tests/adhoc/html/canvas/031.html

The second test:

   http://abrahamjoffe.com.au/ben/opera_canvas_bugs/bug_restore.htm

...is actually a bug in Firefox and Safari:

   http://www.hixie.ch/tests/adhoc/html/canvas/032.html

The third test:

   http://abrahamjoffe.com.au/ben/opera_canvas_bugs/bug_rect.htm

...seems to work everywhere now.

The fourth test:

   http://abrahamjoffe.com.au/ben/opera_canvas_bugs/bug_clip.htm

...is indeed an Opera bug. The spec seems clear, though:

# The clip() method must create a new clipping path by calculating the 
# intersection of the current clipping path and the area described by the 
# current path (after applying the current transformation), using the 
# non-zero winding number rule.

Is that confusing?

I've created this test:

   http://www.hixie.ch/tests/adhoc/html/canvas/034.html

...based on your test.

The fifth and final test:

   http://abrahamjoffe.com.au/ben/opera_canvas_bugs/bug_data.htm

...seems to work everywhere now.

Thanks for the feedback,
-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'