STEPS TO REPRODUCE 1. Create a path. 2. Set a transformation. 3. Fill the path. ACTUAL RESULTS Transformation is ignored. EXPECTED RESULTS Transformation should have an affect. STEPS TO REPRODUCE 1. Set a transformation. 2. Create a path. 3. Undo the transformation. 4. Fill the path. ACTUAL RESULTS Transformation is applied. EXPECTED RESULTS Transformation should have no affect.
TESTCASES http://www.hixie.ch/tests/adhoc/html/canvas/007.html http://www.hixie.ch/tests/adhoc/html/canvas/008.html http://www.hixie.ch/tests/adhoc/html/canvas/020.html http://www.hixie.ch/tests/adhoc/html/canvas/032.html (Some of these may also expose other bugs.)
As mentioned on IRC, it would be really great to have this fixed soon so that we can determine whether this is possible to fix or not. If it's not possible to fix then we should change the spec, but that would make the platform quite sucky so I'd like to avoid doing that if at all possible. (You'd have to redraw the path everytime you wanted to paint it somewhere new, instead of being able to create the path once and rapidly paint it in multiple places.)
Created attachment 14497 [details] Patch that fixes problem Okay, this fixes behaviour to match spec -- still need to confirm that this doesn't destroy dashboard
Hum. This looks like a really easy fix. The only problem is that it sounds like there needs to be some check of dashboard widgets with this patch applied. That's probably best done by some apple person (with access to leopard)... and probably that will only happen post-leopard (possibly compounding the problem).
So this didn't make Safari 3... can we get a post-mortem? Was it too incompatible, requiring this to become WONTFIX? Or is it still something we can do?
For canvas.isPointInPath to be implemented completely and according to spec we need to land this patch. Although i hate to suggest it, should we consider having a dashboard compatibility quirk for this? I don't think we ever confirmed whether or not this would break dashboard. It should be noted however that firefox has this same behaviour.
Oooh, it occurs to me, could we make this dependent on being in html5 "strict"? eg. switch on the presence of the html5 doctype? (this type of compat quirk is new to me so i don't know how it works/if it's possible... i'm hoping hyatt would know :D ) Thoughts anyone?
The spec changed recently, since Firefox and Safari refused to change to this behaviour, and the Opera guys requested that the spec change to match them. Thus I believe this bug is now invalid.
IT's a shame as feel the old specced behaviour was better in a number of ways -- ithnk we should really consider adding a Path object to HTML5