338640 - rotate() shouldn't affect the current path

Status: RESOLVED DUPLICATE of bug 380338

(Core :: Graphics: Canvas2D, defect)

Description

[Anne (:annevk)]

rotate() (and other transformations) should apply just before rendering. They should not affect the next lineTo() etc. directly.

Comment 1

[Anne (:annevk)]

Attachment: testcase

Comment 2

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Er, no; rotate/translate/etc. all modify the current transformation matrix, which is applied to each path operation point when it's added to the path.  Otherwise, there's no consistent way to integrate transforms and path points; you could try something like "the CTM at the time the path is created", but we also have things like the implicit moveTo(0,0) when the canvas is created, which would then lock in an identity matix, which would be broken.

Having the CTM affect each point as its specified keeps everything consistent (in fact, calling rotate() /after/ you specify a path fully but before you call fill() or something should have no effect).

The spec should probably explicitly state this (that path points are transformed by the CTM as path operations are called).  (Do any of the implementations actually behave the way you describe?  I'm pretty sure Safari doesn't, though I can't test at the moment.)

Comment 3

[Anne (:annevk)]

Both Safari and Opera pass the test. (As in, they don't show this star thingie.)

Comment 4

[Anne (:annevk)]

Reopening based on comment 3 and lack of reply from Ian.

Comment 5

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

So Safari 2.0.3 shows the star for me; the latest WebKit nightly does not.  Not sure what changed.