Re: [Gnash-commit] gnash ChangeLog server/parser/shape_character_d...

[strk]

On Mon, Nov 12, 2007 at 08:21:31PM +0100, Udo Giacomozzi wrote:
> Hello strk,
> 
> Monday, November 12, 2007, 3:39:21 PM, you wrote:
> s> Gnash is giving a false positive since your change.
> 
> Right. A new commit is on the way that fixes this...
> 
> s> Could be the cirle fill (defined by curves).
> 
> It is the blue shape (red outline) on the left side. It fails because
> the test coordinate is at the very top of that shape and that's where
> the current implementation fails. But I've fixed this now and will
> commit it in a few moments (running test suite ATM).
> 
> BTW, it's a bit of a problem having so much shapes on the stage at the
> same time and doing hit tests. It can produce false positives and
> negatives (happened at least twice) - which is good and bad at the
> same time. Good because otherwise we probably wouldn't have noticed
> this last bug and bad because it perhaps hides other bugs.

Page 2 of DrawingApiTest.as has less of this problem, in that each
"case" is drawn in its own movieclip, and self-contained hitTest is
used on it.

Anyway, if we did check so closely we wouldn't have cought this
bug, so it's not necessarely a bad approach.

> A general word to test cases: It's pratically *impossible* to create
> test cases for every situation. While working on the point_test
> implementation I found an endless number of situations, each would
> need a dozen test points...

It's a long way, you maybe understand better why I keep insisting
on testcases :)

--strk;