Maybe poly2tri is not the right delaunay triangulation library for axmol? (can not handle complex polygons easily)

[aismann]

see Poly2tri readme:

Because of these results on axmol:

[aismann]

@NgVThangBz

can you made some polygon data like above or this one (i belive poly2tri can not handle this kind of polygon sections without add some holes or steiner stuff):
It will be great if you can check it with poly2tri testbed and tell us the result here.

Maybe this is also an usefull test (rotate the same polygon)

[NgVThangBz]

you can check it:
polyLineData2.txt
polyLineData.txt

[aismann]

Have you check it on poly2tri, is it correct on poly2tri?

[NgVThangBz]

@aismann you need data correct on poly2tri?

[aismann]

Yes. Both.
But if it on poly2tri wrong there is also wrong on axmol (at the moment)
For the cpp-test I need also "correct filled" complex examples

[NgVThangBz]

this is data correct on poly2tri
There are 20 datasets here
polyLineData2.txt

[aismann]

example with autopolygon

only one pixel removed:

Removed some pixel

Seems the distance have to be "4 pixel" between two "lines"

[aismann]

Fazit:
poly2tri works not for all "complex" Polygons, but its better as the "cocos2dx 4.0" implementation:

Cocos2dx 4.0/Axmol (old):

Axmol (new)

Only for me (autopolygon data set?):

[aismann]

@halx99
Where can I put a new math function: bool isConvex(const Vec2* verts, int count)?
Its only used on DrawNopde.cpp

[aismann]

I close it now