pepr3d
pepr3d copied to clipboard
poly2tri is not stable
Guys,
just a note, that the poly2tri library is not stable and it will make the application crash randomly. We at the Slic3r PE team are using the old C glu tesselator code by SGI from a pure despair of not finding a free stable self-contained triangulator (apart from CGAL).
Here is the tesselator dug out from the glu library. https://github.com/prusa3d/Slic3r/tree/master/src/glu-libtess
Hi Vojta!
Thanks for the issue. How often does it happen? Could it be a problem just with certain too complex fonts? We found out that certain fonts are too "weird" to be triangulated properly, but did not notice any problems with classic fonts.
It is sensitive to the input geometry, and it crashes with stack overflow or whatever. We use it for triangulating holes left after splitting an object with a plane, and it crashes a lot. For fonts it may be all right, but the very old C code by SGI is much more stable (and much better tested I suppose)
For inspiration, this is a simple wrapper around glu tesselator. ExPolygon is just a contour with a vector of holes. https://github.com/prusa3d/Slic3r/blob/master/src/libslic3r/Tesselate.hpp https://github.com/prusa3d/Slic3r/blob/master/src/libslic3r/Tesselate.cpp
On Mon, Apr 1, 2019 at 9:15 AM Tomáš Iser [email protected] wrote:
Hi Vojta!
Thanks for the issue. How often does it happen? Could it be a problem just with certain too complex fonts? We found out that certain fonts are too "weird" to be triangulated properly, but did not notice any problems with classic fonts.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/tomasiser/pepr3d/issues/148#issuecomment-478463522, or mute the thread https://github.com/notifications/unsubscribe-auth/AFj5I2qzmNTp6UzAcF1z1eDYhUPyvl9Yks5vcbIggaJpZM4cU-s7 .