[Bf-committers] Freestyle: Polygonalization with small "Error"
rd6t-kjym at asahi-net.or.jp
Thu Feb 14 18:42:16 CET 2013
Thanks again for the other problem report. I confirmed that the suggested
procedure resulted in many "invalid strip vertex" warnings.
As far as I tested, Blender did not crash. Instead in some cases (but not
always) big white polygons appear at random as part of thin strokes.
The warnings indicate that the mesh data being rendered (i.e., those
auto-generated for stroke rendering) contain very large coordinate values.
This is actually not an issue in this specific test case, since we know we have
added very large noise values to stroke geometry.
What seems strange to me is that there is some unexpected randomness
in the results of the Polygonization modifier. There is likely another bug
in the modifier implementation. I am going to look into the problem.
The procedure for reproducing the issue is a kind of stress testing from the
view point of the Polygonization modifier because of the extremely perturbed
stroke geometry. Life would be easier for the modifier if you add another
Sampling geometry modifier between the Perlin 2D Noise and Polygonization
modifiers, since the strokes will no longer be juggy (i.e., stroke vertices do not
do a long jump). This way you should be able to avoid the occasional crash.
With best regards,
KAJIYAMA, Tamito <rd6t-kjym at asahi-net.or.jp>
From: IRIE Shinsuke
Sent: Wednesday, February 13, 2013 7:06 AM
To: bf-committers at blender.org
Subject: Re: [Bf-committers] Freestyle: Polygonalization with small "Error"
Thanks for the fix! It works.
However, another crash sometimes happens when adding large amplitude
noise before Polygonalization modifier.
1. Start Blender and use default cube
2. Enable Freestyle
3. Add a line set
4. Add geometory modifier "Perlin Noise 2D" and set Amplitude=1000
5. Add geometory modifier "Polygonalization" and set Error=1.0
6. Start render by F12
Blender may (or may not) crash after getting warnings like
"strip vertex XXX non valid".
Tamito KAJIYAMA wrote:
> Many thanks for the problem report. Indeed there was a bug in the implementation
> of the Polygonization geometry modifier. I made an attempt to fix it in revision 54518.
> Could you please test the code and see if the issue has been addressed?
> Thank you,
More information about the Bf-committers