Page 1 of 1

Soc project: Booleans -- win/linux (2005/08/31)

Posted: Thu Sep 01, 2005 6:33 am
by Hos
Here are some executables of Marc Freixas' boolean
project for summer of code. Select two mesh objects
and press 'w' as always. Output has less points,
less bad geometry, and uses quads too.

http://bebop.cns.ualberta.ca/~cwant/ble ... -31.tar.gz
http://bebop.cns.ualberta.ca/~cwant/ble ... -08-31.zip

More info on this project is here:

http://wiki.blender.org/bin/view.pl/Ble ... evelopment

Test it out and provide feedback, please.

Regards,
Chris

Posted: Thu Sep 01, 2005 1:58 pm
by LetterRip
I had it crash when doing a union of a cube and a uvsphere, but can't reproduce the crash now. Other than that, it seems to give great results.

LetterRip

Posted: Thu Sep 01, 2005 2:13 pm
by LetterRip
Been able to reproduce the crash three times now, can't simplify it yet, and don't have a debug build of your booleans...

Add a second cube to the default scene, go to object mode, move the cube a bit, intersect the two cubes. Now move both selected cubes away from center. Add a uvsphere, default sections etc. grab in editmode and move it to the location of the the two cubes. Enter object mode, select one cube, do intersection, crash.

Before the last crash I had saved it right before the crash. However if I open the saved file select the cube and do an intersection with the uvsphere it won't crash. So apparently whatever gets wonky before the save gets cleared when the saved file is loaded.

LetterRip

Posted: Thu Sep 01, 2005 2:28 pm
by lucky3
I get some crashes too, but i can't reproduce it either. It looks like to be what LetterRip said, a memory or saving problem. If blender crashes after several tests and then i re-open blender and i redo the same scene than previously it wont' crash.

Mem error

Posted: Thu Sep 01, 2005 4:09 pm
by mfreixas
I'm trying to crash blender, but I can't reproduce the error. This likes a memory error. I have 1Gb of mem and maybe this can hide a memory error. However, I'm checking the code.

Thanks for feedback.

Posted: Fri Sep 02, 2005 12:56 am
by jiri
Hi,
I did few tests and it works much more better then before. Good work, Marc!

Jiri

Posted: Fri Sep 02, 2005 1:15 am
by LetterRip
I'm checking out your branch, and will do a make debug and see if i can crash it again...

LetterRip

Posted: Fri Sep 02, 2005 2:00 am
by LetterRip
doesn't compile - get a bunch of errors once i get to armature.c (so not your code).

Letterrip

Posted: Fri Sep 02, 2005 6:59 pm
by malefico
Works pretty well here. I see that for cubes and other simple geometry it works flawlessly, for UVspheres or more complex meshes, it's still not perfect but better than original.

In Drill blend for instance, when substracting the 3 orthogonal cilynders from the cube sometimes a triangular "hole" is left, like a face not being drawn or something.

Great job !

Regards

malefico.

Posted: Sat Sep 03, 2005 5:26 am
by Gimble
It does a much better job than the old Booleans, but doesn't seem to be very SubSurf friendly.

Windows SoC Boolean build seems to crash after doing a difference between a cube and an UVSphere and then applying SubSurf...

First time after I pushed the SubSurf button, it crashed. Second time it survived pushing the SubSurf button, but then crashed when I tried to enter Edit mode.

Posted: Mon Sep 05, 2005 4:55 pm
by Tochiro
Hi!,
I think that I have find a bug in new boolean.
If I do a union of 2 cilinders with default radius (see the picture)
the result is not perfect,
Image
but if I scale the cilinders, (scale 0.2) the union is perfect (see picture 2 please), why?
Image

Posted: Wed Sep 07, 2005 2:45 pm
by mfreixas
Thanks for all feedback, I just localize one bug (I think). It is a numeric precission error in one of the triangulation operations. I'm working to fix it, as soon as possible I wil post a fixed version.

regards

Posted: Fri Sep 09, 2005 4:11 pm
by Hos