[ODE] Plane-Cylinder collision: not good in 0.7, worse in SVN :(

[Mauro G. Todeschini]

Bram Stolk wrote:
> Ah,
> 
> Excellent investigation!
> Thanks
> 
> I should be able to work with this,
> More later.
I discovered that the problem not only happens for transformed
cylinders, but even between "normal" cylinders and planes. If in
test_cylvssphere.cpp you only change line 185 from:
  dCreatePlane (space,0,0,1,0);
to:
  dCreatePlane (space,0,0,1,-1);
You can reproduce the problem.

Thanks again and Bye

> 
> 
>    Bram
> 
> 
> -----Original Message-----
> From: "Mauro G. Todeschini" <mauro2006todeschini at itia.cnr.it>
> To: "Bram Stolk" <bram at sara.nl>
> Cc: "'ode-list mailing_list'" <ode at q12.org>
> Sent: 10/28/06 8:20
> Subject: Re: [ODE] Plane-Cylinder collision: not good in 0.7, worse in SVN :(
> 
> Bram Stolk wrote:
>> For me, cyl vs plane is OK for both geomoffset and trf geoms.
>>
>> Could you please do this test against svn-tree?
>>
>> - Run test_boxtest and press 'y' a few times.
>>
>> - Edit test_boxstack.cpp and change line 94 to:
>> #define USE_GEOM_OFFSET 0
>>   Rebuild, and re-run the test.
>>
>> For me, both cases are OK.
>> I just tested this with current svn, on linux, with double prec.
> Ok, the test you suggested works correctly in both cases but... I think
> that changing the #define doesn't make any difference for cylinders (ie.
> trf geoms are never used for cylinders). It only makes difference for
> composite objects (which don't contain any cylinder, only capsules).
> Try putting a printf after line 455 and You'll see that you never pass
> from that line with cylinders.
> 
> In my application I discovered that the problem with planes and
> transformed cylinders seems to happen when I define my plane like this
> (gravity is dWorldSetGravity (world,0,-9.8,0);):
> 
> dCreatePlane (space,0,1,0,-20);
> 
> but the problem seems to disappear if:
> 
> dCreatePlane (space,0,1,0,0);
> 
> I think that the negative value in the first definition has something to
> do with the problem.
> 
> Bye and Thanks
> 
>>  Bram
>>
>>
>>
>>
>> On Fri, 2006-10-27 at 18:09 +0200, LR wrote:
>>> Hi Mauro,
>>>
>>> I have similars problems with cylinder & transform geoms (it's with the
>>> sphere/cylinder). To resolve it, i made a compile switch witch enable me to
>>> compile with "transform geoms" or with "GeomOffsets".
>>>
>>> Using geoms offsets solves the cylinder problems (but why ????). 
>>>
>>> LR
>>>
>>>
>>>  
>>>
>>> -----Message d'origine-----
>>> De : ode-bounces at q12.org [mailto:ode-bounces at q12.org] De la part de Mauro G.
>>> Todeschini
>>> Envoyé : vendredi 27 octobre 2006 15:58
>>> À : ode-list mailing_list
>>> Objet : [ODE] Plane-Cylinder collision: not good in 0.7, worse in SVN :(
>>>
>>> Hi,
>>> 	today I downloaded ODE trunk from SVN to test the latest patches for
>>> collisions between cylinder and plane. In ODE 0.7 there is some problem
>>> (explosion) when the cylinder falls with a circular face parallel to the
>>> plane (ground).
>>> In SVN the explosion happens everytime a cylinder hits a plane (not only on
>>> its circular face).
>>> In my tests cylinders are transformed with geomTransforms.
>>> I read in the mailing-list that there has been some work on the issue
>>> (cylinder-plane collision) at the beginning of september, but I fear that It
>>> has not been tested the case of cylinders in geomTransform.
>>> Has anyone verified the behaviour of cylinders in geomTransforms recently
>>> (after the changes)?
>>>
>>> Thanks and Bye
>>>
>>> _______________________________________________
>>> ODE mailing list
>>> ODE at q12.org
>>> http://q12.org/mailman/listinfo/ode
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> ODE mailing list
>>> ODE at q12.org
>>> http://q12.org/mailman/listinfo/ode
>> _______________________________________________
>> ODE mailing list
>> ODE at q12.org
>> http://q12.org/mailman/listinfo/ode
> 
>