yade-dev team mailing list archive

[Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>]

On 12/07/10 18:39, Václav Šmilauer wrote:
> > It doesn't use the closest apparently. If you want to make some tests,
> > you can checkout last commit (where bad dispatch is workarounded) and
> > uncomment the code for Gl1_ChainedCylinder, which is never called in
> > chained-cylinder-spring.py.
> >      
> There is no Gl1_ChainedCylinder in the current code, so I could not
> check.
>    
It was "only" commented, but ok : it needed to think what/how to 
uncomment. I'll see if I can commit uncomments cleanly.

> I added #ifdef YADE_OPENGL to GL-related ChainedCylinder stuff, instead
> of making all the classes dependent on opengl (there is no reason for
> that). I hope we will have no conflicts.
>    
It was not really a problem for me as I don't see everything kept in 
common/DataClass/shape/Cylinder at the end, which makes no sense, but in 
splitted files instead. Having one class per file is better for browsing 
sometimes. Perhaps Gl1_Cylinder and Gl1_Chained in the same file would 
not chock me, but there is clearly too much code in Cylinder.hpp/cpp now 
IMO).
> I will commit, along with an updated scripts/test/dispatcher-torture.py
> script, which gives this now (some dispatches are ambiguous because it
> gives all the functors available to the dispatcher):
>
> http://beta.arcig.cz/~eudoxos/temp/dispatch.html
>
>    
You will commit what, sorry? Yes, the problem with GL dispatching is 
that it's the only one you cannot tune from python script (putting 
wanted functors one by one). Actually this alone would let me solve the 
problem for now (as long as there are no single and chained cylinders in 
simulation at the same time - which will happen sooner that never...).

Bruno