[ODE] some quick notes on 0.6...

Geoff Carlton gcarlton at iinet.net.au
Thu Jun 22 00:21:11 MST 2006


Hi,
I agree that the convex and flat cylinder code needs alot of work, 
however they should really be in SVN ready for updating.  A proper 0.6 
might have branched and stripped them out, but I expect Jason had enough 
work on his hands as it is.  Good point though - any docs should clearly 
state that these two primitives are not in a working state.

Terrain is a different matter.  You mentioned the Y/Z issue.  Also it 
needs a callback mechanism for the common case where the app owns its 
own terrain data (in possibly a different or compressed format).   These 
issues directly affect the interface and even the number of new types 
(two types or one).  Would it be worth adding the legacy terrains just 
to deprecate them next release, thus ending up with three types?  
Perhaps, but nobody volunteered so it stayed out. 

Now 0.6 has been released I'm sure a cleaned up and unified terrain 
would be more than appreciated, it just needs somebody to step up.

Cheers,
Geoff



John Miles wrote:
> - The Win32 test programs don't run with ODE.DLL because export-dif.h 
> is missing the surrounding 'ifdef __cplusplus extern "C" { ... }' 
> clause.  I'm not sure why this worked before; it might be an artifact 
> of my attempt to keep using the 0.5 makefile with the new 0.6 stuff.  
> My build system is neither fish nor fowl at this point, so this may 
> not affect anyone but me... but I'd still suggest adding the extern 
> "C" clause to export-dif.h for consistency's sake.
>  
> - Along those lines, I'm not crazy about the loss of the MSVC 
> makefiles, although in general the new Win32 build setup seems really 
> nice.  Guess they had to be sacrificed for the greater good.  It was 
> nice to see the .def files gone!
>  
> - Wow, that convex collider primitive is broken.  Seriously not ready 
> for prime time.  Did anyone actually try it in test-boxstack?  Not 
> only do convex-convex collisions not work at all -- a situation 
> alluded to by the comments in convex.cpp -- but the convex pieces 
> promptly disable themselves after hitting the ground, and aren't 
> re-enabled when boxes hit them.  (Sphere collisions seem to re-enable 
> them, but not boxes.)
>  
> - I thought flat cylinders weren't going to be promoted to a 
> first-class type until they could collide with each other...?
>  
> In general I don't think much of including half-baked primitives like 
> the convex-hull and flat-cylinder types in the core library while the 
> dTerrainY/dTerrainZ contributions still have to be integrated by 
> hand.  (And that process is not becoming any easier, as Benoit's ODE 
> 0.039-based instructions become more and more dated.)  The argument 
> that they aren't worthy of promotion to the core library due to their 
> hardwired Y-up/Z-up nature falls flat on its face, given the twitchy 
> code that apparently *is* being promoted these days.
>  
> Just some random thoughts.  Despite the major improvements in the new 
> build process, this release is not likely to leave ODE newbies with 
> a good first impression.
>  
> -- jm
>  
> ------------------------------------------------------------------------
>
> _______________________________________________
> ODE mailing list
> ODE at q12.org
> http://q12.org/mailman/listinfo/ode
>   
> ------------------------------------------------------------------------
>
> No virus found in this incoming message.
> Checked by AVG Free Edition.
> Version: 7.1.394 / Virus Database: 268.9.2/372 - Release Date: 21/06/2006
>   


More information about the ODE mailing list