Box2D Forums

Now that 2.0 is out I would like to open up Box2D for community patches. If you have any C++ patches for 2.0, please post them to this thread or send me PM.

I'm looking for:
- projects/makefiles for other compilers/platforms
- source code fixes for othe compilers/platforms
- bug fixes
- features
- comment typo fixes or additions
- manual typo fixes or additions

Please keep your patches as small as possible. For example, please don't post patches that change a bunch of white space or rename lots of variables.

I would like to do this in a more organized fashion than in the past because I lost track of some patches and also forgot to give everyone credit on the contributers page: http://www.box2d.org/contributors.html

XCode 3.0 project for Box2D 2.0

Added new files to project
Fixed missing enable C++ rtti for debug build.

By patch you mean something like an automated script which will change the source files? If so, how can I make one when I have all the changes marked in a blocks of //BLA ... //~BLA ? I am not very expirienced with this

SVN can automatically create a patch file based on your changes. You'll have to work against the SVN version. If you run windows, I recommend using TortoiseSVN, it works beautifully.

Oh, I didn't know it altough I use TortoiseSVN already:)
So I suppose I should delete these //BLA comments or they appear in the code after patching, right?

Hi,

I noticed the latest rev adds a user specific project file:
Build/Xcode/Box2D.xcodeproj/grrussel.pbxuser
You may want to remove it, since it's of no use unless you are grrussel.

Also the testbed doesn't build on the Mac.

I've attached a patch. This also includes a freeglut ifdef for
glutMouseWheelFunc, which isn't in OSX's glut.

Cheers,
Frank.

Frank: thanks, I applied the patch.

BTW, why can't the Mac build freeglut and use that instead of glut?

Certainly possible, but freeglut currently doesn't support native OpenGL on the Mac.
You could build against X11/GL. The X11 dependency means the X server has to be running and prior to 10.5 X11 was an optional install.
Much easier to ifdef one line...

F.

I have a pretty large patch that includes:

1. Fixes to includes for testbed (the includes currently reference "freeglut/gl/glut.h" when the freeglut in Contrib uses "freeglut/GL/glut.h")
2. Gnu Makefiles to build the Contributed software (glui and freeglut), four versions of the library (float, fixed, nintendo DS float and nintendo DS fixed), and two versions of the testbed (float and fixed). You should be able to build things separately or build everything from one top level Makefile, including this very patch!
3. ewjordan's concave polygon decomposition code
4. Fixed Point implementation that redefines the float32 type to be "Fixed". This change requires all references to floating point stuff like FLT_EPLISON, FLT_MAX, etc. to change to something more generic so that they can be redefined between the float and fixed point version. In this port both the fixed and the floating point versions can be built, based on a compile-time flag. Care was taken so that the floating point version should suffer little or no performance degradation when compiled.
5. Nintendo DS version of the fixed point implementation that includes use of the special sqrt and long long division hardware.

The performance of the fixed point is currently not perfect, but usable. All of the testbed examples will run and not assert, but there are clear differences in performance between the float and fixed: biped (balls go through walls, grabbing on guy will cause havoc), car (grabbing onto the car will sometimes send it up in the air) and pyramid (the top square does not fly off and it takes a long time for things to settle to blue, but they do eventually).

This patch is a joint effort by 0xtobw and myself. I am including it here more to start a discussion on what pieces of this will be included an how they might be included. Like I said, it is a big patch with many inter-related pieces.

kavaler: I haven't looked at the patch yet, but here's a few comments:

- First of all, thanks for putting this together. I'm sure a lot of people are interested in the fixed point version.
- I believe Eric (ewjordan) will be integrating the convex decomposition himself (he has SVN access).
- How much does the fixed point code affect the floating point code? I'd rather not complicate the main code line to support fixed point. Mainly because I don't plan to support it directly (not enough time, etc).