I, too, was attracted by the pure Python approach and the bindings of this engine.  But I struggled a lot with CS both to build it and to get started with it. A week later I was really frustrated but I had most of the pieces together and was digging into the docs.  I didn't notice any performance issues with the demos, though.  Python Mac functionality is there, but not without problems.

I then found that OGRE also possesses Python bindings. The down side to that is that OGRE's bindings don't yet work on Mac, but I found OGRE to be a lot simpler to build (less than 2 hours as opposed to 2 days for CS) and use.  I didn't have to get anything from CVS (the CS snapshots are painless, but the dependencies are not). The community is very large and very active with gobs of answers archived in the forums. The wiki is fairly well maintained, the docs are mostly up to date.

I'm not saying OGRE is better, because both engines are good and they do different things (CS provides more game infrastructure out of the box whereas OGRE focuses only on rendering). CS has a lot of potential--especially with the CEL module--but it's currently too hard to get into IMHO. And that's holding back the building up of a community. There's just not enough of a community to feel comfortable if I run into trouble in the middle of a commercial project. The devs are very hard-working and dedicated, that's obvious. What they've done is really amazing. With more "marketing" and some polish on the "getting started" phase, it will really shine.  I will watch Crystal Space's progress with great interest.

Thanks for the info.

You can fire rockets in the walktest?! 

The terrain demo uses shift for faster movement, maybe you could look there for an example?

Changing the config file did make it stop complaining, but there's no sound.  Is there supposed to be sound in the walktest demo?

I tried the sndtest demo but it complained about not initializing the sound and then crashed on exit.  I couldn't find any .cfg file for that one in the data/ subdir.

Ok I'll find the one from CVS.  Thanks for helping!

I seem to be missing the header file "soundcard.h" (according to configure.log).  It's not on my system at all, and apparently I have no sound and test apps report failure to initialize the sound system.  Are there any known issues with this?

I got the snapshot, and configure now says it compiled with OPCODE support, but it still says ODE is not new enough.  It appears to have something to do with the dGeomTriMeshGetTriangleCount() call. Any help appreciated.

configure:35610: checking for ode-config
configure:35639: result: no
configure:35954: checking for libode
configure:36043: g++ -o conftest     -I/usr/local/include    -L/usr/local/lib conftest.cc -framework ode  -lmx  >&5
ld: can't locate framework for: -framework ode
configure:36049: $? = 1
configure: failed program was:
| /* confdefs.h.  */
|
| #define PACKAGE_NAME "crystalspace"
| #define PACKAGE_TARNAME "crystalspace"
| #define PACKAGE_VERSION "0.99"
| #define PACKAGE_STRING "crystalspace 0.99"
| #define PACKAGE_BUGREPORT "crystal-main@lists.sourceforge.net"
| #ifdef __cplusplus
| extern "C" void std::exit (int) throw (); using std::exit;
| #endif
| #define STDC_HEADERS 1
| #define HAVE_SYS_TYPES_H 1
| #define HAVE_SYS_STAT_H 1
| #define HAVE_STDLIB_H 1
| #define HAVE_STRING_H 1
| #define HAVE_MEMORY_H 1
| #define HAVE_STRINGS_H 1
| #define HAVE_INTTYPES_H 1
| #define HAVE_STDINT_H 1
| #define HAVE_UNISTD_H 1
| #define HAVE_CMATH 1
| #define HAVE_LIBMX 1
| #define __CARBONSOUND__
| #define X_DISPLAY_MISSING 1
| /* end confdefs.h.  */
| #include <ode/ode.h>
| int
| main ()
| {
| dWorldCreate()
|   ;
|   return 0;
| }
configure:36043: g++ -o conftest     -I/usr/local/include    -L/usr/local/lib conftest.cc -lode  -lmx  >&5
configure:36049: $? = 0
configure:36053: test -z
          || test ! -s conftest.err
configure:36056: $? = 0
configure:36059: test -s conftest
configure:36062: $? = 0
configure:36101: result: yes
configure:36124: checking if libode is built with OPCODE support
configure:36216: g++ -o conftest     -I/usr/local/include    -L/usr/local/lib conftest.cc  -lode -lmx  >&5
configure:36222: $? = 0
configure:36226: test -z
          || test ! -s conftest.err
configure:36229: $? = 0
configure:36232: test -s conftest
configure:36235: $? = 0
configure:36275: result: yes
configure:36445: checking if libode is sufficiently recent
configure:36537: g++ -o conftest     -I/usr/local/include    -L/usr/local/lib conftest.cc  -lode -lmx  >&5
conftest.cc: In function `int main()':
conftest.cc:30: error: `dGeomTriMeshGetTriangleCount' undeclared (first use
   this function)
conftest.cc:30: error: (Each undeclared identifier is reported only once for
   each function it appears in.)
configure:36543: $? = 1
configure: failed program was:
| /* confdefs.h.  */
|
| #define PACKAGE_NAME "crystalspace"
| #define PACKAGE_TARNAME "crystalspace"
| #define PACKAGE_VERSION "0.99"
| #define PACKAGE_STRING "crystalspace 0.99"
| #define PACKAGE_BUGREPORT "crystal-main@lists.sourceforge.net"
| #ifdef __cplusplus
| extern "C" void std::exit (int) throw (); using std::exit;
| #endif
| #define STDC_HEADERS 1
| #define HAVE_SYS_TYPES_H 1
| #define HAVE_SYS_STAT_H 1
| #define HAVE_STDLIB_H 1
| #define HAVE_STRING_H 1
| #define HAVE_MEMORY_H 1
| #define HAVE_STRINGS_H 1
| #define HAVE_INTTYPES_H 1
| #define HAVE_STDINT_H 1
| #define HAVE_UNISTD_H 1
| #define HAVE_CMATH 1
| #define HAVE_LIBMX 1
| #define __CARBONSOUND__
| #define X_DISPLAY_MISSING 1
| /* end confdefs.h.  */
| #include <ode/ode.h>
| int
| main ()
| {
| dGeomTriMeshGetTriangleCount(0);
|   ;
|   return 0;
| }
configure:36596: result: no

Cool, thanks for such a complete reply. 

I think it's been mentioned but have you guys considered using SCons for builds instead of Jam?  Since there's already a strong connection to Python in the code/engine, it removes the need to learn yet another system/pseudo-language to be able to manipulate the build process.  Just a friendly idea.

Is there a script or other way to remove the CS files that get installed by jam install?  Or maybe there's at least some history or log of what got installed and where?

In the case of the Mac version, I had to copy the data/ directory into the app bundle.  It worked after that.

I did notice it wasn't rendering the woman correctly, though. Maybe it's the sample code, but she didn't appear to respect any z-depth since she appeared "above" all the particle effects (otherwise it all looked normal).  Is this a known issue?

Ok I definitely did that, but it still reports that ODE is not sufficiently new enough.  Maybe I missed an error message somewhere. I'll get the updated snapshot and try again.

It's not clear that pre-built prerequisite packages would make much sense on Mac since most developers are already using Fink or DarwinPorts...

That makes sense. Perhaps this should be noted in the docs for Mac?

It has been my intention to add to the CS documentation...

Can anyone edit the docs? I could update it some once I get everything working and figure out all the pieces.  I would at least be willing to help document Cal3D and ODE since those are currently the hardest to get working and are required for any non-trivial work.

Ah, ok, thank you!

Specifically, in the options file (which the README/INSTALL instructs you to edit), you set the OPCODE variable to 'yes'.

I had edited the ode/config/user-settings file, where the OPCODE_DIRECTORY setting resides. Is that the one you mean? I can't find any other mention of Opcode in the ODE config files. Opcode did build successfully, because I see the .o files in the OPCODE and Ice subdirs.

I may get the latest snapshot and see if configure.log has a better message for me.  Thanks for helping.

Configure currently sets things up to use the Mac OSX 10.3 default Python which is 2.3.  But I have installed 2.4, so how can I build CS to use 2.4 instead?  Is it a matter of just changing a path somewhere?  Any help, please?