libODE and GIMPACT vs OPCODE (car jumps around? look here)

[joevenzon]

R1845 has support for ODE compiled with GIMPACT. If your OPCODE-compiled ODE makes your car jump around or you have other weird collision-related bugs (road is bumpy, fall through road, etc), then we'll need to add an option to the scons files to set a #define for LIBODE_OPCODE, which should fix the problem (see track.cpp line 947).

[abs1nth]

yes this revision is quite unusable with OPCODE ODE as the car jumps around like crazy.

anyway, i've recompiled the osx framework for ODE with GIMPACT which seems to work fine

[xTs]

Hi,

with opcode the car is dancing here, too.

With gimpact i get a segfault and this:
http://paste.debian.net/37095

tried it with the debian lib of ode, with self compiled 0.8, with self compiled 0.8.1rc1 and the svn version. Always a segfault.

[joevenzon]

As I mentioned above it should be easy for someone who knows how to use scons (i.e., not me) to add an option to scons so that opcode works fine.

About the crash with gimpact enabled... that's very odd. It's crashing when trying to compute the AABB, which is just a bunch of if-then-else statements.

[alex25]

About the crash with gimpact enabled... that's very odd. It's crashing when trying to compute the AABB, which is just a bunch of if-then-else statements.

i bet he's running on a 64 bit machine. i am and getting the same crash. i'll try it later tonight on my 32 bit laptop and if it works i'll try to debug it. it's probably some funky size of pointer to type vs type problem but i haven't been able to find it yet (i am still trying to familiarize myself with the gimpact code).

--alex--

[alex25]

i bet he's running on a 64 bit machine. i am and getting the same crash. i'll try it later tonight on my 32 bit laptop and if it works

it works on my 32-bit laptop so i'm sure it's a 64-bit problem.

--alex--

[xTs]

alex: you are right, i'm running a 64bit machine. Must be related to this, cause i had an comparable error with another game.

Would be nice, if you can fix this. (I guess its a problem of ode not of vdrift)

[FFuser]

Code:

Index: SConstruct
===================================================================
--- SConstruct  (revision 1847)
+++ SConstruct  (working copy)
@@ -25,6 +25,7 @@
opts.Add(BoolOption('force_feedback', 'Enable force-feedback support', 0))
opts.Add(BoolOption('profiling', 'Turn on profiling output', 0))
opts.Add(BoolOption('efficiency', 'Turn on compile-time efficiency warnings', 0))
+opts.Add(BoolOption('opcode', 'Set this to one if you have ode compiled with opcode', 0))


#--------------------------#
@@ -316,6 +317,7 @@
       'scons NLS=0' To disable internationalization support (An english-only version will be built)
          'scons efficiency=1' to show efficiency assessment at compile time
          'scons profiling=1' to enable profiling support
+        'scons opcode=1' if you have ODE compiled with opcode
%s

Note: The options you enter will be saved in the file vdrift.conf and they will be the defaults which are used every subsequent time you run scons.""" % opts.GenerateHelpText(env))
@@ -361,6 +363,9 @@
         print 'This system is not configured for internationalized applications (no libintl.h). An english-only version will be built\n'
         env['NLS'] = 0

+if env['opcode']:
+       cppdefines.append("LIBODE_OPCODE")
+
env = conf.Finish()

use with "scons opcode=1", I do not have any idea how to detect it automagically

Btw: driving in the sand gives strange results (Opcode, 64 bit)

[joevenzon]

R1848 has the SConstruct change nathan posted above. I also fixed the driving off-track, I think (I didn't test it, but I'm pretty sure it's fixed).

[xTs]

Did a short test, works. Thanks!

[alex25]

Would be nice, if you can fix this. (I guess its a problem of ode not of vdrift)

i've given up for the time being, i recompiled ode without the gimpact code, set opcode to 1 in vdrift SConstruct. and now i can run vdrift again on my 64-bit machine. i'll keep looking into the gimpact code and if i come up with a solution i'll let you guys know. but first i need to come up with the patch for the skybox drawing offset.

--alex--

[alex25]

i've given up for the time being

i managed to find a patch:
http://fluid.mech.kogakuin.ac.jp/~minnie...amd64.diff
but it doesn't quite work, i still get a seg fault. might be a good starting point if somebody else wants to take a look at this.

--alex--

[xTs]

Uhm... maybe i know the problem already:

Someone didnt write this lib for multi architekture use:
+#define GINT int32_t
+#define GUINT uint32_t

Well, you should always use int32_t if you write code...

[alex25]

Uhm... maybe i know the problem already:

Someone didnt write this lib for multi architekture use:
+#define GINT int32_t
+#define GUINT uint32_t

Well, you should always use int32_t if you write code...

that's what the patch does (and then some). but it's not enough, you still get a seg fault.

--alex--

[xTs]

Will try the patch and have a closer look at ode...