torcs-devel Mailing List for TORCS - The Open Racing Car Simulator

Hi, All.

I'm working now on breaking the "10 cars per robot" and the "10 human players" 
limits (that is the "10 interfaces per dynamically loadable module" limit).

For that, I "simply" changed the signature of the init function of a module :

from : typedef int (*tfModInfo)(tModInfo*);
to   : typedef int (*tfModInfo)(std::vector<tModInfo>&);

This works like a charm under Linux GCC 4.2 (I added an 11th car to BT,
as an exemple), but under Windows XP VC6 SP6, it's a bit harder ;-)

After some searches on the internet, it seems that to be able to pass
std::vector across DLLs, ALL the code MUST be linked against
the "Multi-threaded DLL" (+ debug if relevant) run-time library (MSVCRT)
(ALL means ANY DLL, static lib and executable).

So, question # 1 :
   is it a problem that MSVCRT.DLL (and may be 1 or 2 others ?)
   be distributed with Torcs binaries under Windows ?
   (as far as I remember, it is not part of Win XP installation).

And, question # 2 :
   for the moment, I can't make it work (crashes) under Windows because
   (I suspect) of the static libs and DLLs that are shipped as binaries
   with Torcs (under src/windows/lib and dll) : may be one of them, at least,
   was not linked against the Multi-thread DLL ...
   So, unless you know I'm completly wrong, could you tell me :
   - the exact versions and name of each used library (GLUT or freeGLUT ?
     plib, modified? open AL, zlib, libpng)
   - what modifications, if any, where made in the VC6 project files
     of these libraries to build them.

And if you have any other suggestions to fix the Windows build ...

Thanks in advance for your help.

Jean-Philippe

PS: If needed, I can also give you the patch ...

After updating from CVS and using your proposed flag helped me to compile the whole project in VC6.

Additionally I only had to deactivate "use precompiled headers" in the project options for simuv3 to make it work.

Now only one warning is left:
C:\Programme\torcs\torcs\src\modules\simu\simuv3\aero.cpp(243) : warning C4305: 'argument' : Verkuerzung von 'const double' in 'float'
***************************

Torcs does not crash anymore if I start a race with the simuv3-engine activated. So everything seems to work.


Many thanks for helping me.
Bernhard


Christos wrote:
 Re: [Torcs-users] Can not run torcs with simuv3 under w32
> 
> Oh, OK, it's C++ incompatibility flag in VC.  However,
> you can compile with /GR to fix that.
> 
> Question to all developers:
> 
> Is there a reason for not allowing RTTI? The 'performance' claim seems weak.
> 
> Would the code work without /GR if I used a static cast?
> 
> 
> 

Hi,

I have commited the patch from Jean-Philippe to support wheels with grid 
shifter.
I also updated the setup_win32* files for all the new data.
I commited Visual C++ 2005 project files for those who are interrested.

Have fun,
Eric.

Hi,

I followed the idea of having 4 wheel objects for the different
speed for motion blur.
The root name can be specified in the "Graphic Objects" section
in the attribute "3d wheel". The default value is "wheel".
The objects will have the following names :
<root name><0-3>.acc
so by default:
wheel0.acc for rotation < 20 rad/s (25 km/h for a wheel with a diameter 
of 70 cm)
wheel1.acc for rotation < 40 rad/s (50 km/h)
wheel2.acc for rotation < 70 rad/s (88 km/h)
wheel3.acc for rotation > 70 rad/s

Eric.


Andrew Sumner a écrit :
>
> Hi Eric,
>
> Thanks for evaluating, refining and committing the patch :)
>
> I really love the automatic resize - I would have done it that
> way myself if I'd known how.  It not only makes it tons easier
> to create the 3D wheel models, it means the existing wheel size
> parameters in the XML are still used which is excellent.  I presume
> you resized the wheel models for the clkdtm and sc-f1 before
> committing them?
>
> I do agree with Jean-Phillippe - the model names etc should be
> specified in the car's XML rather than being hardcoded.  I didn't
> do it that way in my patch in order to keep it as small as possible
> to make it trivial to test.
>
> cheers
> Andrew
>
>
> -----Original Message-----
> From: Eric Espie [mailto:eri...@fr...]
> Sent: Tuesday, 25 March 2008 5:23 AM
> To: Andrew Sumner
> Cc: tor...@li...
> Subject: Re: [Torcs-devel] 3D wheel patch revisited
>
>
> Hi all,
>
> I have made some little modifications to the patch to allow the
> automatic resize of the wheel.
> The provided wheel object in the ac3d file should be of size 1m x 1m x 1m
> The object is dynamicaly resized to the values provided in the car's xml
> description.
> No more needs for different objects for front/rear wheels IMHO so I
> 'cut' that part from the code.
>
> I commited the code and wheels for clkdtm and sc-f1 (the initial texture
> has been changed because of trade mark)
> I have also added a wheel3d.png file in the inferno/4 directory to
> illustrate the texture customization by driver.
>
> Have fun,
> Eric.
>
> Andrew Sumner a écrit :
> >
> > Eliam submitted a patch over a year ago to allow use of 3D wheel
> > models in TORCS.  I've
> > revisited the code, reducing the changes to just one file and adding
> > some improvements.
> >
> > With this version, no changes need be made to car XML files - if a 3D
> > wheel model is present
> > it'll be used, otherwise TORCS will revert to the standard generated
> > wheels.  As the 3D ones
> > are .acc files, the wheels will darken when driving into shadows in
> > the same way as cars are
> > darkened.  I've included screenshots and examples for the porsche-gt1
> > and sc-f1 in the zip,
> > along with the sourcecode changes.
> >
> > Please let me know what you think - I'd really like to see this
> > feature make its way into TORCS
> > if possible.
> >
> > http://www2.webng.com/locus/files/3dwheel.zip
> >
> > cheers
> > Andrew
>
> ------------------------------------------------------------------------
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by: Microsoft
> Defy all challenges. Microsoft(R) Visual Studio 2008.
> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
> ------------------------------------------------------------------------
>
> _______________________________________________
> Torcs-devel mailing list
> Tor...@li...
> https://lists.sourceforge.net/lists/listinfo/torcs-devel
>   

Dear everybody,

Now that we are approaching the release of Torcs 1.3.1, with the brand new 
trb1 category, it would be fitting to have the robots everybody loves 
driving the brand new cars. I don't know if anybody else is doing this, but 
I could add a berniw robot (based on current CVS berniw2) that drives all 
car[1-7]-trb1 with some basic setups. Not the most optimised in the world, 
but enough to have downforce in twisty circuits and top speed in fast 
circuits (e-track-4, forza, the ovals and to a lesser extent wheel-2). A 
simple copy command could make Christos' robot also drive those cars with 
decent speed.

I'm also planning to raise the Gear shift threshold slightly, since most 
cars there have maximum power at 8000 rpm and the redline at 8500 rpm. It's 
just a parametre change, so it shouldn't be too troublesome.

Any comments will be appreciated. Regards,

Miguel

Hi Eric,

Thanks for evaluating, refining and committing the patch :)

I really love the automatic resize - I would have done it that
way myself if I'd known how.  It not only makes it tons easier
to create the 3D wheel models, it means the existing wheel size
parameters in the XML are still used which is excellent.  I presume
you resized the wheel models for the clkdtm and sc-f1 before
committing them?

I do agree with Jean-Phillippe - the model names etc should be
specified in the car's XML rather than being hardcoded.  I didn't
do it that way in my patch in order to keep it as small as possible
to make it trivial to test.

cheers
Andrew


-----Original Message-----
From: Eric Espie [mailto:eri...@fr...]
Sent: Tuesday, 25 March 2008 5:23 AM
To: Andrew Sumner
Cc: tor...@li...
Subject: Re: [Torcs-devel] 3D wheel patch revisited


Hi all,

I have made some little modifications to the patch to allow the 
automatic resize of the wheel.
The provided wheel object in the ac3d file should be of size 1m x 1m x 1m
The object is dynamicaly resized to the values provided in the car's xml 
description.
No more needs for different objects for front/rear wheels IMHO so I 
'cut' that part from the code.

I commited the code and wheels for clkdtm and sc-f1 (the initial texture 
has been changed because of trade mark)
I have also added a wheel3d.png file in the inferno/4 directory to 
illustrate the texture customization by driver.

Have fun,
Eric.

Andrew Sumner a écrit :
>
> Eliam submitted a patch over a year ago to allow use of 3D wheel 
> models in TORCS.  I've
> revisited the code, reducing the changes to just one file and adding 
> some improvements.
>
> With this version, no changes need be made to car XML files - if a 3D 
> wheel model is present
> it'll be used, otherwise TORCS will revert to the standard generated 
> wheels.  As the 3D ones
> are .acc files, the wheels will darken when driving into shadows in 
> the same way as cars are
> darkened.  I've included screenshots and examples for the porsche-gt1 
> and sc-f1 in the zip,
> along with the sourcecode changes.
>
> Please let me know what you think - I'd really like to see this 
> feature make its way into TORCS
> if possible.
>
> http://www2.webng.com/locus/files/3dwheel.zip
>
> cheers
> Andrew

Hi all,

I have made some little modifications to the patch to allow the 
automatic resize of the wheel.
The provided wheel object in the ac3d file should be of size 1m x 1m x 1m
The object is dynamicaly resized to the values provided in the car's xml 
description.
No more needs for different objects for front/rear wheels IMHO so I 
'cut' that part from the code.

I commited the code and wheels for clkdtm and sc-f1 (the initial texture 
has been changed because of trade mark)
I have also added a wheel3d.png file in the inferno/4 directory to 
illustrate the texture customization by driver.

Have fun,
Eric.

Andrew Sumner a écrit :
>
> Eliam submitted a patch over a year ago to allow use of 3D wheel 
> models in TORCS.  I've
> revisited the code, reducing the changes to just one file and adding 
> some improvements.
>
> With this version, no changes need be made to car XML files - if a 3D 
> wheel model is present
> it'll be used, otherwise TORCS will revert to the standard generated 
> wheels.  As the 3D ones
> are .acc files, the wheels will darken when driving into shadows in 
> the same way as cars are
> darkened.  I've included screenshots and examples for the porsche-gt1 
> and sc-f1 in the zip,
> along with the sourcecode changes.
>
> Please let me know what you think - I'd really like to see this 
> feature make its way into TORCS
> if possible.
>
> http://www2.webng.com/locus/files/3dwheel.zip
>
> cheers
> Andrew
>
> ------------------------------------------------------------------------
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by: Microsoft
> Defy all challenges. Microsoft(R) Visual Studio 2008.
> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
> ------------------------------------------------------------------------
>
> _______________________________________________
> Torcs-devel mailing list
> Tor...@li...
> https://lists.sourceforge.net/lists/listinfo/torcs-devel
>   

Annick et Jean-Philippe wrote:
> Hi, all.
> 
> Since I've been playing with Torcs, I've always been asking why
> my CPU fan was making so much noise (after a while) while I was only
> moving around in the configuration screens ... OK during the race, to get
> as many frames as possible per second, but in config screens ...
> 

Hm.. probably we can make the CPU go to sleep when waiting for the next
vertical blank too.  I noticed that when I have vblank turned on on my
old system, CPU % went down from 100% to something a bit less (since
anyway TORCS has nothing to do until it can draw the next frame)...
but that probably has more to do with OpenGL.

Hi, all.

Eric Espie discovered a few days ago that my "Gear changing mode configuration"
patch introduced a bug that breaks the mouse and joystick calibration process
(in the associated joystickconfig and mouseconfig screens).

Thanks to him, and blame on me ;-)

As an apologize for this reprehensible omission ;-) ,
here is the fixed version, and as a real patch this time !

To apply it (on a fresh cvs sandbox) :

cd <the directory that contains src>
patch -p1 < <...>/gearChangeModeGUI.patch

Tested under Linux (x86-64 GCC 4.2) and Windows XP SP2 (VC 6 SP 6)
with a mouse and a Logitech G25 Wheel
(but under Linux, no real support available for the moment from the kernel
   for the G25 grid gear lever, so the test was done with one keyboard key
   for each gear).

The patch also fixes a little bug that made the "Autoreverse" editbox
visible at screen load time, even if the current/selected player
had choosen a grid or sequential gear changing mode.

Hope this time is the right one ...

Jean-Philippe.

Hi, all.

Since I've been playing with Torcs, I've always been asking why
my CPU fan was making so much noise (after a while) while I was only
moving around in the configuration screens ... OK during the race, to get
as many frames as possible per second, but in config screens ...

So this is a small patch that puts the CPU to sleep for a (very short) while
each time it executes the glut idle functions that are used in config screens.
Of course (I had a hard time to make it, though ;-), no such rest
in the race screen !

I find it usefull in the following ways :
- more silence during configuration screens,
- globally cooler CPU during game, as it can have a breath
   between races, even for some small seconds,
- energy efficiency (in these days of global warming ... ;-)

To apply it :

cd <the directory that contains src>
patch -p1 < <...>/coolCPUDuringConfig.patch

Tested under Linux (x86-64 GCC 4.2) and Windows XP SP2 (VC 6 SP 6).

Please, let me know if you find it usefull, and tell me about improvements
you would think of ... And especially what you think about the sleep delay
I put in libs/tgfclient/gui.cpp::GfuiIdleCPUCooler (and other Idle* functions
in libs/confscreens/... and the fact that it is a hard-coded constant :
of course, it's bad, ... but, to go further, should it be variable
for any reason ? I can't see any but I may miss something.

Thanks in advance.

Jean-Philippe.

Hi, Andrew, and all.

 > Eliam submitted a patch over a year ago to allow use of 3D wheel models
 > in TORCS.  I've
 > revisited the code, reducing the changes to just one file and adding
 > some improvements.
 >
 > With this version, no changes need be made to car XML files - if a 3D
 > wheel model is present
 > it'll be used, otherwise TORCS will revert to the standard generated
 > wheels.  As the 3D ones
 > are .acc files, the wheels will darken when driving into shadows in the
 > same way as cars are
 > darkened.  I've included screenshots and examples for the porsche-gt1
 > and sc-f1 in the zip,
 > along with the sourcecode changes.
 >
 > Please let me know what you think - I'd really like to see this feature
 > make its way into TORCS

I've tested your patch, and it just works !
I did not notice any frame rate (bad) effect, even on my poor ATI X300 device !

So, I agree with you : this is an intersting feature,
even if it's especially visible for F1 cars
... and we lack of these in Torcs ;-)

By the way, as far as code is concerned, my opinion is that the description
of the 3D wheel model should all be in the car XML files, with the fewest
mandatory assumptions on the file names, and the associated tests and switches
in the code.

But thank you for having made the patch that simple for testing purpose.

Jean-Philippe.

Hello, all (again ;-).

I would like to propose you the following patch on players and control
configuration screens.

Its main purpose is to try and make a little simpler and clearer the process of
configuring the gear changing mode and the associated controls.

Please, feel free to tell me if you consider it an improvement, or not at all
  ;-). Don't hesitate to tell me if you dislike some details or more ...

To do that :
- in the "players configuration" screen,
   * the "transmission" field is renamed to "Gear change",
     and has now 3 possible values : auto, sequential and grid
   * the "autoreverse" field is only visible in "auto" gear change mode
   * the car category and car name have been move down the screen
     (one of my next goals, if you agree, would be to remove them
      from this screen, and be able to choose ones car in race screens ...
      but that's another story)
- in the "control configuration" screen,
   * all the controls that are not dedicated to the gear change
     have been gathered on the left column/half screen
   * the right column is now reserved to the gear changing controls
     (reverse, neutral, up shift, down shift, 1st gear, 2nd gear ... 6th gear)
   * this right column is different for each gear change mode :
     - auto : only reverse, neutral, up shift and down shift,
     - sequential : same
     - grid : only reverse, neutral, 1st gear, 2nd gear ... 6th gear.
- behind the scene, 2 settings are automatically configured :
   * "sequential shifter allows neutral" is set to yes in "sequential" mode
     if no control is configured for neutral gear (no otherwise)
   * "release gear goes neutral" is set to yes in "grid" mode
     if no control is configured for reverse gear (no otherwise).

The "patch" changes only 5 files. I give you the whole files (not real patches),
but if you prefer real patch files, please feel free to ask me
(but in that case, could you tell me what diff options you generally use ?)

So to "install" the "patch", that modifies only 5 files, on a fresh CVS tree :

cd src
tar xvfz GearChangeMode.tgz

Or you can carefully patch the individual files with a diff/merge tool ...

Thanks in advance for your help.

Jean-Philippe.

Hi all.

No a big pb ;-), but one day or another, it will be a compilation error :
nearly everywhere in torcs code, my gcc 4.2.2 issues the following warning :

Warning : deprecated conversion from string constant to 'char*'

Has anyone already fixed this future pb ?

If not, may I suggest to change nearly everywhere the way strings are passed
as parameters from "char *" to "const char *" ?
Of course, I'm conscious that its would change nearly every source file in 
Torcs, so I propose to try and build a script to automatically
and intelligently make the changes (if possible) ; that way, we would be free to 
simply test it home, and then one day decide when we want that massive commit 
(right before a release ?).

Thanks for you input.

Jean-Philippe.

Hi all,

I have made two small patches.

First, MAX_GEARS is the constant number which tells Torcs what the maximum 
number of gears is a car can have (including reverse and neutral). But the 
definition of MAX_GEARS isn't the only place where this is defined. Some 
static char* definitions also implicit define the maximum number of gears. 
The patch "max_gears.patch" solves this.

It maybe also solves the problem that the gear numbers are not displayed 
correctly if the car has no reverse.

Secondly, someone asked for the average fps for automatic testing purposes. I 
made a simple patch which prints the average fps of a race to the console.

Regards,

Mart