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

----- sp4m  wrote:
> Hi David
> 
> Maybe I'm wrong but that mean that every time you upgrade torcs, you 
> have to re-configure it to get your desired settings back?

(not necessarily)

For Linux, TORCS loses the settings when the compile
time is newer than the 'settings' time stamps.
(Also, is there a guarantee that old settings will
always work with new versions?)

If Windows does not lose the settings when upgrading,
all you have to do is copy the the old settings directory
and rename it to the new version.
(You should drive a 'quick race' with the old version and
a different car to 'time stamp' the 'settings' before copying
or executing the new version)

If you want to copy the old settings for the new version
in Linux, you must also 'touch' the files in the copied
directory before executing the new torcs.
(this gives the files new time stamps)

eg.
find .torcs-1.3.8 -type f -exec touch '{}' \;

> Personally, I would probably uninstall such a program after two versions 
> and not touch it again. Unless there is an option to import the settings 
> from an earlier version.

Having different versions of torcs allows you to test drive them.
(especially if you are using a compiler and source code)

> 
> cheers
> Chris
> 
> On 22.05.2017 19:28, David Savinkoff wrote:
> > ----- David Savinkoff wrote:
> >> Hi Bernhard,
> >>
> >> This tiny patch changes ~/.torcs to look like ~/.torcs-1.3.8 ; as an example.
> >> One advantage is that multiple versions of torcs can be used. Please consider.
> >>
> >> Sincerely,
> >> David Savinkoff
> > 
> > This new patch does the same for Windows and Linux
> > (anyone, please test the windows version and report)
> > 
> > 
> > 

Hi David

Maybe I'm wrong but that mean that every time you upgrade torcs, you 
have to re-configure it to get your desired settings back?
Personally, I would probably uninstall such a program after two versions 
and not touch it again. Unless there is an option to import the settings 
from an earlier version.

cheers
Chris

On 22.05.2017 19:28, David Savinkoff wrote:
> ----- David Savinkoff wrote:
>> Hi Bernhard,
>>
>> This tiny patch changes ~/.torcs to look like ~/.torcs-1.3.8 ; as an example.
>> One advantage is that multiple versions of torcs can be used. Please consider.
>>
>> Sincerely,
>> David Savinkoff
> 
> This new patch does the same for Windows and Linux
> (anyone, please test the windows version and report)
> 
> 
> 
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> 
> 
> 
> _______________________________________________
> Torcs-users mailing list
> Tor...@li...
> https://lists.sourceforge.net/lists/listinfo/torcs-users
> 

----- David Savinkoff wrote:
> Hi Bernhard,
> 
> This tiny patch changes ~/.torcs to look like ~/.torcs-1.3.8 ; as an example.
> One advantage is that multiple versions of torcs can be used. Please consider.
> 
> Sincerely,
> David Savinkoff

This new patch does the same for Windows and Linux
(anyone, please test the windows version and report)

Hi Bernhard,

This tiny patch changes ~/.torcs to look like ~/.torcs-1.3.8 ; as an example.
One advantage is that multiple versions of torcs can be used. Please consider.

Sincerely,
David Savinkoff

The latest and greatest patches for torcs-1.3.7 and torcs-1.3.8.test1
Also a gcc compiler options patch for torcs-1.3.7 (not tested on 1.3.8)
The options patch allows you to override the default flags for better
performance (gcc -O3 ... that's what racing is about).

Hi,

So I found a solution:
It turns out that for Torcs 1.3.7 from binary installer the command line works as expected.
For the Torcs 1.3.4 plus scr patch, from examining the binary they switched the –r command line argument with –T. It is not documented anywhere and the change doesn’t allow for external config file usage anymore.

Hope this helps!

From: Liat Sless [mailto:lia...@gm...]
Sent: יום ב 08 מאי 2017 11:30
To: tor...@li...
Subject: [EXTERNAL] [Torcs-users] Torcs command line arguments problem - Windows

Hi,

I'm using torcs version for Windows, using the binary installer.

I haven't been able to run torcs without GUI, i.e., in text mode. The command wtorcs.exe -r <path to config file> opens up the GUI.
Also, the command  wtorcs -h which should show help doesn't do anything other than opening the GUI, which leads me to suspect there is a problem with command line arguments for torcs in Windows.

Both commands execute normally on Ubuntu.

Versions of torcs I checked on Windows (both don't respond to command line arguments):
Torcs 1.3.7 from binary installer
Torcs 1.3.4 plus scr patch from binary installer.

Windows version: 7.

How can I troubleshoot this?

Best,
Liat

Hi,

I remembered that a documentation package was once mentioned
by Bernhard. Here it is:

https://sourceforge.net/projects/torcs/files/api-docs/1.3.7/
https://sourceforge.net/projects/torcs/files/api-docs/1.3.7/torcs-1.3.7-apidoc.tar.bz2/download

To see the contents without navigating a huge directory:
(extract archive, then view)
tar -xjvf torcs-1.3.7-apidoc.tar.bz2
firefox torcs-1.3.7-apidoc/index.html


----- Konstantin Ritt wrote:
> Hi,
> 
> Thanks David for your hint.
> I am still struggling to find my way through to the right file(s) or
> method(s). I actually identified different files as being important to the
> rendering of the driving/game scenes, such as
> src/modules/graphic/ssggraph/grmain.cpp and grscreen.cpp where the actual
> drawing happens.
> I want to state again what I am trying to achieve, maybe someone can direct
> me into the right direction then.
> My goal is to create a FBO (frame buffer object) that is connected to a
> texture unit. To this texture unit I want to render the frames to perform
> post processing.
> I found that the central rendering happens in the following manner:
> 
>    - src/libs/raceengineclient/raceengine.cpp int ReUpdate(void) calls
>    ReInfo->_reGraphicItf.refresh(ReInfo->s) to update the current situation s
>    - src/modules/graphic/ssggraph/grmain.cpp contains this refresh
>    function. In this refresh function it is iterated through all active
>    screens by calling grScreens[i]->update
>    - src/modules/graphic/ssggraph/grscreen.cpp contains this update
>    function. camDraw is called (same class) which again calls grDrawScene()
>    - src/modules/graphic/ssggraph/grscene.cpp contains this grDrawScene
>    function, where the whole scene is passed to ssgCullAndDraw to be rendered.
> 
> In the beginning of grmain.cpp: refresh(Situation s) I set up the FBO. All
> rendering should be performed to the texture of the FBO then (tested all
> this with a smaller application).
> 
> After the call to grScreens[0]->update(s, grFps) I unbind the FBO and
> (currently) use the information from the FBO texture to render back to the
> screen. However, only some textures (of trees, grass etc.) make it to the
> texture in the FBO, but not the whole scene.
> 
> Any ideas what and where it could go wrong?
> 
> I can gladly share the implementation of the FBO as well, but as stated
> above, I tested the principle of the FBO with a simple application and got
> it working. I therefore think the problem lies in my limited understanding
> of the Torcs OpenGL pipeline and that I did not place it correctly.
> 
> Cheers and thanks for all your help,
> 
> Konstantin
> 
> 
> 
> 
> 
> On Thu, Apr 27, 2017 at 7:52 AM, David Savinkoff wrote:
> 
> > ----- Konstantin Ritt wrote:
> > > Hello,
> > >
> > > I just starting using TORCS for my masters thesis. I am currently trying
> > to
> > > gain some knowledge about usage of OpenGL and was wondering if you could
> > > give a quick shout on where and how the OpenGL pipeline is handled in
> > TORCS?
> >
> > I don't know, but see below.
> >
> > > What I am trying to do is to create another buffer to write frames to,
> > as a
> > > texture, before they are displayed on the screen. With these textures I
> > > want to perform postprocessing purely by using shaders to keep all memory
> > > and computation on the GPU in order to reduce impact on the runtime of
> > the
> > > simulation.
> > > I would highly appreciate it if you could give me some short advice to
> > save
> > > me time sorting through the immense framework of TORCS. Thanks a lot in
> > > advance!
> > >
> > > Kind regards,
> > > Konstantin
> >
> > Hi,
> > I found that the following files have 99.99% of torcs gl code in them:
> >
> > torcs/src/libs/tgfclient/gui.cpp
> > torcs/src/libs/tgfclient/img.cpp
> > torcs/src/libs/tgfclient/screen.cpp
> > torcs/src/libs/tgfclient/guibutton.cpp
> > torcs/src/libs/tgfclient/guiedit.cpp
> > torcs/src/libs/tgfclient/guifont.cpp
> > torcs/src/libs/tgfclient/glfeatures.cpp
> > torcs/src/libs/tgfclient/guiscrollist.cpp
> > torcs/src/libs/confscreens/mouseconfig.cpp
> > torcs/src/libs/confscreens/joystickconfig.cpp
> > torcs/src/libs/confscreens/controlconfig.cpp
> > torcs/src/libs/raceengineclient/raceengine.cpp
> > torcs/src/libs/raceengineclient/racegl.cpp
> > torcs/src/libs/client/splash.cpp
> > torcs/src/libs/musicplayer/musicplayer.cpp
> >
> > ( find . -type f -print0 | xargs -0 grep -Iw gl.* )
> >

Hi,

Thanks David for your hint.
I am still struggling to find my way through to the right file(s) or
method(s). I actually identified different files as being important to the
rendering of the driving/game scenes, such as
src/modules/graphic/ssggraph/grmain.cpp and grscreen.cpp where the actual
drawing happens.
I want to state again what I am trying to achieve, maybe someone can direct
me into the right direction then.
My goal is to create a FBO (frame buffer object) that is connected to a
texture unit. To this texture unit I want to render the frames to perform
post processing.
I found that the central rendering happens in the following manner:

   - src/libs/raceengineclient/raceengine.cpp int ReUpdate(void) calls
   ReInfo->_reGraphicItf.refresh(ReInfo->s) to update the current situation s
   - src/modules/graphic/ssggraph/grmain.cpp contains this refresh
   function. In this refresh function it is iterated through all active
   screens by calling grScreens[i]->update
   - src/modules/graphic/ssggraph/grscreen.cpp contains this update
   function. camDraw is called (same class) which again calls grDrawScene()
   - src/modules/graphic/ssggraph/grscene.cpp contains this grDrawScene
   function, where the whole scene is passed to ssgCullAndDraw to be rendered.

In the beginning of grmain.cpp: refresh(Situation s) I set up the FBO. All
rendering should be performed to the texture of the FBO then (tested all
this with a smaller application).

After the call to grScreens[0]->update(s, grFps) I unbind the FBO and
(currently) use the information from the FBO texture to render back to the
screen. However, only some textures (of trees, grass etc.) make it to the
texture in the FBO, but not the whole scene.

Any ideas what and where it could go wrong?

I can gladly share the implementation of the FBO as well, but as stated
above, I tested the principle of the FBO with a simple application and got
it working. I therefore think the problem lies in my limited understanding
of the Torcs OpenGL pipeline and that I did not place it correctly.

Cheers and thanks for all your help,

Konstantin





On Thu, Apr 27, 2017 at 7:52 AM, David Savinkoff <dsa...@te...> wrote:

> ----- Konstantin Ritt wrote:
> > Hello,
> >
> > I just starting using TORCS for my masters thesis. I am currently trying
> to
> > gain some knowledge about usage of OpenGL and was wondering if you could
> > give a quick shout on where and how the OpenGL pipeline is handled in
> TORCS?
>
> I don't know, but see below.
>
> > What I am trying to do is to create another buffer to write frames to,
> as a
> > texture, before they are displayed on the screen. With these textures I
> > want to perform postprocessing purely by using shaders to keep all memory
> > and computation on the GPU in order to reduce impact on the runtime of
> the
> > simulation.
> > I would highly appreciate it if you could give me some short advice to
> save
> > me time sorting through the immense framework of TORCS. Thanks a lot in
> > advance!
> >
> > Kind regards,
> > Konstantin
>
> Hi,
> I found that the following files have 99.99% of torcs gl code in them:
>
> torcs/src/libs/tgfclient/gui.cpp
> torcs/src/libs/tgfclient/img.cpp
> torcs/src/libs/tgfclient/screen.cpp
> torcs/src/libs/tgfclient/guibutton.cpp
> torcs/src/libs/tgfclient/guiedit.cpp
> torcs/src/libs/tgfclient/guifont.cpp
> torcs/src/libs/tgfclient/glfeatures.cpp
> torcs/src/libs/tgfclient/guiscrollist.cpp
> torcs/src/libs/confscreens/mouseconfig.cpp
> torcs/src/libs/confscreens/joystickconfig.cpp
> torcs/src/libs/confscreens/controlconfig.cpp
> torcs/src/libs/raceengineclient/raceengine.cpp
> torcs/src/libs/raceengineclient/racegl.cpp
> torcs/src/libs/client/splash.cpp
> torcs/src/libs/musicplayer/musicplayer.cpp
>
> ( find . -type f -print0 | xargs -0 grep -Iw gl.* )
>

Hello,

I am trying to run torcs on a headless server with GPU support (to train a deep neural network). Specifically, I am trying to install and run it on an Amazon EC2 p2.xlarge instance. I can’t get openGL/DRI to run on this setup, and hence torcs quits at startup (even though I am trying to run it in API only mode).

Can anyone please tell me how to work around the openGL dependency? 
If someone has any advice on how to run torcs on a headless server, I would greatly appreciate it.

I understand how to run torcs in API only mode (using the -r option) and that works correctly on my local machine. The problem is only on the headless server. The headless server is running Ubuntu 16.04 and torcs 1.3.7 with the scr-server patch.

Thanks in advance for the help.

Regards,
Abhinav

Hi,

I'm using torcs version for Windows, using the binary installer.

I haven't been able to run torcs without GUI, i.e., in text mode. The command wtorcs.exe -r <path to config file> opens up the GUI.
Also, the command  wtorcs -h which should show help doesn't do anything other than opening the GUI, which leads me to suspect there is a problem with command line arguments for torcs in Windows.

Both commands execute normally on Ubuntu.

Versions of torcs I checked on Windows (both don't respond to command line arguments):
Torcs 1.3.7 from binary installer
Torcs 1.3.4 plus scr patch from binary installer.

Windows version: 7.

How can I troubleshoot this?

Best,
Liat

Hi

Just from the output I would guess that you have to recompile plib with 
"-fPIC", see "... recompile with -fPIC".

Kind regards

Bernhard

On 05.05.2017 07:42, Nirant Kasliwal wrote:
> Hello,
>
> I am trying to build torcs on Ubuntu 14.04. I have a successful
> configure - as the attached config.log indicates.
> The make fails with the following message:
>
> /usr/bin/ld: /usr/lib/libplibjs.a(jsLinux.o): relocation R_X86_64_32
> against `.rodata.str1.1' can not be used when making a shared object;
> recompile with -fPIC
> /usr/lib/libplibjs.a: error adding symbols: Bad value
> collect2: error: ld returned 1 exit status
> make[4]: *** [libconfscreens.so] Error 1
> make[4]: Leaving directory
> `/home/nirant/torcs/torcs-1.3.4/src/libs/confscreens'
> make[3]: *** [subdirs] Error 1
> make[3]: Leaving directory `/home/nirant/torcs/torcs-1.3.4/src/libs'
> make[2]: *** [subdirs] Error 1
> make[2]: Leaving directory `/home/nirant/torcs/torcs-1.3.4/src'
> make[1]: *** [subdirs] Error 1
> make[1]: Leaving directory `/home/nirant/torcs/torcs-1.3.4'
> make: *** [restart] Error 2
>
>
> If I understand correctly, is this a libplib issue? I have searched
> using the error message and few libplib based ones. Attempts to
> recompile with -fPIC flag did not work.
> What can I do to build torcs?
>
> Thanks,
> Nirant
>
>
>
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
>
>
>
> _______________________________________________
> Torcs-users mailing list
> Tor...@li...
> https://lists.sourceforge.net/lists/listinfo/torcs-users
>

Hello,

I am trying to build torcs on Ubuntu 14.04. I have a successful configure -
as the attached config.log indicates.
The make fails with the following message:

/usr/bin/ld: /usr/lib/libplibjs.a(jsLinux.o): relocation R_X86_64_32
against `.rodata.str1.1' can not be used when making a shared object;
recompile with -fPIC
/usr/lib/libplibjs.a: error adding symbols: Bad value
collect2: error: ld returned 1 exit status
make[4]: *** [libconfscreens.so] Error 1
make[4]: Leaving directory
`/home/nirant/torcs/torcs-1.3.4/src/libs/confscreens'
make[3]: *** [subdirs] Error 1
make[3]: Leaving directory `/home/nirant/torcs/torcs-1.3.4/src/libs'
make[2]: *** [subdirs] Error 1
make[2]: Leaving directory `/home/nirant/torcs/torcs-1.3.4/src'
make[1]: *** [subdirs] Error 1
make[1]: Leaving directory `/home/nirant/torcs/torcs-1.3.4'
make: *** [restart] Error 2


If I understand correctly, is this a libplib issue? I have searched using
the error message and few libplib based ones. Attempts to recompile with
-fPIC flag did not work.
What can I do to build torcs?

Thanks,
Nirant

Hi,

There are only two possibilities to fix this one because you either
add or subtract a vector component to get the right answer.
I suggest testing both to see which one is better (unless you
decide to trace the value to its origin for correctness - note also
that a steering wheel can be turned 1/2 rad to the right, or
the ground can be steered 1/2 rad to the left depending on
your frame of reference)

The car should power slide better.