visualpython-users

[Visualpython-users] no-go with new (and old!) visual on linux

[Martin G. <mar...@co...>]

I am running Debian testing, but got boost-python-dev and dependencies
from unstable (since only unstable has boost 1.31 at this time).

Compilation of visual-3.0 seemed to go fine.  I used
PYTHON=/usr/bin/python ./configure --prefix=/usr/local
(/usr/bin/python is a link to python2.3, and I want all locally built
software in /usr/local)

Importing visual in a python2.3 shell goes OK, but I get a segfault on
trying to do anything that generates a display on-screen.  I see a display 
window appear, then immediately close.  Simply instantiating a display
(e.g. firstdisplay=visual.display()) does not throw a segfault.

What's worse, my old visual setup (python2.2, visual-2.1.9) no longer
works:  
Python 2.2.3+ (#1, Jun 20 2004, 13:32:48)
[GCC 3.3.4 (Debian)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import visual
Visual-2003-10-05
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
  File "/usr/local/lib/python2.2/site-packages/visual/__init__.py", line 17, 
in ?
    import cvisual
ImportError: /usr/local/lib/python2.2/site-packages/cvisualmodule.so: 
undefined symbol: __pure_virtual

I don't see how this could be the fault of installing the new visual--could
it have come about due to an upgrade of python2.2 since I built that
version of visual?

Thanks for your attention.

-- 
Martin Gelfand

Associate Professor.............Phone: 970 491 5263
Department of Physics.............Fax: 970 491 7947
Colorado State University.......Email: ge...@la...
Fort Collins CO 80523-1875

Re: [Visualpython-users] no-go with new (and old!) visual on linux

[Jonathan B. <jbr...@ea...>]

On Fri, 2004-07-23 at 15:49, Martin Gelfand wrote:
> I am running Debian testing, but got boost-python-dev and dependencies
> from unstable (since only unstable has boost 1.31 at this time).
> 
> Compilation of visual-3.0 seemed to go fine.  I used
> PYTHON=/usr/bin/python ./configure --prefix=/usr/local
> (/usr/bin/python is a link to python2.3, and I want all locally built
> software in /usr/local)

/usr/local is configure's default prefix, and so long as /usr/bin/python
is the first Python named 'python' in your PATH, that is the default as
well.

> Importing visual in a python2.3 shell goes OK, but I get a segfault on
> trying to do anything that generates a display on-screen.  I see a display 
> window appear, then immediately close.  Simply instantiating a display
> (e.g. firstdisplay=visual.display()) does not throw a segfault.

visual.display has "initialize on first use" semantics, and merely
instantiating one isn't enough to constitute 'use'; you have to also
instantiate a rendered object.

I'm using visual 3.0 on Debian Sid right now, without any problems.  Can
you please run the following to get a backtrace?:

gdb python
[displays some stuff and gives you the (gdb) prompt]
(gdb) run
[displays more stuff and gives you the python prompt]
>>> from visual import *
>>> import sys
>>> sys.path.append( 
...   "/usr/local/lib/python2.3/site-packages/visual/demos")
... 
>>> import stars
[ This will attempt to run the stock "stars" demo program ]
[ When (if) it crashes, control will return to gdb with the (gdb)
prompt.]
(gdb) bt

Please send me the output of the last command.

Also, to partially rule out general OpenGL problems, please see if
glxgears runs.

Are you by chance using the R200 DRI driver?

> What's worse, my old visual setup (python2.2, visual-2.1.9) no longer
> works:  
> Python 2.2.3+ (#1, Jun 20 2004, 13:32:48)
> [GCC 3.3.4 (Debian)] on linux2
> Type "help", "copyright", "credits" or "license" for more information.
> >>> import visual
> Visual-2003-10-05
> Traceback (most recent call last):
>   File "<stdin>", line 1, in ?
>   File "/usr/local/lib/python2.2/site-packages/visual/__init__.py", line 17, 
> in ?
>     import cvisual
> ImportError: /usr/local/lib/python2.2/site-packages/cvisualmodule.so: 
> undefined symbol: __pure_virtual
> 
> I don't see how this could be the fault of installing the new visual--could
> it have come about due to an upgrade of python2.2 since I built that
> version of visual?

I don't know for certain what this could be.  The only place that I can
find this symbol on my system is in libgcc.a from GCC 2.95.  Try
rebuilding VPython 2.1.9 with the current compiler.  (you will have to
specify PYTHON=python2.2 to build against Python 2.2 since `python` is
2.3).

-Jonathan

Re: [Visualpython-users] no-go with new (and old!) visual on linux

[Martin G. <mar...@co...>]

In answer to the questions from J.B. ...
* I am using an nvidia card with the nvidia drivers
* The output of everything after import stars...

>>> import stars

Right button drag to rotate view.
Left button drag up or down to move in or out.

[New Thread 32769 (LWP 10189)]
[New Thread 16386 (LWP 10190)]

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 16386 (LWP 10190)]
0x40118c0f in strlen () from /lib/libc.so.6
(gdb) bt
#0  0x40118c0f in strlen () from /lib/libc.so.6
#1  0x40a90464 in visual::GLDevice::render_control ()
   from /usr/local/lib/python2.3/site-packages/cvisualmodule.so
#2  0x40a9187b in visual::GLDevice::callback ()
   from /usr/local/lib/python2.3/site-packages/cvisualmodule.so
#3  0x40e98ec4 in g_main_set_poll_func () from /usr/lib/libglib-1.2.so.0
#4  0x40e9803d in g_get_current_time () from /usr/lib/libglib-1.2.so.0
#5  0x40e984f4 in g_get_current_time () from /usr/lib/libglib-1.2.so.0
#6  0x40e98724 in g_main_run () from /usr/lib/libglib-1.2.so.0
#7  0x40c4c25f in gtk_main () from /usr/lib/libgtk-1.2.so.0
#8  0x40aca12b in visual::mainloop ()
   from /usr/local/lib/python2.3/site-packages/cvisualmodule.so
#9  0x4002ee51 in pthread_start_thread () from /lib/libpthread.so.0
#10 0x4002eecf in pthread_start_thread_event () from /lib/libpthread.so.0
#11 0x4017a69a in clone () from /lib/libc.so.6

On Friday 23 July 2004 03:23 pm, Jonathan Brandmeyer wrote:
> On Fri, 2004-07-23 at 15:49, Martin Gelfand wrote:
> > I am running Debian testing, but got boost-python-dev and dependencies
> > from unstable (since only unstable has boost 1.31 at this time).
> >
> > Compilation of visual-3.0 seemed to go fine.  I used
> > PYTHON=/usr/bin/python ./configure --prefix=/usr/local
> > (/usr/bin/python is a link to python2.3, and I want all locally built
> > software in /usr/local)
>
> /usr/local is configure's default prefix, and so long as /usr/bin/python
> is the first Python named 'python' in your PATH, that is the default as
> well.
>
> > Importing visual in a python2.3 shell goes OK, but I get a segfault on
> > trying to do anything that generates a display on-screen.  I see a
> > display window appear, then immediately close.  Simply instantiating a
> > display (e.g. firstdisplay=visual.display()) does not throw a segfault.
>
> visual.display has "initialize on first use" semantics, and merely
> instantiating one isn't enough to constitute 'use'; you have to also
> instantiate a rendered object.
>
> I'm using visual 3.0 on Debian Sid right now, without any problems.  Can
> you please run the following to get a backtrace?:
>
> gdb python
> [displays some stuff and gives you the (gdb) prompt]
> (gdb) run
> [displays more stuff and gives you the python prompt]
>
> >>> from visual import *
> >>> import sys
> >>> sys.path.append(
>
> ...   "/usr/local/lib/python2.3/site-packages/visual/demos")
> ...
>
> >>> import stars
>
> [ This will attempt to run the stock "stars" demo program ]
> [ When (if) it crashes, control will return to gdb with the (gdb)
> prompt.]
> (gdb) bt
>
> Please send me the output of the last command.
>
> Also, to partially rule out general OpenGL problems, please see if
> glxgears runs.
>
> Are you by chance using the R200 DRI driver?
>
> > What's worse, my old visual setup (python2.2, visual-2.1.9) no longer
> > works:
> > Python 2.2.3+ (#1, Jun 20 2004, 13:32:48)
> > [GCC 3.3.4 (Debian)] on linux2
> > Type "help", "copyright", "credits" or "license" for more information.
> >
> > >>> import visual
> >
> > Visual-2003-10-05
> > Traceback (most recent call last):
> >   File "<stdin>", line 1, in ?
> >   File "/usr/local/lib/python2.2/site-packages/visual/__init__.py", line
> > 17, in ?
> >     import cvisual
> > ImportError: /usr/local/lib/python2.2/site-packages/cvisualmodule.so:
> > undefined symbol: __pure_virtual
> >
> > I don't see how this could be the fault of installing the new
> > visual--could it have come about due to an upgrade of python2.2 since I
> > built that version of visual?
>
> I don't know for certain what this could be.  The only place that I can
> find this symbol on my system is in libgcc.a from GCC 2.95.  Try
> rebuilding VPython 2.1.9 with the current compiler.  (you will have to
> specify PYTHON=python2.2 to build against Python 2.2 since `python` is
> 2.3).
>
> -Jonathan
>
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by BEA Weblogic Workshop
> FREE Java Enterprise J2EE developer tools!
> Get your free copy of BEA WebLogic Workshop 8.1 today.
> http://ads.osdn.com/?ad_id=4721&alloc_id=10040&op=click
> _______________________________________________
> Visualpython-users mailing list
> Vis...@li...
> https://lists.sourceforge.net/lists/listinfo/visualpython-users

-- 
Martin Gelfand

Associate Professor.............Phone: 970 491 5263
Department of Physics.............Fax: 970 491 7947
Colorado State University.......Email: ge...@la...
Fort Collins CO 80523-1875

Re: [Visualpython-users] no-go with new (and old!) visual on linux

[Jonathan B. <jbr...@ea...>]

On Fri, 2004-07-23 at 18:42, Martin Gelfand wrote:
> In answer to the questions from J.B. ...
> * I am using an nvidia card with the nvidia drivers
> * The output of everything after import stars...

[snipped]

Thanks for the info.

First, apply the attached patch to cvisual/gldevice.cpp [1], rebuild
vpython, and see if that fixes the problem.
If it does, please upgrade to the latest nvidia-glx and nvidia-kernel
packages in unstable, and rebuild vpython from the unpatched source, to
see if upgrading the nvidia drivers fixes the problem.

If applying the patch corrects the crash and upgrading the driver does
not, please run glxinfo and send me the output.

Thanks,
-Jonathan


[1] Save the attached file as cvisual/gldevice.cpp.diff (in the visual
source tree)
cd into cvisual
run `patch < gldevice.cpp.diff`

Re: [Visualpython-users] no-go with new (and old!) visual on linux

[Jonathan B. <jbr...@ea...>]

On Fri, 2004-07-23 at 22:49, Jonathan Brandmeyer wrote:
> On Fri, 2004-07-23 at 18:42, Martin Gelfand wrote:
> > In answer to the questions from J.B. ...
> > * I am using an nvidia card with the nvidia drivers
> > * The output of everything after import stars...
> 
> [snipped]
> 
> Thanks for the info.
> 
> First, apply the attached patch to cvisual/gldevice.cpp [1], rebuild
> vpython, and see if that fixes the problem.
> If it does, please upgrade to the latest nvidia-glx and nvidia-kernel

Make that nvidia-glx-dev rather than nvidia-glx.

> packages in unstable, and rebuild vpython from the unpatched source, to
> see if upgrading the nvidia drivers fixes the problem.
> 

-Jonathan

Re: [Visualpython-users] no-go with new (and old!) visual on linux

[Martin G. <mar...@co...>]

I've upgraded the nvidia drivers/libs to the version in unstable,
and recompiled visual with the patched gldevice.cpp (I didn't
try going back to the original version), and am happy to report
that all seems well so far.  Many thanks!  Perhaps a note
to the effect that the linux nvidia drivers version 1.0.5336 do
not work with visual, upgrade to 1.0.6106, should be added
to the linux download page.

Love the new helix!   --Marty

On Friday 23 July 2004 08:49 pm, Jonathan Brandmeyer wrote:
> On Fri, 2004-07-23 at 18:42, Martin Gelfand wrote:
> > In answer to the questions from J.B. ...
> > * I am using an nvidia card with the nvidia drivers
> > * The output of everything after import stars...
>
> [snipped]
>
> Thanks for the info.
>
> First, apply the attached patch to cvisual/gldevice.cpp [1], rebuild
> vpython, and see if that fixes the problem.
> If it does, please upgrade to the latest nvidia-glx and nvidia-kernel
> packages in unstable, and rebuild vpython from the unpatched source, to
> see if upgrading the nvidia drivers fixes the problem.
>
> If applying the patch corrects the crash and upgrading the driver does
> not, please run glxinfo and send me the output.
>
> Thanks,
> -Jonathan
>

Re: [Visualpython-users] no-go with new (and old!) visual on linux

[Martin G. <mar...@co...>]

I've resolved the problem with the old visual.  On rebuilding, there
was an error that libGL.so did not exist.  It did, but it was a link
to libGL.so.1.2, which did not exist.  However, libGL.so.1 did exist,
as a link to libGL.so.1.0.5336.  Pointing libGL.so to libGL.so.1 allowed
visual to build, and it runs fine again.

Does that constitute a bug in xlibmesa-gl-dev (the package which
appears to own /usr/lib/libGL.so)?

Onward to visual-3.0...  And thanks, Jonathan, for your assistance.


> On Friday 23 July 2004 03:23 pm, Jonathan Brandmeyer wrote:
> > On Fri, 2004-07-23 at 15:49, Martin Gelfand wrote:
> > > What's worse, my old visual setup (python2.2, visual-2.1.9) no longer
> > > works:
> > > Python 2.2.3+ (#1, Jun 20 2004, 13:32:48)
> > > [GCC 3.3.4 (Debian)] on linux2
> > > Type "help", "copyright", "credits" or "license" for more information.
> > >
> > > >>> import visual
> > >
> > > Visual-2003-10-05
> > > Traceback (most recent call last):
> > >   File "<stdin>", line 1, in ?
> > >   File "/usr/local/lib/python2.2/site-packages/visual/__init__.py",
> > > line 17, in ?
> > >     import cvisual
> > > ImportError: /usr/local/lib/python2.2/site-packages/cvisualmodule.so:
> > > undefined symbol: __pure_virtual
> > >
> > > I don't see how this could be the fault of installing the new
> > > visual--could it have come about due to an upgrade of python2.2 since I
> > > built that version of visual?
> >
> > I don't know for certain what this could be.  The only place that I can
> > find this symbol on my system is in libgcc.a from GCC 2.95.  Try
> > rebuilding VPython 2.1.9 with the current compiler.  (you will have to
> > specify PYTHON=python2.2 to build against Python 2.2 since `python` is
> > 2.3).
> >

Re: [Visualpython-users] no-go with new (and old!) visual on linux

[Jonathan B. <jbr...@ea...>]

On Sun, 2004-07-25 at 15:12, Martin Gelfand wrote:
> I've resolved the problem with the old visual.  On rebuilding, there
> was an error that libGL.so did not exist.  It did, but it was a link
> to libGL.so.1.2, which did not exist.  However, libGL.so.1 did exist,
> as a link to libGL.so.1.0.5336.  Pointing libGL.so to libGL.so.1 allowed
> visual to build, and it runs fine again.
> 
> Does that constitute a bug in xlibmesa-gl-dev (the package which
> appears to own /usr/lib/libGL.so)?

I think that the symlink from libGL.so to the correct file is provided
by nvidia-glx-dev.

> Onward to visual-3.0...  And thanks, Jonathan, for your assistance.

You're welcome,
-Jonathan