Hi Stephan,
thanks for looking at this.
As an aside, I'm Uli, not Andreas, and - I believe we spent some time
together at Art+Com :-).
This is the first call:
#0 0x00688bba in osgViewer::PixelBufferCarbon::closeImplementation ()
#1 0x092a012f in osg::GraphicsContext::close ()
#2 0x0066f910 in osgViewer::Viewer::~Viewer ()
#3 0x08c65eb8 in avg::OSGNode::disconnect (this=0x1ef447d0,
bKill=4098347009) at /usr/local/include/osg/Referenced:183
and this the second:
#0 0x00688bba in osgViewer::PixelBufferCarbon::closeImplementation ()
#1 0x092a012f in osg::GraphicsContext::close () at
function_template.hpp:1010
#2 0x00688e23 in osgViewer::PixelBufferCarbon::~PixelBufferCarbon ()
#3 0x0920e825 in osg::Referenced::unref () at function_template.hpp:
1010
#4 0x092049c0 in osg::ref_ptr<osg::GraphicsContext>::~ref_ptr () at
function_template.hpp:1010
#5 0x09201aff in osg::Camera::~Camera () at function_template.hpp:1010
#6 0x08c65ed8 in avg::OSGNode::disconnect (this=0x1ef447d0,
bKill=4098347009) at /usr/local/include/osg/Referenced:183
The setup looks like this: The context is created on a pbuffer, but
this pbuffer is never rendered to and doesn't have a camera. All
rendering happens in an FBO. The camera in the second stack trace
isn't attached to the pbuffer at all - it's rendering to the FBO.
This is OSG 2.8.3.
Applying your patch seems to fix (or at least hide) it, but as I said:
The error is hard to reproduce and I'm not sure whether it's really
gone.
Cheers,
Uli
On Apr 30, 2010, at 5:02 PM, Stephan Maximilian Huber wrote:
Hi Andreas,
the underlying problem is the double-free. Can you provide two
stack-traces for the two calls of
PixelBufferCarbon::closeImplementation?
as the x11-implementation of closeImplementation does nothing fancy,
there's nothing which can crash ;-)
A quick and dirty fix for PixelBufferCarbon would be:
if (_pbuffer) { aglDestroyPBuffer(_pbuffer); _pbuffer = NULL; }
if (_context) { aglDestroyContext(_context); _context = NULL; }
if (_pixelformat) { aglDestroyPixelFormat(_pixelformat);
_pixelformat =
NULL; }
can you provide some more details about how your viewer-setup?
cheers,
Stephan
Am 30.04.10 15:14, schrieb Ulrich von Zadow:
Hi,
my OSG program crashes on end:
Thread 0 Crashed:
0 com.apple.agl 0x91e52a71
aglDestroyPixelFormat + 60
1 libosgViewer.65.dylib 0x00688c16
osgViewer::PixelBufferCarbon::closeImplementation() + 150
2 libosg.65.dylib 0x092a012f
osg::GraphicsContext::close(bool) + 1019
3 libosgViewer.65.dylib 0x00688e35
osgViewer::PixelBufferCarbon::~PixelBufferCarbon() + 47
4 osgnode.so 0x08c65f7b
avg::OSGNode::disconnect(bool) + 475 (Referenced:183)
The crash isn't really reproducible, but it happens only on a mac,
never
under linux.
What's irritating me (besides the crash itself ;-)) is that
PixelBufferCarbon::closeImplementation() is called twice. Here is the
osg logging output with output added to the closeImplementation call:
Viewer::~Viewer():: start destructor getThreads = 0
close(1)0x1ef4f600
Releasing GL objects for Camera=0xa0d000 _state=0x3ab730
Closing still viable window 0 _state->getContextID()=0
Doing Flush
Done Flush
PixelBufferCarbon::closeImplementation: 0x1ef4f600
_pixelformat: 0x1ef4f728
Doing discard of deleted OpenGL objects.
decrementUsageCount()0
GraphicsContext::decrementContextIDUsageCount(0) to 0
Viewer::~Viewer() end destrcutor getThreads = 0
Destructing osgViewer::View
Destructing osg::View
close(1)0x1ef4f600
PixelBufferCarbon::closeImplementation: 0x1ef4f600
_pixelformat: 0x1ef4f728
close(0)0x1ef4f600
GraphicsContext::unregisterGraphicsContext 0x1ef4f600
Done destructing osg::View
The crash happens (when it does) at the second
PixelBufferCarbon::closeImplementation line.
Any idea what I might be doing wrong? The logging output is a
response
to dereferencing the Viewer - nothing more.
Cheers,
Uli
--
Ulrich von Zadow
Software Engineer (Dipl. Inf.)
Exhibit Development
Tel +49 (0)30 / 2000 577 12
Fax +49 (0)30 / 2000 577 20
Skype: uzadow
Archimedes Solutions GmbH
Saarbrücker Str. 24 10405 Berlin
www.archimedes-solutions.com
Geschäftsführung:
W. Rien | J. Schmidtsiefen | S. Spenling
Amtsgericht: Berlin Charlottenburg
HR Nr.: 107563 B
UST-ID Nr.: DE-253.771.793
_______________________________________________
osg-users mailing list
[email protected]
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
_______________________________________________
osg-users mailing list
[email protected]
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
--
Ulrich von Zadow
Software Engineer (Dipl. Inf.)
Exhibit Development
Tel +49 (0)30 / 2000 577 12
Fax +49 (0)30 / 2000 577 20
Skype: uzadow
Archimedes Solutions GmbH
Saarbrücker Str. 24 10405 Berlin
www.archimedes-solutions.com
Geschäftsführung:
W. Rien | J. Schmidtsiefen | S. Spenling
Amtsgericht: Berlin Charlottenburg
HR Nr.: 107563 B
UST-ID Nr.: DE-253.771.793
_______________________________________________
osg-users mailing list
[email protected]
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
