Hi Serge,
I can't right out whether your new code is safe, I would for starts
check the graphics context is being passed in by case the object to a
GraphicsContext* as well as the Viewer* just to make sure that it's be
called from the correct thread.
Rather than roll your own tests, could you just try enabling the
compile context via the setting the env var OSG_COMPILE_CONTEXTS to
ON, then running osgviewer on one your paged databases.
Robert.
On Mon, Aug 4, 2008 at 5:46 PM, Serge Lages <[EMAIL PROTECTED]> wrote:
> You're right, my test wasn't correct.
> I've made a new one (see below), and it doesn't work on any OSG version...
> :/ It crash everywhere on glGetString because the makeCurrent failed on the
> PixelBuffer.
> I'll try to work on it tomorrow, any idea on where to look ? :) Are
> PixelBuffer not working at all under Windows or just having troubles with
> the CompileContext ?
>
> Here is my new test :
>
> class CustomOperation : public osg::Operation
> {
> public:
>
> CustomOperation() : osg::Operation("CustomOperation", true) {}
>
> void operator()(osg::Object *object)
> {
> osgViewer::Viewer *viewer = dynamic_cast<osgViewer::Viewer
> *>(object);
>
> // Do my task
> if (viewer == NULL && _node == NULL)
> {
> osg::Image *image = osgDB::readImageFile("test.jpg");
> osg::StateSet *stateset = new osg::StateSet();
> osg::Texture2D *texture = new osg::Texture2D();
>
> texture->setImage(image);
> texture->setUnRefImageDataAfterApply(true);
>
> texture->setInternalFormatMode(osg::Texture::USE_S3TC_DXT5_COMPRESSION);
> stateset->setTextureAttributeAndModes(0, texture,
> osg::StateAttribute::ON);
>
>
> stateset->compileGLObjects(*osg::GraphicsContext::getCompileContext(0)->getState());
> if (texture->getImage() == NULL)
> {
> osg::notify(osg::NOTICE) << "Image successfully applied !"
> << std::endl;
> }
> _node = createTexturedQuad(osg::Vec2(), 100, 100, stateset);
> }
> // Update the scene
> else if (_node != NULL &&
> viewer->getSceneData()->asGroup()->getNumChildren() == 0)
> {
> osg::notify(osg::NOTICE) << "Node merged to the graph !" <<
> std::endl;
> viewer->getSceneData()->asGroup()->addChild(_node.get());
> }
> }
>
> private:
>
> osg::ref_ptr<osg::Node> _node;
> };
>
> int main(int ac, char **av)
> {
> osgViewer::Viewer viewer;
>
> viewer.setThreadingModel(osgViewer::ViewerBase::SingleThreaded);
> osg::DisplaySettings::instance()->setCompileContextsHint(true);
> osg::DisplaySettings::instance()->setMaxNumberOfGraphicsContexts(1);
> viewer.setUpViewInWindow(100, 100, 800, 600);
> viewer.setCameraManipulator(new osgGA::TrackballManipulator());
> viewer.setSceneData(new osg::Group());
> viewer.realize();
>
> if (osg::GraphicsContext::getCompileContext(0) != NULL &&
> osg::GraphicsContext::getCompileContext(0)->getGraphicsThread() !=
> NULL)
> {
> CustomOperation *operation = new CustomOperation();
>
>
> osg::GraphicsContext::getCompileContext(0)->getGraphicsThread()->add(operation);
> viewer.addUpdateOperation(operation);
> osg::notify(osg::NOTICE) << "Operation successfully added !" <<
> std::endl;
> }
>
> viewer.run();
> }
>
> On Mon, Aug 4, 2008 at 6:22 PM, Robert Osfield <[EMAIL PROTECTED]>
> wrote:
>>
>> Hi Serge,
>>
>> The code looks to be a bit suspect to me. The problem is that the
>> CompileContextHint will tell osgViewer to try and create a compile
>> context for each window it creates, and then start a thread for each
>> of these compile context, this thread will have the a makeCurrent()
>> internally and then assume that this thread has the context.
>> However, with your code you go ahead a do a
>> makeCurrent()/releaseContext() in the main thread so that the compile
>> contexts graphics thread will no longer have the context current and
>> any further OpenGL calls in that thread will fail. The fact that
>> this ever worked is surprising...
>>
>> If you do want to do a precompile in the way you are doing it'd be
>> best to manually create the compile context, use it, then start the
>> thread on it, or just then discard the context.
>>
>> Robert.
>>
>> On Mon, Aug 4, 2008 at 4:31 PM, Serge Lages <[EMAIL PROTECTED]> wrote:
>> > Hi all,
>> >
>> > I've made some investigations. First of all, I've tested several OSG
>> > versions with the following code :
>> >
>> > osgViewer::Viewer viewer;
>> >
>> > viewer.setThreadingModel(osgViewer::ViewerBase::SingleThreaded);
>> > osg::DisplaySettings::instance()->setCompileContextsHint(true);
>> > osg::DisplaySettings::instance()->setMaxNumberOfGraphicsContexts(1);
>> > viewer.setUpViewInWindow(100, 100, 800, 600);
>> > viewer.setCameraManipulator(new osgGA::TrackballManipulator());
>> > viewer.realize();
>> > if (osg::GraphicsContext::getCompileContext(0) != NULL)
>> > {
>> > osg::Image *image = osgDB::readImageFile("test.jpg");
>> > osg::StateSet *stateset = new osg::StateSet();
>> > osg::Texture2D *texture = new osg::Texture2D();
>> >
>> > texture->setImage(image);
>> > texture->setUnRefImageDataAfterApply(true);
>> >
>> > texture->setInternalFormatMode(osg::Texture::USE_S3TC_DXT5_COMPRESSION);
>> > stateset->setTextureAttributeAndModes(0, texture,
>> > osg::StateAttribute::ON);
>> >
>> > osg::GraphicsContext::getCompileContext(0)->makeCurrent();
>> >
>> >
>> > stateset->compileGLObjects(*osg::GraphicsContext::getCompileContext(0)->getState());
>> > osg::GraphicsContext::getCompileContext(0)->releaseContext();
>> > if (texture->getImage() == NULL)
>> > {
>> > osg::notify(osg::NOTICE) << "Texture successfully applied !"
>> > <<
>> > std::endl;
>> > }
>> > viewer.setSceneData(createTexturedQuad(0, 0, 100, 100,
>> > stateset));
>> > }
>> > viewer.run();
>> >
>> > (the createTexturedQuad function just returns a geode with a quad and
>> > the
>> > texture applied on it).
>> >
>> > Here are my results :
>> >
>> > - OSG 2.2 and OSG 2.4 : it seems to work, I have the "Texture
>> > successfully
>> > applied !" message and my quad have the texture on it. Note : if I don't
>> > specify the setMaxNumberOfGraphicsContexts, my image is not unref but it
>> > still works.
>> >
>> > - Latest SVN : If setMaxNumberOfGraphicsContexts is not set to 1, it
>> > crash.
>> > If it's set, I got the "Texture successfully applied !" message but my
>> > quad
>> > has no texture applied on it.
>> >
>> > Robert, have you any hint on where to look next ? You told it works on
>> > Linux, so do you think it will be usefull to start looking at the
>> > modifications on the Win32 implementation of osgViewer since OSG 2.4 ?
>> >
>> > On Mon, Aug 4, 2008 at 2:14 PM, Serge Lages <[EMAIL PROTECTED]>
>> > wrote:
>> >>
>> >> Some time ago it worked for me (the same time when we were able to make
>> >> glGetString calls outside a draw callbacks).
>> >>
>> >> I'll try to make some tests this afternoon and let you know how it
>> >> goes.
>> >>
>> >> On Mon, Aug 4, 2008 at 1:56 PM, Glenn Waldron <[EMAIL PROTECTED]>
>> >> wrote:
>> >>>
>> >>> FWIW I have never been able to get compile contexts to work on Windows
>> >>> since the feature was introduced. So I believe the issue goes back a
>> >>> ways
>> >>> (i.e., is not particular to 2.6).
>> >>>
>> >>> HTH- Glenn
>> >>>
>> >>> On Mon, Aug 4, 2008 at 6:17 AM, Adrian Egli OpenSceneGraph (3D)
>> >>> <[EMAIL PROTECTED]> wrote:
>> >>>>
>> >>>> Hi Robert,
>> >>>>
>> >>>> i just come tested under windows vista, same as problem as Serge
>> >>>> reported :-(
>> >>>>
>> >>>> PixelBufferWin32::makeCurrentImplementation, wglMakeCurrent error:
>> >>>> The
>> >>>> requested resource is in use.
>> >>>>
>> >>>>
>> >>>> 2008/8/4 Robert Osfield <[EMAIL PROTECTED]>
>> >>>>>
>> >>>>> Hi Serge,
>> >>>>>
>> >>>>> I've been doing testing the compile context support using osgviewer
>> >>>>> on
>> >>>>> paged database I have and it all works without error so far. I'm
>> >>>>> using Kubuntu 7.10 and 7800GT. I enable compile contexts via:
>> >>>>>
>> >>>>> export OSG_COMPILE_CONTEXTS=ON
>> >>>>> osgviewer pagedmodel.ive --SingleThreaded
>> >>>>>
>> >>>>> I tried the whole range of threading models without incident.
>> >>>>>
>> >>>>> So this problem looks to be a Windows issue. My guess is that the
>> >>>>> error report of about a problem with make current is what is the
>> >>>>> early
>> >>>>> warning of something not working quite right. I'm afraid I can't do
>> >>>>> much to investigate this myself though as I have no Windows box to
>> >>>>> test against.
>> >>>>>
>> >>>>> Robert.
>> >>>>>
>> >>>>>
>> >>>>> On Thu, Jul 31, 2008 at 9:44 AM, Serge Lages <[EMAIL PROTECTED]>
>> >>>>> wrote:
>> >>>>> > Hi all,
>> >>>>> >
>> >>>>> > I am currently trying to use CompileContexts on my application,
>> >>>>> > but
>> >>>>> > just
>> >>>>> > doing this :
>> >>>>> >
>> >>>>> > osgViewer::Viewer viewer;
>> >>>>> >
>> >>>>> > viewer.setThreadingModel(osgViewer::ViewerBase::SingleThreaded);
>> >>>>> > osg::DisplaySettings::instance()->setCompileContextsHint(true);
>> >>>>> > return (viewer.run());
>> >>>>> >
>> >>>>> > The application crash into glExtensions.cpp line 46 (trying to get
>> >>>>> > the GL
>> >>>>> > version with glGetString( GL_VERSION )).
>> >>>>> > Before the crash I have this error message :
>> >>>>> >
>> >>>>> > PixelBufferWin32::makeCurrentImplementation, wglMakeCurrent error:
>> >>>>> > La
>> >>>>> > ressource demandé est en cours d'utilisation.
>> >>>>> > or in english something like that :
>> >>>>> > PixelBufferWin32::makeCurrentImplementation, wglMakeCurrent error:
>> >>>>> > the
>> >>>>> > wanted resource is already being used.
>> >>>>> >
>> >>>>> > Any idea on what's happening ?
>> >>>>> > My system : WinXP with SVN head, VS 8 Express, NVIDIA 8600GTS
>> >>>>> >
>> >>>>> > Thanks in advance !
>> >>>>> >
>> >>>>> > --
>> >>>>> > Serge Lages
>> >>>>> > http://www.tharsis-software.com
>> >>>>> >
>> >>>>> > _______________________________________________
>> >>>>> > 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
>> >>>>
>> >>>>
>> >>>>
>> >>>> --
>> >>>> ********************************************
>> >>>> Adrian Egli
>> >>>>
>> >>>> _______________________________________________
>> >>>> osg-users mailing list
>> >>>> [email protected]
>> >>>>
>> >>>>
>> >>>> http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
>> >>>>
>> >>>
>> >>>
>> >>>
>> >>> --
>> >>> Glenn Waldron : Pelican Mapping : http://pelicanmapping.com :
>> >>> +1.703.652.4791
>> >>>
>> >>> _______________________________________________
>> >>> osg-users mailing list
>> >>> [email protected]
>> >>>
>> >>> http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
>> >>>
>> >>
>> >>
>> >>
>> >> --
>> >> Serge Lages
>> >> http://www.tharsis-software.com
>> >
>> >
>> >
>> > --
>> > Serge Lages
>> > http://www.tharsis-software.com
>> >
>> > _______________________________________________
>> > 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
>
>
>
> --
> Serge Lages
> http://www.tharsis-software.com
>
> _______________________________________________
> 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
