Hi Jannik, I had to change the code to always assign a VertexBufferObject as I couldn't find a way of implementing VAO without forcing changes to client code, in order to make sure the OSG works on GLcore profile implementations that require VAO all arrays have to have a VertexBufferObject.
The observation that display lists are being used the VBO side shows a bug, display lists and VBO shouldn't be invoked though, we'll need to review osg::Geometry to make sure VBO's aren't used when display lists are being generated.. Robert. On 16 February 2017 at 21:14, Jannik Heller <[email protected]> wrote: > Hi, > > I've noticed some odd API usage in the OSG that seems to have started with > the VAO support changes in OSG 3.5. Vertex buffer objects are being used > inside a display list. This happens on a standard Geometry with default > settings. Example: > > glNewList(6, GL_COMPILE) > glGenBuffers(1, [5]) > glBindBuffer(GL_ARRAY_BUFFER, 5) > glBufferData(...) > glBufferSubData(...) x4 > glVertexPointer(3, GL_FLOAT, 0, NULL) > glNormalPointer(GL_FLOAT, 0, 0x90) > glColorPointer(4, GL_FLOAT, 0x180) > glTexCoordPointer(2, GL_FLOAT, 0, 0x120) > glDrawElements(...) > glBindBuffer(GL_ARRAY_BUFFER, 0) > glEndList() > > The source of this change in behavior is here: > https://github.com/openscenegraph/OpenSceneGraph/blob/master/src/osg/Geometry.cpp#L187 > OSG will now always assign VBOs, even if they weren't requested. > > First off all I find this usage very odd. The GL standard doesn't actually > specify what is supposed to happen to VBOs in a display list - is the VBO > simply ignored or could that cause a new VBO to be created with every call to > the display list? Even if this works, the calls are completely redundant and > could be removed. > > More importantly, though, and the reason I'm bringing it up, is that this > usage seems to be causing a crash with some versions of Mesa drivers, and the > crash goes away when I disable display lists. > > Ubuntu 14.04 - Nvidia: works > Ubuntu 14.04 - Intel HD: broken > Ubuntu 16.04 - Intel HD: broken > > I've attached a minimal-ish apitrace you can use to see if your driver is > affected - run it with 'apitrace replay <file>' , if you get 'caught signal > 11' that means the driver has crashed. > > Cheers, > Jannik > > ------------------ > Read this topic online here: > http://forum.openscenegraph.org/viewtopic.php?p=70219#70219 > > > > > _______________________________________________ > 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
