Hi Wojciech, First of thanks for the efforts in tracking down this issue.
I think that accepting a 32bit buffer when 24bit was sought would be fine, but if the stick_adherance is set then it shouldn't be acceptable. Could you post changes to osg-submissions mailing list too, as this will make it much easier for me and Don to track submissions - the osg-users mailing list can often be swamped so fixes could be missed otherwise. A general note for other win32 users, could you guys test out these changes? Robert. On 7/12/06, Wojciech Lewandowski <[EMAIL PROTECTED]> wrote:
Hi again, I took me longer than I thought but I finally made the tests. I can confirm your observations. It seems that there is an issue with VisualChooser::choose method implementation in Win32. Good thing for me this is not my code but this (my changes were in RenderSurface ;-). I did look further whats actually happening. It seems that list of _visual_attributes contains 2 recommended depth buffer sizes: 16 bit and 24 bit. ChoosePixelFormat (Line 654 VisualChooser.cpp ver 1.26) returns PIixelFormat using 32 bit depth. Code in lines 655-682 does aditional check on attributes returned from DescribePixelFormat and rejects visual because bit depth does not match exactly. And here lies the problem: this code resets visualId but forgets to set failed flag which is used later. So the remainning code assumes that visual was found but visual id is reset and incorrect empty visualInfo structure is returned. If Don or Robert reads this I have attached simple fix to VisualChooser.cpp. This fix makes sure that VisualChooser::choose will return null if PixelFormat gets rejected in lines 655-682. But I am not sure if this in accordance wiht original intent. Should 32 bit depth buffer be rejected if 16 or 24 bit was sought ? If not further changes in the code should be made... And what is the role of strict_adherence flag ? Best, Wojtek Lewandowski ----- Original Message ----- From: Frank Bergmann To: 'osg users' Sent: Sunday, July 09, 2006 5:22 AM Subject: RE: [osg-users] Working towards OpenSceneGraph-1.1 release Hello Wojtek, Sorry for taking my sweet time to answer this one … I believe the problem to be with the choose method in VisualChooser.cpp … in line 683 the visual id will be set to zero in case one of the tests fail. On an older (or virtual, or remote machine) there will not be a better visual than the one already found. So what I did in my build was to display a warning in console mode and in window mode my application is silent about the test-failure and work just fine. I know that this is a hack and I would like to see this issue solved in a better way. Best Frank From: Wojciech Lewandowski [mailto:[EMAIL PROTECTED] Sent: Wednesday, June 28, 2006 6:54 AM To: osg users Subject: Re: [osg-users] Working towards OpenSceneGraph-1.1 release Hi Frank, What you describe seems a bit difficult for me for me. I don't have easy access to second machine to test these scenarios. In fact I have a lot of other work to do till the end of this week. Hopefully situation will change and I will look at the problem next week. If you can't wait you may try to revert my changes and see if that helps. Don submitted them around 12th April 2006. Look at the thread with following topic: "Attaching OGL initialized Window to RenderSurface +RenderSurface build broken" Wojtek ----- Original Message ----- From: Frank Bergmann To: 'osg users' Sent: Tuesday, June 27, 2006 7:28 PM Subject: RE: [osg-users] Working towards OpenSceneGraph-1.1 release Hello Wojtek, One easy way to reproduce the failure would be to use a remote desktop connection to your Win32 development machine. When using Remote Desktop it will seem to windows as if a different graphics card is used. When tracing the call I see a test for the DepthSize (VisualChooser.cpp Line 678) failing … maybe it would be enough for me to reset the attribute from 32 to 24 bit … And yes it occurs with the osgDemos … osgTeapot will result in an endless loop of OpenGL errors, osgPick is an example of an application crashing (access violation) with: "Error: In Texture::Extensions::setupGLExtensions(..) OpenGL version test failed, requires valid graphics context." The same error occurs when testing the application in a virtual WIN32 machine (VMware) … while it works fine with virtual machines with linux systems. Thanks Frank From: Wojciech Lewandowski [mailto:[EMAIL PROTECTED] Sent: Tuesday, June 27, 2006 6:59 AM To: osg users Subject: Re: [osg-users] Working towards OpenSceneGraph-1.1 release Frank & Robert, I may be somehow responsible for the above issues. I once asked Don to provide support for render surface to use window with gl context already created. Don implemented this feature in Linux and made empty placeholder for Windows version. I wrote neccessary portion of code and Don put this into repository. I was quite a long ago and I don't remember details but may try to help you if you provide some aditional details. Do you see this problem when using osg Demos ? If not how you create render surface ? Do you work with non standard windows toolkit ? Can you trace the code and see which WGL extensions are missing ? Wojtek Lewandowski P.S. I don't feel responsible for use of MessageBoxes though ;-) They were there already... ----- Original Message ----- From: Frank Bergmann To: 'osg users' Sent: Monday, June 26, 2006 8:07 PM Subject: RE: [osg-users] Working towards OpenSceneGraph-1.1 release The current CVS version works fine for me with one noticeable change since the 1.0 release: - The Producer will no longer work with older graphics cards not supporting certain WGL extensions on WIN32. Osg::notify prints: "Producer::VisualChooser: unable to setup a valid visual with WGL extensions, switching to compatibility mode." even more annoying is a MessageBox popping up: "Producer: Internal Error: Can't assign Visual (PIXELFORMAT) to surface. The parameter is incorrect." after clicking the message box away one of several things happens: Either I get an error message: "Error: OpenGL version test failed, requires valid graphics context." And the program dies because of access violations … or The program continues but each draw attempt will just print that realize has to be called first. Obviously the "switching to compatibility mode" does not work out well. Also the whole notion that a dependency library will open MessageBoxes is not something I relish, I prefer exceptions. Under Linux (FC5, ubuntu 6.06) the problem does not exist … also OSG 1.0 and 0.9.9 on Win32 … works just fine, confirming that the problem seems to confined to "RenderSurface_Win32.cpp". Cheers Frank From: Robert Osfield [mailto:[EMAIL PROTECTED] Sent: Monday, June 26, 2006 7:21 AM To: osg users Subject: [osg-users] Working towards OpenSceneGraph-1.1 release Hi All, I would like to start working towards getting the OpenSceneGraph-1.1 release out in July, and would very much appreciate assistance from the community in this effort. 1) The first phase has to be testing out the CVS version of OpenThreads/Produer and OpenSceneGraph on as many platforms as possible, fixing build issues on each of these platforms. Testing against your own applications will also be invaluable. 2) The second phase will be the a series of release candidates. This phase will only commense once we have general consensus that the CVS versions are all compiling and working ok. 3) The thrid phase will be the actual release. This will be once we have a release candidate that looks solid w.r.t source code and the distribution tarballs themselves. The third phase is mainly down to me updating the website in all the appropriate places, but its also about then telling all the various relevant news websites about the new release - something the community can certainly help significantly with. I'd like everything to be complete in time for Siggraph, but the earlier the better as those heading to Siggraph will no doubt be busy, and well I'm gunna be pretty busy preparing for me house move. Phase 1 can start right now. I'll try and turn around build fixes ASAP to get things back and tested. I am still checking in a few submissions, more on this on follow up emails in this thread. Thanks in advance for your help, Robert. _______________________________________________ osg-users mailing list [email protected] http://openscenegraph.net/mailman/listinfo/osg-users http://www.openscenegraph.org/ ________________________________ _______________________________________________ osg-users mailing list [email protected] http://openscenegraph.net/mailman/listinfo/osg-users http://www.openscenegraph.org/ ________________________________ _______________________________________________ osg-users mailing list [email protected] http://openscenegraph.net/mailman/listinfo/osg-users http://www.openscenegraph.org/ ________________________________ _______________________________________________ osg-users mailing list [email protected] http://openscenegraph.net/mailman/listinfo/osg-users http://www.openscenegraph.org/ _______________________________________________ osg-users mailing list [email protected] http://openscenegraph.net/mailman/listinfo/osg-users http://www.openscenegraph.org/
_______________________________________________ osg-users mailing list [email protected] http://openscenegraph.net/mailman/listinfo/osg-users http://www.openscenegraph.org/
