View Issue Details [ Jump to Notes ] | [ Print ] | ||||||||
ID | Project | Category | View Status | Date Submitted | Last Update | ||||
0003842 | ParaView | (No Category) | public | 2006-10-04 17:00 | 2007-02-14 17:19 | ||||
Reporter | Clinton Stimpson | ||||||||
Assigned To | Ken Moreland | ||||||||
Priority | normal | Severity | major | Reproducibility | always | ||||
Status | closed | Resolution | fixed | ||||||
Platform | OS | OS Version | |||||||
Product Version | |||||||||
Target Version | Fixed in Version | ||||||||
Summary | 0003842: crash on Mac connecting to server | ||||||||
Description | Apparently the Mac doesn't like glClearColor being called unless there is a valid current context, or it'll crash. This happens when connecting a ParaView3 client on the Mac to a server. The call stack is glClearColor() vtkOpenGLRenderer::Clear() vtkClientCompositeManager::StartRender() // line 424 StartRender in vtkParallelRenderManager.cxx line 1816 ... vtkRenderWindow::Render() // line 239 where it sends the StartEvent. At the time of the crash, the vtkCarbonRenderWindow::ContextId is NULL, and aglGetCurrentContext() returns NULL, because the vtkCarbonRenderWindow had been Finalized() and moved to a different parent by pqRenderModule::setWidgetParent(). vtkRenderWindow::Start() has not yet been called, and would be called if it could continue pas the crash on to vtkRenderWindow::DoStereoRender(). Found this testing Prism with server connections for a demo. I've yet to connect it, or ParaView3, to any server at all. | ||||||||
Tags | No tags attached. | ||||||||
Project | |||||||||
Topic Name | |||||||||
Type | |||||||||
Attached Files | |||||||||
Relationships | |
Relationships |
Notes | |
(0005193) Clinton Stimpson (developer) 2006-10-04 17:11 |
I also noticed that vtkClientCompositeManager doesn't check that the context for a window is the current context. |
(0005194) Ken Moreland (manager) 2006-10-04 17:24 |
Wait a minute. Why is vtkClientCompositeManager being used at all? I don't think any of the IceT render modules use that class. Make sure that you are using vtkSMIceTDesktopRenderModuleProxy as your render module. |
(0005196) Clinton Stimpson (developer) 2006-10-04 17:41 |
The Mac client tells the Server Manager to use the IceTDesktopRenderModule. However, the client is not compiled with MPI, but is connecting to a server compiled with MPI and IceT. |
(0005197) Clinton Stimpson (developer) 2006-10-04 17:44 |
And yes, the debugger shows that the vtkSMIceTDesktopRenderModuleProxy constructor is called on the client side. |
(0005198) Ken Moreland (manager) 2006-10-04 17:46 |
Is there any possible way to trace back who is instantiating the vtkClientCompositeManager? |
(0005201) Clinton Stimpson (developer) 2006-10-04 18:00 |
Agh. I was debugging why a Mac wouldn't connect to a Linux box where the client issues a start command, so I thought I'd try it local. Only now do I realize that the local build doesn't use IceT. Well, in any case, a pvserver exectuable was created, and it didn't work with the Mac ParaView3 client. |
(0005349) Berk Geveci (administrator) 2006-10-11 15:22 |
We should replace MPIRenderModule with IceTRenderModule even when MPI is off. |
(0005350) Berk Geveci (administrator) 2006-10-11 15:24 |
We should also deprecate MPIRenderModule after ParaView 2.6. |
(0006420) Ken Moreland (manager) 2007-02-14 16:44 |
OK. The MPIRenderModule is now officially deprecated (and removed from the repository). This bug should now be fixed. Clint, please verify that you are no longer seeing any client/server connect problems on Mac, even when IceT is not available, and then close this bug. |
Notes |
Issue History | |||
Date Modified | Username | Field | Change |
2011-06-16 13:10 | Zack Galbreath | Category | => (No Category) |
Issue History |
Copyright © 2000 - 2018 MantisBT Team |