UEFI: Dell v: A34 date: 10/19/2020
CPU: Topology: 2x Quad Core model: Intel Xeon E5-2623 v3 bits: 64 type: MT MCP SMP L2 cache: 20.0 MiB
Speed: 1200 MHz min/max: 1200/3500 MHz Core speeds (MHz): 1: 1204 2: 1198 3: 1198 4: 1198 5: 1198
6: 1198 7: 1200 8: 1198 9: 1241 10: 1199 11: 1197 12: 1198 13: 1197 14: 1197 15: 1198 16: 1198
Graphics: Device-1: NVIDIA GK208B [GeForce GT 730] driver: nvidia v: 470.129.06
Device-2: NVIDIA GK210GL [Tesla K80] driver: nvidia v: 470.129.06
Device-3: NVIDIA GK210GL [Tesla K80] driver: nvidia v: 470.129.06
Display: x11 server: [X.Org](http://X.Org) 1.20.13 driver: nvidia unloaded: fbdev,modesetting,nouveau,vesa
resolution: 1280x1024~60Hz
OpenGL: renderer: NVIDIA GeForce GT 730/PCIe/SSE2 v: 4.6.0 NVIDIA 470.129.06
Audio: Device-1: Intel C610/X99 series HD Audio driver: snd\_hda\_intel
Device-2: NVIDIA GK208 HDMI/DP Audio driver: snd\_hda\_intel
Sound Server: ALSA v: k5.13.0-51-generic
Network: Device-1: Intel Ethernet I217-LM driver: e1000e
IF: enp0s25 state: down mac: 48:4d:7e:f6:a7:77
Device-2: Intel I210 Gigabit Network driver: igb
IF: enp8s0 state: up speed: 100 Mbps duplex: full mac: 48:4d:7e:f6:a7:92
Drives: Local Storage: total: 465.76 GiB used: 31.68 GiB (6.8%)
ID-1: /dev/sda vendor: Seagate model: ST500DM009-2F110A size: 465.76 GiB
Partition: ID-1: / size: 456.96 GiB used: 31.68 GiB (6.9%) fs: ext4 dev: /dev/sda2
Sensors: System Temperatures: cpu: 49.0 C mobo: N/A sodimm: 26.0 C gpu: nvidia temp: 40 C
Fan Speeds (RPM): cpu: 1001 gpu: nvidia fan: 40%
Info: Processes: 434 Uptime: 13d 2h 56m Memory: 62.82 GiB used: 4.03 GiB (6.4%) Shell: bash inxi: 3.0.38
brorson@brorson-Precision-Tower-7910:~/VTKLambdaPoly$
---
### Author:  [@andreasbuykx](https://discourse.vtk.org/u/andreasbuykx)
#### Post date: [July 4, 2022, 1:30pm UTC](https://discourse.vtk.org/t/crash-when-starting-vtkrenderer/8851/2 "2022-07-04T13:30:20Z")
Do you really mean that you restart the vtkRenderer or is it the vtkRenderWindow? It looks like that from the stack trace.
Is it really necessary to restart the renderer (renderwindow) and interactor? A more conventional approach would be to have a callback of some sort update the pipeline and re-render the window.
I’m not sure if this is the case here but initializing the window will claim graphics resources which are held until destruction. If you still hold to some resource-owning instances then re-starting a new window may lead to conflicts.
---
### Author:  [@Stuart\_Brorson](https://discourse.vtk.org/u/Stuart_Brorson)
#### Post date: [July 4, 2022, 1:50pm UTC](https://discourse.vtk.org/t/crash-when-starting-vtkrenderer/8851/3 "2022-07-04T13:50:05Z")
Thank you for the ideas.
Regarding restarting the vtkRenderer or the vtkRenderWindow, my reading of the documentation is that you start the renderer by invoking the Start() method on the renderWindow – the Start method in the renderer itself is an abstract class. Therefore, that’s what I did. But others who are more knowlegable that I can correct me if I am wrong.
Regarding callbacks, yes, I am considering that, but my understanding (again, potentially wrong) is that one can either use the standard event loop with callbacks or write a customer InteractorStyle which implements the desired mouse behaviors. I used the second method for reasons of history. Is it better to use the callback method? And why would my InteractorStyle cause the Renderer to barf?
Here’s what I am doing. The crash happens the second time renWin-\>Render() is called.
* * *
while(1) {
std::cout \<\< “----------------------------------------” \<\< std::endl;
std::cout \<\< "Computing polynomial … " \<\< std::flush;
computeMandelbrot(rImageData, N); // My fcn to compute the poly over the whole complex plane.
std::cout \<\< “done!\n” \<\< std::endl;
```
// Tell pipeline to update
colorComplexPlane->Modified();
imageMapper->Modified();
imageActor->Modified();
renderer->Modified();
// Start rendering thread
cout << "Start rendering thread......" << endl;
renWin->Render();
// Start interactor event loop
cout << "Start interactor event loop......" << endl;
iren->Start();
// If we get here it's because an event stopped the loop.
// Handle the captured events then loop again.
```
}
* * *
Thanks!
---
### Author:  [@andreasbuykx](https://discourse.vtk.org/u/andreasbuykx)
#### Post date: [July 4, 2022, 1:57pm UTC](https://discourse.vtk.org/t/crash-when-starting-vtkrenderer/8851/4 "2022-07-04T13:57:36Z")
You should start iren only once, outside the loop. The separate Render is not necessary since the interactor will take care of that. Also all the Modified calls except the one on imageMapper are unnecessary.
See here [https://kitware.github.io/vtk-examples/site/Cxx/Interaction/MouseEventsObserver/](https://kitware.github.io/vtk-examples/site/Cxx/Interaction/MouseEventsObserver/) for inspiration on how to respond to mouse events.
---
### Author:  [@Stuart\_Brorson](https://discourse.vtk.org/u/Stuart_Brorson)
#### Post date: [July 4, 2022, 2:44pm UTC](https://discourse.vtk.org/t/crash-when-starting-vtkrenderer/8851/5 "2022-07-04T14:44:59Z")
Thank you for the suggestions. I tried them out and found I get different crash behavior. That’s progress of a sort 🙂
The latest code is below. When I do something with the mouse then the render window crashes, but my program continues running and captures a stream of (nonexistant) mouse events. I guess the render window thread died but the main thread lives on…
One interesting observation I made based on your comments is this: My custom interactor inherits from vtkInteractorStyleImage. This is the version which crashes. When I change the custom interactor to inherit from vtkInteractorStyleTrackball, then I get no crash – but my custom methods are not invoked. Rather, the standard Trackball interactor mouse bindings seem to be used. Not sure what’s up with that… Any thoughts? And why would the interactor determine whether the renderer crashes or not?
I will try using callbacks instead of a custom interactor.
Thanks, Stuart
* * *
vtkSmartPointer iStyle =
vtkSmartPointer::New();
iren-\>SetInteractorStyle(iStyle);
// Start rendering thread
cout \<\< “Start rendering thread…” \<\< endl;
renWin-\>Render();
//-------------------------------------------------------------------
while(1) {
std::cout \<\< “----------------------------------------” \<\< std::endl;
std::cout \<\< "Computing polynomial … " \<\< std::flush;
computeMandelbrot(rImageData, N); // Compute the whole set.
std::cout \<\< “done!\n” \<\< std::endl;
```
// Tell pipeline to update
colorComplexPlane->Modified(); // Image won't update without this.
imageMapper->Modified();
//imageActor->Modified();
//renderer->Modified();
// Start interactor event loop
cout << "Start interactor event loop......" << endl;
iren->Start();
// If we get here it's because an event stopped the loop.
// Process pending events.
```
}
---
### Author:  [@Stuart\_Brorson](https://discourse.vtk.org/u/Stuart_Brorson)
#### Post date: [July 10, 2022, 1:42pm UTC](https://discourse.vtk.org/t/crash-when-starting-vtkrenderer/8851/6 "2022-07-10T13:42:55Z")
Hello again.
Thank you for your suggestions regarding my crash problem. I have fixed the problem.
My old program would start the event loop, then when the user caused a mouse event, the program would exit the loop, handle the event (in this case, recompute a result), then loop back and restart the event loop – which would call render and crash.
Based on your suggestions, I now start the event loop and leave it running. Then, when the user causes a mouse event, the callbacks call the thing which recomputes the result and then re-renders the image. I will guess this is how VTK is supposed to work.
Now that my program is working, please allow me to show it off. The working program is a CUDA-based Mandelbrot set explorer using VTK for visualization. I put it up on GitHub for public enjoyment here:
> **[GitHub - brorson/MandelbrotExplorer: Interactive Mandelbrot set explorer...](https://github.com/brorson/MandelbrotExplorer)**
>
> Interactive Mandelbrot set explorer using CUDA acceleration and VTK for graphics/interaction - GitHub - brorson/MandelbrotExplorer: Interactive Mandelbrot set explorer using CUDA acceleration and V...
Hopefully, it shows off the correct way to use VTK to recompute and update an image as the user interacts with it.
Thanks,
Stuart Brorson