-
Notifications
You must be signed in to change notification settings - Fork 825
Program crash upon exiting visualizer #760
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Comments
Judging by the odd behavior of no segfault, could the visualizer window somehow be inheriting the window id of the main window and exiting both as a result? EDIT: I guess not.
|
For "running a visualization works but upon exiting the visualizer window the entire program exits.", do you mean the visualization shown when doing Edit->View point clouds? Currently I have an issue on Ubuntu 20.04 default VTK version, which crashes rtabmap everytime we close a visualization window in which we enabled EDL shader. The crash doesn't happen with more recent VTK versions. |
Yeah exactly that. View point clouds is the path that leads to the crash condition when closing the window. I can at least confirm it still does it on VTK 8.2.0 (if that is the issue) |
I cannot reproduce the problem with:
when opening a database and do Edit->View Clouds. Just if I enable EDL shader in the view cloud window then close it:
With the following build, there is no crash:
|
So in relation to #769, PCL 1.10 is what I need? I've been unable to retest the crash as I cannot compile rtabmap against PCL 1.11/1.12/master (It was working on 1.11 at some point) VTK may be harder as 8.2.0 is what my system ships with. The struggle of up-to-date libraries continues :/ |
At some point, PCL maintainers moved from VTK6-7-8 to VTK9 at PCL 1.11. It is unclear which PCL version is the most compatible with VTK 8.2, as default VTK version on Ubuntu 18.04 is VTK6, then on Ubuntu 20.04 it is VTK7. On Ubuntu 21.10, it is still VTK7, and PCL version is 1.11.1: https://packages.ubuntu.com/impish/libpcl-dev There was a change in how Opengl windows were handled between VTK7 and VTK8 (along migration from Qt4 to Qt5), and on PCL side, they fixed it for VTK9 directly. |
Video
|
Do you know which Qt version it is used on that machine? I'll start a build with VTK 8.2 to see...
|
QMake version 3.1 |
Did a test with VTK 8.2.0:
no problem. I did the same steps than in your video, no crash. The visualization window did appear in front of the main window. On Ubuntu 20.04, it is Qt 5.12.8. On my Mac build, I do have the problem of the visualization window appearing behind the main window, Qt version is 5.15.2, however at that time I didn't push more the debugging because it is on Mac Silicon, so I was expecting that Qt would be not stable yet, but looks like it is more a Qt 5.15.2 compatibility problem (independent of the os). But there is no crash at least, here is the config on the Mac (last time I built with default homebrew packages):
|
Strange. Thanks for checking. Still trying to see if I can get a good breakpoint debug trace. It doesn't segfault so it's been trickier trying to hook the correct spot (This is also why I initially thought the child window was somehow controlling the main, because all it prints is the X window error) |
So on a smaller scan it doesn't crash fully but locks up when exiting visualizer. Seems to get stuck in xcb_wait_for_special_event. I can continue to Ctrl C out of the loop and it just repeats.
|
Great news! (Kinda...) No longer seems to crash on VTK 9.2 and QT 5.15.2 https://gist.github.com/parkerlreed/09af1075135fd25108a9344607fa0552 I'll close this out for now. I'll have to open a new ticket as now dragging the visualizer window to move around the scene is dragging the entire window... |
Uh oh!
There was an error while loading. Please reload this page.
Regarding #425 (comment)
I finally got around to testing this on an Ubuntu setup.
Rundown: Performing a scan (Freenect2 in my case) and running a visualization works but upon exiting the visualizer window the entire program exits.
Ubuntu 20.04
ROS Noetic
rtabmap-ros (still using as standalone/had to recompile in ros mode for freenect2)
All of the deps from the noetic package
I built in debug mode and installed all the dbgsym packages I could think of, but it doesn't seem to segfault so I haven't been unable to get a backtrace,
Judging by the current output, it may have never been a vtk issue to begin with.
The text was updated successfully, but these errors were encountered: