SEGV when running code on roborio

Yesterday when running our code on the robot, we were randomly getting a SEGV fault on the roborio.
It also seemed to occur randomly, happening when a SmartDashboard line was added, or even when a function call was removed. The stack trace that preceded it always pointed to the line where the vision measurement is added to the pose estimation.

     var vposeOpt = poseSupplier.getEstimatedGlobalPose(getPose());
      if(vposeOpt.isEmpty()) return;
      var vpose = vposeOpt.get();
      SmartDashboard.putString("vpose", vpose.estimatedPose.toString());
   //this line is what the stack trace points to \/
poseEstimator.addVisionMeasurement(vpose.estimatedPose.toPose2d(), 
vpose.timestampSeconds);
      visionPose= vpose.estimatedPose.toPose2d();

Surrounding that line with try/catch with Exception in the catch statement did not solve it.
The code is at code
Right now the code does not segfault but uncommenting this line
causes it to segfault.

We are also using the following software/libs
roborio@2023.1
wpilib@2023.4.2
pathplannerlib@2023.4.3
photonlib@2023.4.2
wpilibNewCommands@1
REVlib@2023.1.1
PhoenixProAnd5@23.0.4

roborio log:
hs_err_pid1682.txt (59.0 KB)

See this thread: WPILIB Pose Estimators killing robot code?

A fix has been merged but it’s not in a release yet.

We are using java, but we can test with the latest artifact from github actions.

I wasn’t able to find a combined build artifact, only the separate hal/wpimath/etc files. The pose estimation also works perfectly without that one line.

Oh, sorry, didn’t read closely enough. That is bizarre for that SmartDashboard line to cause a SIGSEGV from Java. Can you capture a core dump? Instructions are in this post: WPILib 2023.4.2 Release - #6 by Peter_Johnson

We were having the issue yesterday night but for some reason I am not able to recreate the error. Everything is working including pose estimation. No clue what caused it or what fixed it but if it happens again I will get the core dump.