free(): invalid next size (normal)

[MrOCW]

Hi, running DLIO with a Livox MID-360 on the ros2 branch, on a Jetson Orin AGX

I got a free(): invalid next size (normal) which then led to a process has died [pid 53, exit code -6, .....

Here is the last statistics seen before the crash

[nubertj]

I am facing the same issue, but only when running on nvidia jetson.
It seems to be particularly triggered under high load, resulting in more issues when compiling with clang.

We also experience high load when compiling with clang, as described here: #35

[JunghaWang]

i have the same problem with low performing cpu. something to do with cpu performance?

[wagdeer]

I am currently locating this issue, and I suspect that it is caused by the incorrect memory allocation method of PCL in the ROS 2 branch.

[wagdeer]

Hi, running DLIO with a Livox MID-360 on the ros2 branch, on a Jetson Orin AGX

I got a free(): invalid next size (normal) which then led to a process has died [pid 53, exit code -6, .....

Here is the last statistics seen before the crash

I have identified the root cause of this issue: it arises because multithreading was enabled without proper locking when publishing ROS messages.
You can resolve this by disabling the multithreading for the message-sending section and replacing it with a regular function call. Currently, I have conducted a stress test for 20 hours, and no issues were detected.

[naefjo]

I have identified the root cause of this issue: it arises because multithreading was enabled without proper locking when publishing ROS messages. You can resolve this by disabling the multithreading for the message-sending section and replacing it with a regular function call. Currently, I have conducted a stress test for 20 hours, and no issues were detected.

Hi @wagdeer! I'm running into the same issue. can you point me to the lines you changed? are you refering to these?

direct_lidar_inertial_odometry/src/dlio/odom.cc

Lines 844 to 845 in c8acc37

	this->publish_thread = std::thread( &dlio::OdomNode::publishToROS, this, published_cloud, this->T_corr );
	this->publish_thread.detach();

Also, have you identified further issues with PCL?

Looking forward to your response!