rtabmap_ros icon indicating copy to clipboard operation
rtabmap_ros copied to clipboard

Unable to colcon build rtabmap_sync package

Open sa32953 opened this issue 1 year ago • 3 comments

Hi, I am trying to build from source as mentioned here in ros2 branch of the repo: https://github.com/introlab/rtabmap_ros/tree/ros2.

I am building it on Ubuntu 22.04 on a VM with ROS2 Humble as distribution. I am unable to build a specific pakcages i.e. rtabmap_sync. When I try to build it, it completes 21% and the system freezes every time. Here's the colcon command I am running in the workspace.

colcon build --symlink-install --cmake-args -DRTABMAP_SYNC_MULTI_RGBD=ON -DRTABMAP_SYNC_USER_DATA=ON -DCMAKE_BUILD_TYPE=Release --packages-select rtabmap_sync

Any hints what could be the reason for this?

I also tried building by limiting the parallel workers using the below command. Same issue persists!

colcon build --symlink-install --cmake-args -DRTABMAP_SYNC_MULTI_RGBD=ON -DRTABMAP_SYNC_USER_DATA=ON -DCMAKE_BUILD_TYPE=Release --packages-select rtabmap_sync --parallel-workers 2

sa32953 avatar Dec 10 '24 15:12 sa32953

You may ran out of RAM. Try:

export MAKEFLAGS="-j1"

before doing colcon build. Increase number of threads accordingly to your RAM.

matlabbe avatar Dec 14 '24 18:12 matlabbe

Hey @matlabbe. Thanks for the reply. I will try that out. Do you recommend any optimal systems requirements for compiling and runtime? Also I dont have a GPU, do you recommend running it without GPU?

sa32953 avatar Dec 21 '24 14:12 sa32953

GPU is not required, unless you want to visualize 3D point clouds live. The underlaying visual odometry and loop closure / grpah optimization performance should be fairly the same without GPU.

With an i7, it would work as expected at runtime. For compilation, having at least 4 GB of RAM is necessary (when using make -j1), but having 16GB or 32 GB you can build more files in parallel.

matlabbe avatar Dec 22 '24 22:12 matlabbe