Release/1.5.1

[yk-fujii]

Status

DEVELOPMENT

Description

This PR to resolve the 1.5.0 build problem.
In addition, I will raise the version to 1.5.1.

Target issues

• Build Error @1.5.0 autoware_ai#982 ((pre-merged 9608344 )
• Compilation flag was not enough on ubuntu 14.04 (trusty)

Todos

• Tests
• Tested with Ubuntu 16.04 and ROS Kinetic
• Tested with Ubuntu 14.04 and ROS Indigo by Docker

[amc-nu]

@yk-fujii I checked compilation against Ubuntu 14.04.
However, it's failing on my Ubuntu 16 without GPU setup.
I found the problem in:
https://github.com/CPFL/Autoware/blob/master/ros/src/computing/perception/detection/lib/image/dpm_ttic/CMakeLists.txt#L44

The library is being added in both setups (CPU and GPU), regardless if CUDA was found or not.

[yk-fujii]

@amc-nu
I could not reproduce the problem in the similar environment.(ubuntu 14.04 without GPU Setup)
It seems that code for GPU is added only when CUDA is found.
https://github.com/CPFL/Autoware/blob/master/ros/src/computing/perception/detection/lib/image/dpm_ttic/CMakeLists.txt#L25

"CUDA_FOUND" is not working?

[amc-nu]

@yk-fujii Found the problem, but we will need to modify later this part.
My setup is Ubuntu 16.04, without GPU, but CUDA installed.
CUDA_FOUND is working correctly, but the linking is made manually instead of using the Vars created by FindCUDA.

[yk-fujii]

@amc-nu
Is this PR OK? Can you approve?

Why do you need CUDA in an environment without GPU?
The same problem occurs with other nodes such as ssd, yolo and dpm_ocv.
How would you like to fix it?

[amc-nu]

Is this PR OK? Can you approve?

I'm still checking please wait a few more minutes.

Why do you need CUDA in an environment without GPU?

In some cases, I test only compilation and linking of CUDA code.
In other cases, we can run the nodes you mentioned using only CPU support (extremely slow thought)

[amc-nu]

The fix for the ladybug node still needs to be fixed.

[yk-fujii]

@amc-nu
Thanks!