Apollo: cyber_visualizer can't start (Could not initialize GLX)

[dwrr]

Hi Apollo!

I can't seem to open cyber_visualizer due to an error. Please see my steps below.

Thanks

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 14.04): Ubuntu 16.04
• Apollo installed from (source or binary): source
• Apollo version (1.0, 1.5, 2.0, 2.5, 3.0): 5.0

Steps to reproduce the issue:

• Please use bullet points and include as much details as possible:
• ./docker/scripts/dev_start.sh
• ./docker/scripts/dev_into.sh - now inside docker.
• source cyber/setup.bash
• cyber_visualizer

Result:

dan@in_dev_docker:/apollo$ cyber_visualizer
Could not initialize GLX
Aborted (core dumped)

Supporting materials (screenshots, command lines, code/script snippets):

[fengliu00]

nVidia's OpenGL is not working perfectly in docker. The suggestion is to run cyber_visualizer out of docker, but have to install the necessary dependencies.

[dwrr]

Thanks for the reply @fengliu00. Do you have a list of the required dependencies?

[fengliu00]

1. install libboost.54 if non-existent
2. install ros console bridge 0.2.0, fastrtps 1.5.0 along with fastcdr 1.0.7, gflags 2.2.0, glog 0.3.5, protobuf 3.3.0 and qt 5.5.1 if non-existent

[ShepelIlya]

@fengliu00 Do you have same list for Ubuntu 18.04? All libs versions are the same?

[fengliu00]

Almost as same as 14.04, except QT. QT5.9.8 should be fit for 18.04.

[alexbuyval]

Finally, I have run the cyber_visualizer in the docker.

I used the master branch and the last version of the docker image.

However, you need to have exactly the same version of nvidia drivers in the docker as on the your host.
For instance, I have the nvidia driver 418.67 on the host, so I download the needed version and install it into the docker.

wget http://us.download.nvidia.com/tesla/418.67/NVIDIA-Linux-x86_64-418.67.run
sudo chmod +x NVIDIA-Linux-x86_64-418.67.run
sudo ./NVIDIA-Linux-x86_64-418.67.run -a -N --ui=none --no-kernel-module

[qwetqwe]

Finally, I have run the cyber_visualizer in the docker.

I used the master branch and the last version of the docker image.

However, you need to have exactly the same version of nvidia drivers in the docker as on the your host.
For instance, I have the nvidia driver 418.67 on the host, so I download the needed version and install it into the docker.

wget http://us.download.nvidia.com/tesla/418.67/NVIDIA-Linux-x86_64-418.67.run
sudo chmod +x NVIDIA-Linux-x86_64-418.67.run
sudo ./NVIDIA-Linux-x86_64-418.67.run -a -N --ui=none --no-kernel-module

How did you do that, apollo-master use host's NVIDIA driver by NVIDIA docker, and will be same as host driver. Is it necessary setup NVIDIA driver again?

[alexbuyval]

@qwetqwe Yes, I have installed the Nvidia driver in the docker.

[qwetqwe]

@alexbuyval Do you build visualizer tools by yourself? Because apollo team disabled bazel build visualizer in apollo.sh. What else dependencies do you install in the latest image? QT also disappeared in latest image...

[alexbuyval]

@qwetqwe
I use the following:

sudo apt-get update
sudo apt-get install qt5-default
bazel build //modules/tools/visualizer:cyber_visualizer

[qwetqwe]

@alexbuyval Finally，I succeed in your way.
Thanks a lot.

[ntutangyun]

1. install libboost.54 if non-existent
2. install ros console bridge 0.2.0, fastrtps 1.5.0 along with fastcdr 1.0.7, gflags 2.2.0, glog 0.3.5, protobuf 3.3.0 and qt 5.5.1 if non-existent

Hi, @fengliu00 may i know what ROS distribution is this ? thank you.

[ntutangyun]

1. install libboost.54 if non-existent
2. install ros console bridge 0.2.0, fastrtps 1.5.0 along with fastcdr 1.0.7, gflags 2.2.0, glog 0.3.5, protobuf 3.3.0 and qt 5.5.1 if non-existent

Apollo seems to be migrated to Ubuntu 18.4. but libboost 1.54.0 seems to be no longer supported in ubuntu 1804...

currently I've installed

• fast-rtps 1.9.3
• libgflags-dev
• libgoogle-glog-dev
• Protobuf 3.3.0
• Libpocofoundation9v5
• boost 1.54.0

however, after i source .../apollo/cyber/setup.bash and execute cyber_monitor,
i got following results:

cyber_monitor: symbol lookup error: cyber_monitor: undefined symbol: _ZN3fLS13FLAGS_log_dirE

[zhanghming]

@alexbuyval Hi, I follow your way and build the cyber_visualizer OK, but when I run it, get the following errors:

zhanghm@in_dev_docker:/apollo$ cyber_visualizer
QStandardPaths: XDG_RUNTIME_DIR not set, defaulting to '/tmp/runtime-zhanghm'
libGL error: No matching fbConfigs or visuals found
libGL error: failed to load driver: swrast
Unrecognized OpenGL version
Unrecognized OpenGL version
QOpenGLFunctions created with non-current context
ASSERT: "QOpenGLFunctions::isInitialized(d_ptr)" in file /usr/include/x86_64-linux-gnu/qt5/QtGui/qopenglfunctions.h, line 861
Aborted (core dumped)

Do you know why?

Update:
Sovled this problem, refer this

[storypku]

@alexbuyval Hi, I follow your way and build the cyber_visualizer OK, but when I run it, get the following errors:

zhanghm@in_dev_docker:/apollo$ cyber_visualizer
QStandardPaths: XDG_RUNTIME_DIR not set, defaulting to '/tmp/runtime-zhanghm'
libGL error: No matching fbConfigs or visuals found
libGL error: failed to load driver: swrast
Unrecognized OpenGL version
Unrecognized OpenGL version
QOpenGLFunctions created with non-current context
ASSERT: "QOpenGLFunctions::isInitialized(d_ptr)" in file /usr/include/x86_64-linux-gnu/qt5/QtGui/qopenglfunctions.h, line 861
Aborted (core dumped)

Do you know why?

Update:
Sovled this problem, refer this

Hi @zhanghming and all, the cyber_visualizer issue should have been fix now in lastest master. And it works on realtime kernels if you follow our latest RT kernel installation guide here (English , Chinese)

Thank you.

[daohu527]

Finally, nvidia never asked to install the driver in docker! Ref link