The eclipse plugin directory '/usr/local/eclipse/plugins/' is owned by root and the eclipse container is running by default as user. Therefore, it is not possible to update the plugins by clicking 'Help/Checking for Updates'. I will try to fix this by changing the ownership of the plugin directory to the running user.

and change apt installation to git clone and build like in the cmake module

Starting vscode-1.49.1 with

docker run -d -v /tmp/.X11-unix:/tmp/.X11-unix:ro -e DISPLAY braintwister/ubuntu-20.04-vscode-1.49.1:0.4

shows following error message using docker logs:

/usr/share/code/bin/../code: error while loading shared libraries: libxcb-dri3.so.0: cannot open shared object file: No such file or directory

Directory '/usr/local/cuda-10.1/nsightee_plugins' is missing.

JupyterLab is the next-generation user interface of Jupyter and really great.

https://github.com/jupyterlab/jupyterlab

Release of version 3.0 is planned on 31 Jul 2020.

Running the docker image braintwister/ubuntu-18.04-cuda-9.2-cmake-3.12-gcc-7-conan-1.8-docker-18.06-eclipse-cpp-2018-09 on a docker host with CUDA 9.2, following error occurs:

WARNING: Image for service eclipse was built because it did not already exist. To rebuild this image you must use `docker-compose build` or `docker-compose up --build`.
Creating eclipse-cpp_eclipse_1 ... error

ERROR: for eclipse-cpp_eclipse_1  Cannot start service eclipse: OCI runtime create failed: container_linux.go:348: starting container process caused "process_linux.go:402: container init caused \"process_linux.go:385: running prestart hook 1 caused \\\"error running hook: exit status 1, stdout: , stderr: exec command: [/usr/bin/nvidia-container-cli --load-kmods configure --ldconfig=@/sbin/ldconfig.real --device=all --compute --utility --require=cuda>=10.0 brand=tesla,driver>=384,driver<385 --pid=25159 /var/lib/docker/aufs/mnt/d8ebef6b2e08d52c075d029202420724d033c1f3d7542c59d0bec296db60d5a7]\\\\nnvidia-container-cli: requirement error: invalid expression\\\\n\\\"\"": unknown

ERROR: for eclipse  Cannot start service eclipse: OCI runtime create failed: container_linux.go:348: starting container process caused "process_linux.go:402: container init caused \"process_linux.go:385: running prestart hook 1 caused \\\"error running hook: exit status 1, stdout: , stderr: exec command: [/usr/bin/nvidia-container-cli --load-kmods configure --ldconfig=@/sbin/ldconfig.real --device=all --compute --utility --require=cuda>=10.0 brand=tesla,driver>=384,driver<385 --pid=25159 /var/lib/docker/aufs/mnt/d8ebef6b2e08d52c075d029202420724d033c1f3d7542c59d0bec296db60d5a7]\\\\nnvidia-container-cli: requirement error: invalid expression\\\\n\\\"\"": unknown
ERROR: Encountered errors while bringing up the project.

At the moment the user information is only locally available in the main entrypoint script:

By exporting them they will be available also in the module entrypoint scripts. To be sure not disturbing systems, where these variables are used, we prefix BT_ to them.

In the base image following will be configured:

git config user.name "$GIT_USER_NAME"
git config user.email "$GIT_USER_EMAIL"
git config credential.helper 'cache --timeout=3600'

Error message by building tensorflow-13.1:

Cuda Configuration Error: Cannot find cuda library libcublas.so.10.1

Instead of updating the version argument in the Dockerfile, it should be taken from the module name. Therefore, the full version string is needed, e.g. not neglecting the patch version.

I want to add lcov and gcovr to code coverage, will you help me on this?

The module versions are not enough. Changes in the configuration will overwrite published images. Therefore, I will provide image versioning to ensure that published images will not change and it is easier to determine their build state.

The module cmake and conan will be removed and the packages will be implicitly install by pip in the ubuntu base image.

If the module provide patch versions (x.y.z), the docker module providing only a upper level of version, e.g.minor version (x.y) should automatically take the largest patch level available during build-time.

vscode is starting in detached mode, which closes the docker container immediately after vs code was started. The option --wait will fix this, but the container will still exit when all tabs are closed. As a workaround I keep the container running as long as the PIDs, got with pidof code, are active.

Sometimes eclipse.org is very slow or even not reachable. In this cases it would be helpful to switch to a other download site. Also the checksum of the downloaded file should be checked.

To simplify the image list we can use a combinatorial list.

The line

 - ['ubuntu-18.04', ['gcc-9', 'clang-9'], ['eclipse-cpp-2019-06', 'vscode-1.41']]

will build following images:

• ubuntu-18.04-gcc-9-eclipse-cpp-2019-06
• ubuntu-18.04-clang-9-eclipse-cpp-2019-06
• ubuntu-18.04-gcc-9-vscode-1.41
• ubuntu-18.04-clang-9-vscode-1.41

At the moment all intermediate images must be explicitly listed in front of the final image. For instance if the image ubuntu-16.04-cmake-3.11-gcc-7-conan-1.2-docker-18.03-eclipse-cpp-4.7.3 is needed, the image list must be

• ['ubuntu-16.04']
• ['ubuntu-16.04', 'cmake-3.11']
• ['ubuntu-16.04', 'cmake-3.11', 'gcc-7']
• ['ubuntu-16.04', 'cmake-3.11', 'gcc-7', 'conan-1.2']
• ['ubuntu-16.04', 'cmake-3.11', 'gcc-7', 'conan-1.2', 'docker-18.03']
• ['ubuntu-16.04', 'cmake-3.11', 'gcc-7', 'conan-1.2', 'docker-18.03', 'eclipse-cpp-4.7.3']

To be more clear, the image list should only contain the final images. Additionally, only the final images have to be pushed to the docker repository, which also reduce the needed storage.