dsd-dbs / capella-dockerimages Goto Github PK

If the user has set wrong permissions for the Capella archive, the build fails. We should set the permissions properly.

• https://github.com/hadolint/hadolint
• https://github.com/aquasecurity/trivy

In some sessions the user does see "Invalid license". The license does not get injected in the right way.

As discussed in a conversation on #108 (comment) , ENV and SHELL are inherited when using FROM, so we can remove these two lines in all of our dockerfiles having base as the base image (which should be all files).

The description editor in Capella doesn't work with the latest GTK version. For the installation of older GTK versions we depend on the focal apt repository of Ubuntu. Versions for architectures like arm are missing in the registry. It would be nice to build them from source.

The project name is escaped to BZB2.0%40ARCH, but T4C expects BZB2.0@ARCH.

We should add an option to install a specific Java version instead of using the integrated Java version of Capella.

Add Github templates for diagram cache and model validation (same behaviour as Gitlab CI templates: https://github.com/DSD-DBS/capella-dockerimages/blob/main/ci-templates/gitlab/README.md)

Hi,

I recently noticed that there exists a simple line formatting error in teams for capella web instance. For example in the picture below there is an arrow which has added by line formatting i.e. using right click -> format -> arrow type.

The issue here is that, the arrow or any other formatting with regards to line thickness or line type done on the line vanishes on session restart i.e. Starting a session post termination.

The metrics script on the remote container runs 100% CPU.

The script is busy-looping at the end of the file. This is not needed for a simple HTTP server that should only perform a bit of work when invoked.

When the script is invoked via EASE, one doesn't get any stacktrace. We should move all parts which are unrelated to the workspace import into a separate script. When running externally, we can get the stracktrace, which makes debugging a lot easier.

Users want to see their original commits in the history of their GitLab repository. Currently the backup job is just doing one commit and all individual commits with their messages are not included in the model repository on GitLab. Having the original commits migrated, enables many features/actions available on GitLab. For e.g. issues on the model repo, or Jira tickets can be referenced and their status can be changed (e.g. closed).

Clone command/submodule command cannot update this submodule:

git submodule update --init --recursive

remote: Repository not found.
fatal: repository 'https://github.com/DSD-DBS/pyease.git/' not found
fatal: clone of 'https://github.com/DSD-DBS/pyease.git' into submodule path 'DS_Capella/capella-dockerimages/ease/pyease' failed

What is pyease and is it a closed component?
Thanks!

We should run pytest for the read-only and remote containers in Github Actions.

The success message is different for Capella 5.0.0.

Currently, we load all archives for each Capella version into the Docker build context. This is not a good solution as it slows down the whole build process. We need to generate a dockerignore file before we start the build in the Makefile, which ignores all other Capella versions, so that only the relevant version is loaded into the build context.

The images have to be tested with Capella 6.0. Most likely, it will just work.

We need a service within a session container (capella-dockerimages/remote) to expose user inactivity time (in seconds) on port 9118 (configurable) in OpenMetrics format

The service start should be managed by supervisord so that the service is available even before a user is logged in

Currently, the TeamForCapella backup flattens the directory structure, leading to failed backups when an image appears two times in different directories.

With -archiveProject, the directory structure is not flattened. We can use -archiveProject and unzip the archive automatically.

How to reproduce: Start the remote docker container, try to type in '@'. Nothing happens.

To use the CDO session view, administrators need to enable the Team for Capella - Administration in Window > Preferences > General > Capabilties.

This can be automated, administrators should have it enabled automatically.
Eventually, we can also pre-fill the server location.

Many Capella models depend on specific dropins.

We should offer three sets of dropins for all Capella based images:

• none - No dropins are included (how it is handled currently)
• selected - The following dropins are included:
  • System to Subsystem Transition
  • XHTML Documentation Generation
  • Filtering
  • Property Values Management Tools (PVMT)
  • Requirements Viewpoint
  • Textual Editor
• all - All open source dropins (with EPL license): https://www.eclipse.org/capella/addons.html

Hi guys!
Thank you very much for this images are really useful. (I was getting mad about the downgrade dependencies of gtk) .

My little contribution is the following (how I use it):
I've created a docker-compose.yml like the following:

version: "3"
services:
  capella:
    image: capella/base:latest
    environment:
      - DISPLAY=${DISPLAY}
      - GTK_THEME=Adwaita
    volumes:
      - /tmp/.X11-unix:/tmp/.X11-unix
      - ./git:/git
      - ./extworkspace:/extworkspace
    network_mode: host
    entrypoint: ./capella/capella -data /extworkspace

Basically it sets the DISPLAY variable accordingly to execute GUIs application inside the docker and sets my git and an external workspace to use by default.

here some reference on GUI apps on docker.
https://gursimar27.medium.com/run-gui-applications-in-a-docker-container-ca625bad4638

1 projects exports failed in the log should lead to error code 1.

Our Gitlab CI template should get a bit more advanced:

Split each jobs into three stages: build, test (optional) and push
Either run all steps in one job in a sequence (I'd say the best solution) or use nested dynamic pipelines: https://www.youtube.com/watch?v=C5j3ju9je2M

In Capella 6.0.0, the arguments of the importer script have been changed. Thus, the backups fail as we haven't noticed it.
We need a full integration test.

Add a new pytest for the backup job:

• Spawn the T4C server in the pytest
• Populate the server with an existing project. The data has to be placed in a specific location, e.g. t4c/server/tests/data/$CAPELLA_VERSION
• Start a Git server. The git http-backend command might be helpful.
• Run the backup Docker container
• Check if the Git server contains the project with all important files.

In addition, the backup image should have support for Capella 6.0.0:

• Fix the backup Dockerfile / Python script to support Capella 6.0.0.

In a team-for-capella model, I tried to add a model element link to a logical function description. The link dialog showed up, I successfully selected a model element and confirmed the link dialog. Then, when I'd have expected the link to be added to the description text, the description window freezes and I get this error prompt:

The Capella session doesn't recover after that, repeatedly error messages pop up when trying to edit any description. The only way out is to close it and open a new one.
The error also occurs when I don't select a model element and cancel the link dialog.

Add a quick start section to build all Dockerimages in their default configurations. We should write a Makefile with targets for all images.
Additionally, we should have some goals to run images.

Check compatibility with Capella version 6.1.0.

Our number of images has grown, leading to long build times. We can simplify it a bit:

• Move the CLI features into capella/base and t4c/client/base, deprecate capella/cli and t4c/client/cli.
• Move the readonly features into capella/base as "preparation script", deprecate capella/readonly
• Move backup and exporter features to t4c/client/base, deprecate t4c/client/backup and t4c/client/exporter.

Consider the environment variables that are used in the Makefile as variables in the test for the image.
DOCKER_REGISTRY and DOCKER_TAG are the interesting ones.

Git repositories should be cloned into /workspace/git instead of /home/techuser/git. In the Git preferences, this can be changed. The workspace should be prepared properly.

Production builds should only work on tags - not on branch names. A simple regex v[0-9]+\.[0-9]+\.[0-9]+ is ok as check.
This prevents "messing" up the production registry.

Change the username to login in the entrypoint. This should allow login with a different username in xrdp.

In case of a SIGTERM, Capella should be closed properly.
Instructions: https://stackoverflow.com/questions/6391705/how-to-cleanly-shutdown-eclipse-from-linux-command-line

• Catch SIGTERM, terminate the tool and supervisord
• Implement a graceful termination of Capella (see StackOverflow)

Add image for Papyrus: https://www.eclipse.org/papyrus/download.html

The read-only session fails when using a starting / in the entrypoint.

While following the build guide on my quest to get a running capella-ease container I am stuck in download-eclipse-packages-manually. Running the presented commands with capellac ... won't work since my machine doesn't know of this command. Yet there is no explanation on how to use capella/eclipse to download all files from a given Eclipse software repository URL.

Add the following lines to all Dockerfiles:

SHELL ["/bin/bash", "-euo", "pipefail", "-c"]
ENV SHELL=/bin/bash

XRDP seems to run as non root, so there should not be a dependency on the techuser. We should be able to spawn the server regardless of the user.

https://github.com/neutrinolabs/xrdp/wiki/Running-the-xrdp-process-as-non-root

While the authentication for normal Git repositories works fine, cloning a Git LFS repository from the Capella GUI results in an error. It tries to use the GIT_AKSPASS, which fails because the GIT_USERNAME and GIT_PASSWORD environment variables are not set.

The pure::variants license path /home/techuser/pure-variants-5 is hardcoded for pure::variants version 5. It is should depend on the version, for version 6 it should be /home/techuser/pure-variants-6.

Possible workaround:

1. The user gets this prompt when opening a model:
   
2. Click "yes"
3. Navigate to "Home":
   
4. Double-click on pure-variants-5
   
5. Double-click on de.license
6. Accept the certificate:
   

Due to the GIT_ASKPASS configuration, the usage of the Git integration of Eclipse is not possible. However, users would like to clone and push Git repositories with authentication. Therefore, we have to set a proper git credential manager during preparation.

It is possible to set an entry point (GIT_REPO_ENTRYPOINT) that specifies where the .aird file should be placed. Therefore, we should also perform a backup test using this environment variable.

It should be possible to run pure::variants in a separate container without Capella.

After reconnecting to a session, copy/paste doesn't work anymore.

Some potential workarounds are listed here: neutrinolabs/xrdp#1188