Explore options for interactive debugging

[bschilder]

One weak point of CI/CD platforms that are runs remotely is that you can't directly debug in these environments.

However, running CI/CD on GitHub may offer some unique solutions to this as believe the GH Codespaces use the same VMs as GH Actions. Will explore this further and add to the rworkflows documentation if I find anything useful.

Access your existing codespaces here:
https://github.com/codespaces

Quickly access a repo-specific Codespace with the following link format. However, it seems this version of the Codespace it quite limited as you can't use it to run code in the terminal (it prompts you do create a Codespace with a certain amount of compute and then generates it for you):
https://github.dev/neurogenomics/rworkflows

GH Codespaces docs

• https://github.com/features/codespaces
• https://docs.github.com/en/codespaces/developing-in-a-codespace/creating-a-codespace-for-a-repository
• https://docs.github.com/en/codespaces/developing-in-a-codespace/creating-a-codespace-from-a-template
• https://docs.github.com/en/codespaces/getting-started/deep-dive
• https://docs.github.com/en/codespaces/prebuilding-your-codespaces
• https://code.visualstudio.com/docs/remote/codespaces
• https://docs.github.com/en/codespaces/setting-up-your-project-for-codespaces/adding-a-dev-container-configuration/introduction-to-dev-containers
• https://containers.dev/supporting

Pricing

[bschilder]

Dev containers

You can configure a dev container using a .json, but it's not clear how to translate the GHA yaml files to this format, or whether all the options even translate:
https://github.com/neurogenomics/rworkflows/new/dev?dev_container_template=1&filename=.devcontainer%2Fdevcontainer.json

You can customize these dev envs by adding "Features" to the config file, via the Marketplace window:

Prebuilds

Prebuilds are basically Codepsaces that have already been setup and installed with a particular configuration, reducing the time it takes to spin up a new Codespace:
https://docs.github.com/en/codespaces/prebuilding-your-codespaces

[bschilder]

Potential solutions

1. Create a rworkflows Codespace from the dev container on GitHub Container Registry.

• Limitations
  • The way I create Docker containers atm is rather inefficient and doesn't follow all steps used in the action beforehand; Make Docker step less redundant #9
  • Probably won't have access to all the same inputs. and env. variables that are accessible in the GHA VM.

[bschilder]

Got a version of the rworkflows codespace working. Not sure how closely it mimics the GHA VM, but i'm at least able to do some debugging with conda:

I also noticed that the env.GITHUB_TOKEN is indeed present within the Codespace:

Remaining issues:

1. Not sure how closely it mimics the GHA VM.
2. Not sure how similar the ghcr.io/devcontainers/features/conda:1 feature I'm using mimics the conda-incubator/setup-miniconda@v2 action.
3. Additional setup steps are still required for Visual Studio Code, such installing R plugins. Looking for ways to set this up automatically. Setting up dotfiles may help. Or better yet, specifying to install these extensions within the devcontainer.json file itself. More specific instructions here.
4. Certain VSC key bindings don't work (e.g CMD+W). Tho creating a Codespace App may help with this.

[bschilder]

I've set up a basic dev container config file for rworkflows. It doesn't perfectly replicate the GHA runners and the actions used to install deps, but it does get somewhat close. This can be useful for debugging directly within GitHub.

Also created a use_codespace helper function to set up a dev container config file.