Leave Helm unconfigured during Terraform entrypoint?

[mogul]

Regarding https://github.com/abdennour/dockerfiles/commit/b227fff369c8842608d696da3b747fb3e15731a7 : The purpose of adding the Helm CLI binary to the Terraform image was to allow easy use of the Terraform helm_provider. This change goes further than that, though; it mirrors the Helm Dockerfile in that you can specify plugins, repositories, etc. as env vars.

That caused some problems per PR https://github.com/abdennour/dockerfiles/pull/4; now the Dockerfile isn't simply a wrapper for calling terraform without installing it. The Helm provider already has attributes for specifying the plugins_path, repositories, etc.. So it feels like doing Helm setup during the Terraform entry point is the wrong thing. Rather than specifying these things through the Docker environment variables, it feels "more IaC" to have them specified in the .tf that sets up the provider. What do you think?

[abdennour]

I totally agree. As far as i know, env vars prefixed by TF_ will be used by Terraform as values for TF variables. I am not sure if mentioning this feature will help here or not.