tfe_workspace_run: Adding an `apply` block requires two deploys

[tmatilai]

Terraform Configuration Files

Initial configuration:

resource "tfe_workspace" "test" {
  name = "foo"

  # VCS and other configurations
  # [...]
}

resource "tfe_workspace_run" "test" {
  workspace_id = tfe_workspace.test.id

  destroy {
    manual_confirm = false
  }
}

Next iteration:

resource "tfe_workspace_run" "test" {
  # Same as before
  # [...]

  apply {
    manual_confirm = false
  }
}

Expected Behavior

The tfe_workspace_run resource would have been recreated (because of the ID change), and a run triggered.

Actual Behavior

1. A Terraform run with the initial configuration adds the workspace_run resource to the state, with a random ID.
2. After adding an apply block to the configuration, the next terraform apply detects that the apply block doesn't exist in the state, but doesn't check the new configuration. The resource is kept in the state with the same random ID, updated with the new ( apply) configuration, but nothing else is done as the resourceTFEWorkspaceRunUpdate() does nothing.
3. Next Terraform run (without any config changes), finds the apply block in the state, but obviously fails to find a Run matching the random ID. Thus it drops the old state, and lets Terraform to create a new resource, now also triggering the run to the workspace.

[tmatilai]

If #1604 is accepted, there might not be many real world use cases for this. But I think that the current behavior is still wrong. 🙂