Add check for binary to determine whether to execute script

[decentral1se]

ansible-nvm-node/tasks/main.yml

Lines 19 to 24 in c9f1a77

	- name: Execute the nvm installation script
	command: "{{ nvm_dir }}/install_nvm_{{ nvm_node_version }}.script"
	when: nvm_installation_script_template is defined and nvm_installation_script_template.changed
	# the tag is needed so that linter wouldn't complain about this not being handler,
	# and handlers are so unstable I eventually decided not to use them in this role
	tags: skip_ansible_lint

I saw a case where the script was put in place but for some reason didn't run the install (maybe ctrl-c was the culprit during runtime) and then on 2nd run, it failed at the symlink task because it never ran the script.

[grzegorznowak]

Hey @decentral1se !
Yeah I can see how it might not be handling edge cases properly with half-ran envs etc.
It would be more idempotent had we have a variable registered that denotes whether the installation ran or not at all.

Can you put this before the installation step

- name: Probe the env for the nvm binary
  shell: which nvm || echo "nvm missing"
  register: nvm_probe

and then augment the install step with

when: nvm_installation_script_template is defined and (nvm_installation_script_template.changed or nvm_probe.stdout == "nvm missing")

EDIT: tweaked the when: part for more sense