refactor(api): add info to debug warning on how IS_ROBOT is determined

[theosanderson]

If you run a robot without RUNNING_ON_PI in the environment you get this warning message (and substantially reduced functionality), but it is not clear how to resolve it. This provides that information. In my case it was because I was invoking one Python script from another with subprocess launching a Python script without the conventional shell.

overview

changelog

review requests

risk assessment

[sfoster1]

Looks good to me, thanks!

Out of curiosity, what situation were you in where running this on the robot didn't have the environment variable set? Shells on the ot-2 in our buildroot build should have it set

[theosanderson]

Awesome.

I wrote a basic launcher script which essentially lists .py files in a directory for you to choose a protocol (requiring input) to run, which it does with subprocess python script.py. Then it was a bit confusing to debug why this issue happened specifically in that case (confounded by wondering if it was the unrelated GPIO-lock issue).

[sfoster1]

Awesome.

I wrote a basic launcher script which essentially lists .py files in a directory for you to choose a protocol (requiring input) to run, which it does with subprocess python script.py. Then it was a bit confusing to debug why this issue happened specifically in that case (confounded by wondering if it was the unrelated GPIO-lock issue).

Hmm, and that was running from ssh shell? or from a jupyter terminal? I think subprocess should mirror the env vars from the parent I guess unless you're specifying other ones.

[theosanderson]

Ahhh.. you're right, misdiagnosis on my part. The issue is that I was starting the launcher script not from the shell but from ttyd (and then screen), and so lost the environment. Still think this is a (very slightly) useful change as it also would happen when running on e.g. Raspbian.

[sfoster1]

Agreed, yeah.