GitHub actions

[djetelina]

Hello, I'm trying to cover my application with tests using the snapshot extension, works great locally, but in GH Actions it's failing. I belive it's due to NoScreen: node has no screen error. Is there a workaround for it?

Failed run: https://github.com/djetelina/tofuref/actions/runs/15112170160/job/42473937234?pr=8

[willmcgugan]

I suspect the error is a red herring. There is no reason why GH should run any different that locally if you are using our test framework. I would suspect something environmental. You might want to compare with our test suite and dependencies.

[djetelina]

Switched from my apple sillicon to my linux machine and it seems like my error is coming from the fact that I'm treating widgets as singletons, importing them initialized in some cases. In the tests, the application runs aren't separated completely and something doesn't get cleared along the way:

tests/test_snapshots.py::test_welcome 
--------------------------------------------------------------------------------------------------------------------------------------------------- live log call ----------------------------------------------------------------------------------------------------------------------------------------------------
2025-05-19 21:32:43 [    INFO] Importing widgets (widgets.py:14)
2025-05-19 21:32:43 [    INFO] Importing widgets done (widgets.py:102)
2025-05-19 21:32:43 [    INFO] Composing UI (main.py:51)
2025-05-19 21:32:44 [    INFO] HTTP Request: GET https://api.opentofu.org/registry/docs/providers/index.json "HTTP/1.1 200 OK" (_client.py:1740)
PASSED                                                                                                                                                                                                                                                                                                         [  8%]
tests/test_snapshots.py::test_welcome_fullscreen 
--------------------------------------------------------------------------------------------------------------------------------------------------- live log call ----------------------------------------------------------------------------------------------------------------------------------------------------
2025-05-19 21:32:47 [    INFO] Composing UI (main.py:51)
FAILED       

Although what's odd is that in the Actions, even the first test fails. That might have more to do with some unhandled IO issue, but I guess that'll be my next step, if I don't give up on tests again :D

I guess I have some refactoring to do. Shame, the singleton approach is really handy and doesn't do anything wrong in normal app runs from what I can tell.

edit: small update, the tests fail at "random" points of my code, meaning that something else that holds the asyncio loop during that time is causing the crashes. I still think it's due to my singleton approach...

edit2: confirmed, singletons were the root of all evil