♻️ Remove overloaded loops in tests

[sanderegg]

What do these changes do?

In pytest the loop fixture is function scoped. The idea behind is to have a clean Event Loop for each test.

For different reasons, we used to overload that loop with a module scoped, or session scoped loop. This leads to weird behaviours in some tests that became a hurdle and a pain. Sometimes the loop gets replaced and the typical error message of "a future is waiting in a different loop than the current loop" would appear.

Rationale:

• we are using pytest-aiohttp for pytest
• this creates a loop fixture that is function scoped
• when we use async fixtures this gets detected automatically by pytest-aiohttp and then auto-use the loop fixture

--> To simplify and start from a clean slate, all the module/session scoped loops were removed.

Also some of the autouse fixtures are removed. PLEASE DO NOT USE THAT FEATURE, it is not explicit and is painful when debugging.

Hint: with pytest one casn use pytest --setup-show, this shows how the fixtures are started

Related issue/s

How to test

Checklist

[codecov]

Codecov Report

Merging #2674 (a059446) into master (670cb22) will decrease coverage by 2.8%.
The diff coverage is 80.0%.

@@           Coverage Diff            @@
##           master   #2674     +/-   ##
========================================
- Coverage    78.0%   75.1%   -2.9%     
========================================
  Files         636     636             
  Lines       26064   26073      +9     
  Branches     2524    2525      +1     
========================================
- Hits        20342   19602    -740     
- Misses       5053    5847    +794     
+ Partials      669     624     -45     

Flag	Coverage Δ
integrationtests	82.0% <60.0%> (+15.3%)	⬆️
unittests	73.6% <80.0%> (-0.1%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
...rector_v2/modules/comp_scheduler/dask_scheduler.py	91.8% <ø> (ø)
...-sidecar/src/simcore_service_dask_sidecar/tasks.py	81.2% <70.0%> (-3.0%)	⬇️
...ctor_v2/models/schemas/dynamic_services/service.py	100.0% <100.0%> (ø)
...r-v2/src/simcore_service_director_v2/utils/dask.py	93.5% <100.0%> (+<0.1%)	⬆️
...simcore_service_webserver/computation_subscribe.py	27.2% <0.0%> (-65.7%)	⬇️
...imcore_service_webserver/exporter/export_import.py	34.1% <0.0%> (-61.0%)	⬇️
...re-sdk/src/simcore_sdk/node_ports/serialization.py	25.0% <0.0%> (-57.4%)	⬇️
...vice_webserver/exporter/formatters/formatter_v2.py	31.9% <0.0%> (-56.4%)	⬇️
...vice_webserver/exporter/formatters/formatter_v1.py	24.5% <0.0%> (-54.2%)	⬇️
...c/simcore_service_webserver/users_to_groups_api.py	46.1% <0.0%> (-53.9%)	⬇️
... and 43 more

[GitHK]

Nice cleanup!

[GitHK]

Why was this changed to non async?

[pcrespov]

@GitHK because, as stated in the PR, we can only use async fixtures with a function scope, otherwise (i.e. module or session) they must be synchronous. This is a "reasonable" limitation provided the complex situations we found when mixing loops at different scopes :-)

[sanderegg]

read the PR description ;)
async def fixtures are auto-detected by pytest-aiohttp

[GitHK]

Nice, I've missed that part. I got it, no more module or session scoped async fixtures, only function wise async.

[pcrespov]

thx! greatly appreciated the changes in this PR

[pcrespov]

@GitHK because, as stated in the PR, we can only use async fixtures with a function scope, otherwise (i.e. module or session) they must be synchronous. This is a "reasonable" limitation provided the complex situations we found when mixing loops at different scopes :-)