fix: reset Cores created by the GUI when it closes

[gselzer]

Proper device shutdown requires calling CMMCore.reset() before terminating the application. This solution ensures that any CMMCore created by the application is reset before the main GUI closes.

Note that callers passing a CMMCore instance to MicroManagerGUI or create_mmgui are responsible for cleaning it up (they may want to continue using that instance after the GUI closes)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 82.69%. Comparing base (3e40fed) to head (7601003).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #168      +/-   ##
==========================================
+ Coverage   81.87%   82.69%   +0.82%     
==========================================
  Files          29       29              
  Lines        1810     1815       +5     
==========================================
+ Hits         1482     1501      +19     
+ Misses        328      314      -14     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[tlambert03]

I'm ok with this... but it smells a little bit like the thing we just spent a long time debugging (see pymmcore-plus/pymmcore-plus#572 and related issues).

this one is probably safer because it's a higher level object cleaning up the core... but, I'll point out that the C++ destructor of CMMCore itself calls reset. So, it's not strictly necessary for the user to call reset() (it's not the case that proper device shutdown always demands this). It's definitely possible that you found a race condition where object cleanup didn't happen or something, but can you confirm what you noticed that made you feel like this was necessary?

[gselzer]

It's definitely possible that you found a race condition where object cleanup didn't happen or something, but can you confirm what you noticed that made you feel like this was necessary?

@marktsuchida and I were testing a new device adapter on one of our scopes and found that devices were not being shutdown (specifically that (a) behavior programmed into shutdown was not being observed and (b) that no log messages relating to unloading devices were present in the core log) when we pressed the close button. Then I was able to replicate this on my own machine with the demo config.

Yeah, I was a bit suprised too that the CMMCore destructor does call this, but it's possible that with GUI shutdown behavior the C++ destructor doesn't get called in time?

[tlambert03]

Arg. That's annoying :) I guess we'll just have to use belt and suspenders then