Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.43%. Comparing base (e1fe62f) to head (8cf652d).
Report is 639 commits behind head on master.

@@           Coverage Diff           @@
##           master   #15511   +/-   ##
=======================================
  Coverage   98.43%   98.43%           
=======================================
  Files         161      161           
  Lines       21281    21302   +21     
=======================================
+ Hits        20948    20969   +21     
  Misses        333      333           

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

Code size report:

   bare-arm:    +0 +0.000% 
minimal x86:    +0 +0.000% 
   unix x64:  +536 +0.064% standard[incl +64(data)]
      stm32:  +236 +0.060% PYBV10
     mimxrt:  +248 +0.068% TEENSY40
        rp2:  +240 +0.027% RPI_PICO_W
       samd:  +260 +0.098% ADAFRUIT_ITSYBITSY_M4_EXPRESS

Thanks for this, it looks like a pretty useful feature (I've never used it myself though in CPython).

Especially if MicroPython will never support full CPython metaclasses, this is a good (and more modern) approach to it.

I'm currently looking to see how to address the issue with the init_subclass recursion now given #15533.

I'm currently looking to see how to address the issue with the init_subclass recursion now given #15533.

Oh, that's interesting, that the fix in 15533 breaks this PR.

I think 15533 is necessary and should be merged. It's a very simple fix and fixes an edge case that's very difficult to notice in a running program, let alone debug and work around.

Oh, that's interesting, that the fix in 15533 breaks this PR.

No, mostly it just makes one of the behavioral compromises in the original version of this PR obsolete and unnecessary -- since it was possible to fix that fairly quickly, merging the original half-measure is unnecessary.

For that matter, it actually fixes cpydiff/core_class_initsubclass_recursive, so I'll be moving it back to tests -- though frustratingly, it wasn't quite enough to fix cpydiff/core_class_initsubclass_multi the way I'd originally hoped, even if I think I understand why now.

I think the fix for that will be tweaking to super_attr to special-case __init_subclass__ lookups to default to a no-op function instead of failing. (Alternatively, we could just add a default __init_subclass__ implementation to base object; I just think it's better to limit the scope of the performance impact and tweak super instead.)

For that matter, it actually fixes cpydiff/core_class_initsubclass_recursive, so I'll be moving it back to tests

Very good! Then merging #15533 is the right way to move forward.

though frustratingly, it wasn't quite enough to fix cpydiff/core_class_initsubclass_multi the way I'd originally hoped, even if I think I understand why now.

I think the fix for that will be tweaking to super_attr to special-case __init_subclass__ lookups to default to a no-op function instead of failing. (Alternatively, we could just add a default __init_subclass__ implementation to base object; I just think it's better to limit the scope of the performance impact and tweak super instead.)

Hmm, I'm not sure at this stage what the best option is... do what you think is best then we can review it.

Retracting in favor of further discussion and consensus about how to implement this part of PEP 487.