feat(test): Failure test case for mypyc dictionary subclass views

[BobTheBuidler]

I've created a failing test case for a mypyc bug with dict views as seen here: https://github.com/BobTheBuidler/checksum_dict/actions/runs/14843222374/job/41670896987

I added some new (working) test cases:

• dict
  • keys
  • values
  • items
• dict subclass
  • keys
  • values
  • items

and the failure cases:

• dict subclass with 2 bases
  • keys
  • values
  • items

[BobTheBuidler]

@JukkaL Would you happen to know what I need to change so the tests are able to find the DefaultDict object in the typing module? Or is there someone better to ask about this?

[BobTheBuidler]

Fascinating, I've managed to make it work using a type: ignore comment, and now we have our successful (wut?) failure case!

[JukkaL]

It's unclear what's the best approach here. @mypyc_attr(native_class=False) could act as a workaround (with some performance cost). We could also declare this as an unsupported use case, and generate an error with a note suggesting the @mypyc_attr workaround, in case it works.

[BobTheBuidler]

Hmm. That could work but I'd first like to try a bit to resolve it, I just need some help to know where I should look for the relevant code. Is there not a strightforward way to tell mypyc to only use the dict.keys c function when obj.keys actually resolves to dict.keys, and fallback to python's name resolution when it does not?