scoder

>What I have here works, but maybe feels like the wrong solution. I'll have another look. Yes, I was wondering the some thing on a quick look. It's to early...

https://github.com/cython/cython/issues/2747#issuecomment-1889945532 might be related as well. I always considered it hacky that we only remember the `cname` of `final` methods separately, instead of having a real `Entry` available that we...

We currently create a new scanner for each f-string expression and parse it as if it was inside of parentheses. CPython's parser now only creates a new parser context, stacking...

>`strip_string_literals` in Dependencies.py is a non obvious thing that'll need updating with this. It seems a case where a regex based scanner would make the parsing simpler and safer, not...

Might be a memory issue: ``` 3289s powerpc64le-linux-gnu-gcc: fatal error: Killed signal terminated program cc1 3289s compilation terminated. 3292s ERROR ``` It's a test that generates a fairly large C...

An issue with this is that the "score" is really just "a number". It's not exactly the number of C-API calls, for example. And the total number of Python (object)...

> most of the tests are run in `language_level=2` We should change that, BTW. Most projects, even with legacy code, will probably change to LL3 or have changed already. It...

In any case, we cannot just remove these bits from the tests and expect them to work. There's more to be done. See the long list of failures in https://github.com/cython/cython/pull/5839...

> I suspect it makes some non-loop cases marginally slower of course We could remember whether we're currently inside of a loop body and apply it only there. But that...

> temps created for indexing memoryview slice are > not cleaned up after they are finished with in normal use - > they are left until the end of the...