Feature

StopIteration.value is stored in the instance __dict__ instead of as a member (struct field) like CPython. This makes StopIteration().__dict__ non-empty and causes value to return the wrong result when the dict is cleared.

Same class of issue as SystemExit.code (#8230, fixed in #8282). Other built-in exceptions share it (AttributeError, ImportError, SyntaxError, UnicodeDecodeError); this issue targets StopIteration first.

Wrong value:

e = StopIteration(5)
e.__dict__.clear()
print(e.value)
# CPython 3.14: 5      (value is a member, unaffected by dict)
# RustPython:   None   (value lives in the dict, so it's gone)

Also observable:

type(StopIteration.value)   # CPython: member_descriptor   RustPython: NoneType
vars(StopIteration(5))      # CPython: {}                  RustPython: {'value': 5}

Cause: value is a class-level None set via set_attr into the instance dict in slot_init, instead of a struct field. The hot-path helper VirtualMachine::new_stop_iteration (crates/vm/src/vm/vm_new.rs) also allocates a fresh dict per call.

Proposed fix: Move value to a struct field on PyStopIteration (the #[repr(C)] + base-field approach used by PyOSError and now SystemExit in #8282), exposed via #[pygetset]. Since StopIteration is raised on a very hot path, this also removes the per-instance dict allocation.

I'll take this and open a PR, following the same approach as #8282.

Python Documentation or reference to CPython source code

https://docs.python.org/3/library/exceptions.html#StopIteration

value is defined as a PyMemberDef (a struct member, not an instance-dict entry) in CPython's Objects/exceptions.c.

Reproduction verified against CPython 3.14.6 and RustPython locally. Issue drafted with AI assistance (Claude), reviewed and edited by me.