Crash when overriding field as property/method on frozen derived class

[jjlee]

I see this on python/mypy master (41b40aa) and Python 3.7.5.

Following on from #7535 (fixed recently by @msullivan), this triggers another crash (really just adding the frozen=True from my example on that bug report, though I also dropped the args to field since I realised I could):

from dataclasses import dataclass, field

@dataclass
class Base:
    foo: bool = field()

@dataclass(frozen=True)
class Derived(Base):
    @property
    def foo(self):
        pass

Traceback (seems slightly different from #6715 -- the assertion here is in method _freeze, not method collect_attributes):

$ mypy --show-traceback /home/me/minimal3.py
/home/me/minimal3.py:10: error: INTERNAL ERROR -- Please try using mypy master on Github:
https://mypy.rtfd.io/en/latest/common_issues.html#using-a-development-mypy-build
Please report a bug at https://github.com/python/mypy/issues
version: 0.770+dev.41b40aa4fc7cdeb32f5ee4fec97f5379ebbe795d
Traceback (most recent call last):
  File "/home/me/env/3/moving-target/bin/mypy", line 11, in <module>
    load_entry_point('mypy', 'console_scripts', 'mypy')()
  File "/home/me/dev/mypy/mypy/__main__.py", line 8, in console_entry
    main(None, sys.stdout, sys.stderr)
  File "/home/me/dev/mypy/mypy/main.py", line 89, in main
    res = build.build(sources, options, None, flush_errors, fscache, stdout, stderr)
  File "/home/me/dev/mypy/mypy/build.py", line 166, in build
    sources, options, alt_lib_path, flush_errors, fscache, stdout, stderr, extra_plugins
  File "/home/me/dev/mypy/mypy/build.py", line 234, in _build
    graph = dispatch(sources, manager, stdout)
  File "/home/me/dev/mypy/mypy/build.py", line 2631, in dispatch
    process_graph(graph, manager)
  File "/home/me/dev/mypy/mypy/build.py", line 2938, in process_graph
    process_stale_scc(graph, scc, manager)
  File "/home/me/dev/mypy/mypy/build.py", line 3031, in process_stale_scc
    mypy.semanal_main.semantic_analysis_for_scc(graph, scc, manager.errors)
  File "/home/me/dev/mypy/mypy/semanal_main.py", line 77, in semantic_analysis_for_scc
    process_top_levels(graph, scc, patches)
  File "/home/me/dev/mypy/mypy/semanal_main.py", line 202, in process_top_levels
    patches)
  File "/home/me/dev/mypy/mypy/semanal_main.py", line 330, in semantic_analyze_target
    active_type=active_type)
  File "/home/me/dev/mypy/mypy/semanal.py", line 363, in refresh_partial
    self.refresh_top_level(node)
  File "/home/me/dev/mypy/mypy/semanal.py", line 374, in refresh_top_level
    self.accept(d)
  File "/home/me/dev/mypy/mypy/semanal.py", line 4682, in accept
    node.accept(self)
  File "/home/me/dev/mypy/mypy/nodes.py", line 939, in accept
    return visitor.visit_class_def(self)
  File "/home/me/dev/mypy/mypy/semanal.py", line 1014, in visit_class_def
    self.analyze_class(defn)
  File "/home/me/dev/mypy/mypy/semanal.py", line 1091, in analyze_class
    self.analyze_class_body_common(defn)
  File "/home/me/dev/mypy/mypy/semanal.py", line 1100, in analyze_class_body_common
    self.apply_class_plugin_hooks(defn)
  File "/home/me/dev/mypy/mypy/semanal.py", line 1146, in apply_class_plugin_hooks
    hook(ClassDefContext(defn, decorator, self))
  File "/home/me/dev/mypy/mypy/plugins/dataclasses.py", line 369, in dataclass_class_maker_callback
    transformer.transform()
  File "/home/me/dev/mypy/mypy/plugins/dataclasses.py", line 182, in transform
    self._freeze(attributes)
  File "/home/me/dev/mypy/mypy/plugins/dataclasses.py", line 355, in _freeze
    assert isinstance(var, Var)
AssertionError: 
/home/me/minimal3.py:10: : note: use --pdb to drop into pdb

[anis-campos]

Is there any workaround that would make this less of a crash but more of a warning ? warning that could therefore be ignore just like the others ones ?

It's so sad no being able to properly override inherited properties with @property, it's becomes very difficult to impose immutability of the dataclasses without some cumbersome workaround via __post_init__ and object.setattr

[JukkaL]

In #8976 there was a similar-looking crash when overriding a frozen attribute with a method:

from dataclasses import dataclass


@dataclass(frozen=True)
class Foo:
    attr: int


@dataclass(frozen=True)
class Bar(Foo):
    def attr(self) -> int:
        return 1

[anis-campos]

Isn't this related to #13475 ? Well I guess not completely, as for a frozen dataclass, a read-only property should be enough