This PR fixes a crash in pylint when processing dataclasses that use __init_subclass__ to modify the __init__ method. The issue occurred when inheriting from a generic dataclass without providing explicit type parameters.

Problem

When __init_subclass__ modifies the __init__ method by assignment (e.g., cls.__init__ = wrapped_init), astroid's dataclass brain module would crash with an AttributeError because it expected __init__ to always be a FunctionDef node, but after reassignment it becomes an AssignAttr node.

from dataclasses import dataclass
from typing import ParamSpec

@dataclass
class Foo[T]:
    def __init_subclass__(cls) -> None:
        cls.__init__ = wrap(cls.__init__)  # This causes the crash

@dataclass
class Bar(Foo): ...  # pylint crashes here

Solution

Added a monkey patch in pylint/checkers/dataclass_checker.py that replaces astroid's _find_arguments_from_base_classes function with a version that safely handles non-FunctionDef nodes. When __init__ has been modified by __init_subclass__, the patched function simply skips processing that base class rather than crashing.

The fix is minimal and surgical:

• Only affects the specific case where __init_subclass__ modifies __init__
• Preserves all existing dataclass functionality
• Gracefully handles edge cases like multiple inheritance and various assignment patterns

Testing

Added a comprehensive functional test (dataclass_generic_init_subclass.py) that verifies the fix works for both generic and non-generic dataclass inheritance scenarios. All existing dataclass and regression tests continue to pass.

Fixes #10519.

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.