为何我在简单ctypes示例中遇到Access Violation错误？

Let's break down why you're hitting that access violation and how to fix it quickly.

The Root Cause #

The problem boils down to missing type annotations for your AComplicatedCPPObj_new function in Python. By default, ctypes assumes every function returns a 32-bit c_int—but on 64-bit systems, pointers are 8 bytes wide. When your C wrapper returns a pointer to the C++ object, ctypes truncates that 8-byte address to 4 bytes, leaving you with an invalid, garbage pointer stored in self.obj. When you try to use that broken pointer in getA, it tries to access a non-existent memory address, triggering the access violation.

The Fix #

You need to explicitly define argtypes and restype for all C functions you call via ctypes. Here's the corrected Python code:

import ctypes as c
import os, glob, sys

WORKING_DIRECTORY = os.path.dirname(__file__)
if sys.platform == "linux":
    so_path = os.path.join(WORKING_DIRECTORY, "INSTALL/lib/libctypesTest.so")
else:
    so_path = os.path.join(WORKING_DIRECTORY, "INSTALL/bin/ctypesTest.dll")
if not os.path.isfile(so_path):
    raise FileNotFoundError(so_path)
lib = c.CDLL(so_path)

# Set up types for all exported functions ONCE, outside the class
lib.AComplicatedCPPObj_new.argtypes = [c.c_int]
lib.AComplicatedCPPObj_new.restype = c.c_void_p

lib.AComplicatedCPPObj_getA.argtypes = [c.c_void_p]
lib.AComplicatedCPPObj_getA.restype = c.c_int

class AComplicatedCPPObj:
    def __init__(self, a):
        self.obj = lib.AComplicatedCPPObj_new(a)
    def getA(self):
        return lib.AComplicatedCPPObj_getA(self.obj)

if __name__ == '__main__':
    o = AComplicatedCPPObj(4)
    print(o)
    print(o.getA())  # This should now output 4 without errors

Key Changes Explained #

1. Added type info for AComplicatedCPPObj_new:
   • argtypes = [c.c_int] tells ctypes the function expects an integer input (matching your C wrapper's parameter).
   • restype = c.c_void_p specifies the function returns a void pointer (which aligns with the C++ object pointer returned by the wrapper).
2. Moved type setup outside __init__: You only need to configure these types once when loading the library, not every time you create an instance of the Python class.

Extra Tips to Avoid Future Issues #

• Always define argtypes and restype for every ctypes function you use. This catches type mismatches early and prevents undefined behavior like access violations.
• On Windows, ensure your exported functions use the correct calling convention: your extern "C" wrapper uses cdecl, which matches ctypes.CDLL (if you used stdcall, you'd need ctypes.WinDLL instead).

内容的提问来源于stack exchange，提问作者CiaranWelsh