主页 / coding / 问题 / 78933467
Accepted
vmonteco
Asked: 2024-08-31 03:20:11 +0800 CST

如何从 Python 访问传递给 C 函数并由 C 函数返回的指针值？

  • 772

我的 python 代码可以访问通过调用的 C 函数接收和返回的实际指针值ctypes吗？

如果是，我该如何实现？

我想测试传递给共享库函数并从共享库函数返回的指针值，以使用 pytest 测试分配（这里，测试它strdup没有返回相同的指针而是返回指向不同地址的新指针）。

我把要实现的函数之一（strdup）包装在一个名为的文件中的一个新 C 函数中，wrapped_strdup.c以显示指针值和内存区域的内容：

/*
** I'm compiling this into a .so the following way:
**   - gcc -o wrapped_strdup.o -c wrapped_strdup.c
**   - ar rc wrapped_strdup.a wrapped_strdup.o
**   - ranlib wrapped_strdup.a
**   - gcc -shared -o wrapped_strdup.so -Wl,--whole-archive wrapped_strdup.a -Wl,--no-whole-archive
*/
#include <stdio.h>
#include <stdlib.h>
#include <string.h>

char *wrapped_strdup(char *src){
    char *dst;

    printf("From C:\n");
    printf("- src address: %X, src content: [%s].\n", src, src);
    dst = strdup(src);
    printf("- dst address: %X, dst content: [%s].\n", dst, dst);
    return dst;
}

我还在同一目录中创建了一个名为的 pytest 测试文件test_strdup.py：

#!/usr/bin/env python3

import ctypes
import pytest

# Setting wrapped_strdup:
lib_wrapped_strdup = ctypes.cdll.LoadLibrary("./wrapped_strdup.so")
wrapped_strdup = lib_wrapped_strdup.wrapped_strdup
wrapped_strdup.restype = ctypes.c_char_p
wrapped_strdup.argtypes = [ctypes.c_char_p]

@pytest.mark.parametrize("src", [b"", b"foo"])
def test_strdup(src: bytes):
    print("")
    dst = wrapped_strdup(src)

    print("From Python:")
    print(f"- src address: {hex(id(src))}, src content: [{src!r}].")
    print(f"- dst address: {hex(id(dst))}, dst content: [{dst!r}].")
    
    assert src == dst
    assert hex(id(src)) != hex(id(dst))

然后，运行测试得到以下输出：

$ pytest test_strdup.py --maxfail=2 -v -s
=================================== test session starts ====================================
platform linux -- Python 3.12.5, pytest-8.3.2, pluggy-1.5.0 -- /usr/bin/python
cachedir: .pytest_cache
rootdir: /home/vmonteco/code/MREs/MRe_strdup_test_with_ctypes
plugins: anyio-4.4.0, cov-5.0.0, typeguard-4.3.0
collected 2 items                                                                          

test_strdup.py::test_strdup[] 
From C:
- src address: C19BDBE8, src content: [].
- dst address: 5977DFA0, dst content: [].
From Python:
- src address: 0x75bcc19bdbc8, src content: [b''].
- dst address: 0x75bcc19bdbc8, dst content: [b''].
FAILED
test_strdup.py::test_strdup[foo] 
From C:
- src address: BF00A990, src content: [foo].
- dst address: 59791030, dst content: [foo].
From Python:
- src address: 0x75bcbf00a970, src content: [b'foo'].
- dst address: 0x75bcbefc18f0, dst content: [b'foo'].
PASSED

========================================= FAILURES =========================================
______________________________________ test_strdup[] _______________________________________

src = b''

    @pytest.mark.parametrize("src", [b"", b"foo"])
    def test_strdup(src: bytes):
        print("")
        dst = wrapped_strdup(src)
    
        print("From Python:")
        print(f"- src address: {hex(id(src))}, src content: [{src!r}].")
        print(f"- dst address: {hex(id(dst))}, dst content: [{dst!r}].")
    
        assert src == dst
>       assert hex(id(src)) != hex(id(dst))
E       AssertionError: assert '0x75bcc19bdbc8' != '0x75bcc19bdbc8'
E        +  where '0x75bcc19bdbc8' = hex(129453562518472)
E        +    where 129453562518472 = id(b'')
E        +  and   '0x75bcc19bdbc8' = hex(129453562518472)
E        +    where 129453562518472 = id(b'')

test_strdup.py:22: AssertionError
================================= short test summary info ==================================
FAILED test_strdup.py::test_strdup[] - AssertionError: assert '0x75bcc19bdbc8' != '0x75bcc19bdbc8'
=============================== 1 failed, 1 passed in 0.04s ================================

此输出显示了两件事：

  • b''尽管从底层角度看地址不同，但 Python 中引用的变量的地址无论如何都是相同的（即同一个对象）。这与一些纯 Python 测试一致，我猜这可能是某种优化功能。
  • C 和 Python 中的地址值dst和src变量实际上似乎没有关联。

因此，上述尝试实际上不可靠，无法检查函数是否返回指向不同区域的指针。

我还可以尝试检索指针值本身，并通过更改属性进行第二次测试运行以专门检查此部分restype：

#!/usr/bin/env python3                                                                      

import ctypes
import pytest

# Setting wrapped_strdup:                                                                   
lib_wrapped_strdup = ctypes.cdll.LoadLibrary("./wrapped_strdup.so")
wrapped_strdup = lib_wrapped_strdup.wrapped_strdup
wrapped_strdup.restype = ctypes.c_void_p   # Note that it's not a c_char_p anymore.
wrapped_strdup.argtypes = [ctypes.c_char_p]

@pytest.mark.parametrize("src", [b"", b"foo"])
def test_strdup_for_pointers(src: bytes):
    print("")
    dst = wrapped_strdup(src)

    print("From Python:")
    print(f"- retrieved dst address: {hex(dst)}.")

以上代码给出以下输出：

$ pytest test_strdup_for_pointers.py --maxfail=2 -v -s
=================================== test session starts ====================================
platform linux -- Python 3.12.5, pytest-8.3.2, pluggy-1.5.0 -- /usr/bin/python
cachedir: .pytest_cache
rootdir: /home/vmonteco/code/MREs/MRe_strdup_test_with_ctypes
plugins: anyio-4.4.0, cov-5.0.0, typeguard-4.3.0
collected 2 items                                                                          

test_strdup_for_pointers.py::test_strdup_for_pointers[] 
From C:
- src address: E15BDBE8, src content: [].
- dst address: 84D4D820, dst content: [].
From Python:
- retrieved dst address: 0x608984d4d820.
PASSED
test_strdup_for_pointers.py::test_strdup_for_pointers[foo] 
From C:
- src address: DEC7EA80, src content: [foo].
- dst address: 84EA7C40, dst content: [foo].
From Python:
- retrieved dst address: 0x608984ea7c40.
PASSED

==================================== 2 passed in 0.01s =====================================

这将提供实际的地址（或至少是看起来相关的信息）。

但如果不知道 C 函数接收的值，它就没有多大帮助。

附录：我从马克的回答中得出的结论（并且有效）：

这是一个测试，它实现了所接受的答案中建议的解决方案：

#!/usr/bin/env python3

import ctypes
import pytest

# Setting libc:
libc = ctypes.cdll.LoadLibrary("libc.so.6")
strlen = libc.strlen
strlen.restype = ctypes.c_size_t
strlen.argtypes = (ctypes.c_char_p,)

# Setting wrapped_strdup:
lib_wrapped_strdup = ctypes.cdll.LoadLibrary("./wrapped_strdup.so")
wrapped_strdup = lib_wrapped_strdup.wrapped_strdup
# Restype will be set directly in the tests.
wrapped_strdup.argtypes = (ctypes.c_char_p,)


@pytest.mark.parametrize("src", [b"", b"foo"])
def test_strdup(src: bytes):
    print("")  # Just to make pytest output more readable.

    # Set expected result type.
    wrapped_strdup.restype = ctypes.POINTER(ctypes.c_char)

    # Create the src buffer and retrieve its address.
    src_buffer = ctypes.create_string_buffer(src)
    src_addr = ctypes.addressof(src_buffer)
    src_content = src_buffer[:strlen(src_buffer)]

    # Run function to test.
    dst = wrapped_strdup(src_buffer)

    # Retrieve result address and content.
    dst_addr = ctypes.addressof(dst.contents)
    dst_content = dst[: strlen(dst)]

    # Assertions.
    assert src_content == dst_content
    assert src_addr != dst_addr

    # Output.
    print("From Python:")
    print(f"- Src content: {src_content!r}. Src address: {src_addr:X}.")
    print(f"- Dst content: {dst_content!r}. Dst address: {dst_addr:X}.")


@pytest.mark.parametrize("src", [b"", b"foo"])
def test_strdup_alternative(src: bytes):
    print("")  # Just to make pytest output more readable.

    # Set expected result type.
    wrapped_strdup.restype = ctypes.c_void_p

    # Create the src buffer and retrieve its address.
    src_buffer = ctypes.create_string_buffer(src)
    src_addr = ctypes.addressof(src_buffer)
    src_content = src_buffer[:strlen(src_buffer)]

    # Run function to test.
    dst = wrapped_strdup(src_buffer)

    # Retrieve result address and content.
    dst_addr = dst
    # cast dst:
    dst_pointer = ctypes.cast(dst, ctypes.POINTER(ctypes.c_char))
    dst_content = dst_pointer[:strlen(dst_pointer)]

    # Assertions.
    assert src_content == dst_content
    assert src_addr != dst_addr

    # Output.
    print("From Python:")
    print(f"- Src content: {src_content!r}. Src address: {src_addr:X}.")
    print(f"- Dst content: {dst_content!r}. Dst address: {dst_addr:X}.")

输出 ：

$ pytest test_strdup.py -v -s            
=============================== test session starts ===============================
platform linux -- Python 3.10.14, pytest-8.3.2, pluggy-1.5.0 -- /home/vmonteco/.pyenv/versions/3.10.14/envs/strduo_test/bin/python3.10
cachedir: .pytest_cache
rootdir: /home/vmonteco/code/MREs/MRe_strdup_test_with_ctypes
plugins: anyio-4.4.0, stub-1.1.0
collected 4 items                                                                 

test_strdup.py::test_strdup[] 
From C:
- src address: 661BBE90, src content: [].
- dst address: F5D8A7A0, dst content: [].
From Python:
- Src content: b''. Src address: 7C39661BBE90.
- Dst content: b''. Dst address: 57B4F5D8A7A0.
PASSED
test_strdup.py::test_strdup[foo] 
From C:
- src address: 661BBE90, src content: [foo].
- dst address: F5E03340, dst content: [foo].
From Python:
- Src content: b'foo'. Src address: 7C39661BBE90.
- Dst content: b'foo'. Dst address: 57B4F5E03340.
PASSED
test_strdup.py::test_strdup_alternative[] 
From C:
- src address: 661BBE90, src content: [].
- dst address: F5B0AC50, dst content: [].
From Python:
- Src content: b''. Src address: 7C39661BBE90.
- Dst content: b''. Dst address: 57B4F5B0AC50.
PASSED
test_strdup.py::test_strdup_alternative[foo] 
From C:
- src address: 661BBE90, src content: [foo].
- dst address: F5BF9C20, dst content: [foo].
From Python:
- Src content: b'foo'. Src address: 7C39661BBE90.
- Dst content: b'foo'. Dst address: 57B4F5BF9C20.
PASSED

================================ 4 passed in 0.01s ================================
python

1 个回答

  1. Best Answer

    返回类型为ctypes.c_char_p“有用”，并将返回值转换为 Python 字符串，丢失实际的 C 指针。用于ctypes.POINTER(ctypes.c_char)保留指针。

    返回类型ctypes.c_void_p也很“有用”，它将返回的 C 地址转换为 Python 整数，但可以转换为更具体的指针类型来访问地址处的数据

    要找到它的地址，请使用指针的内容ctypes.addressof；否则您将获得指针存储的地址。

    我使用它char* strcpy(char* dest, const char* src)作为示例，因为返回的指针与dest指针的地址相同，并且它显示 C 地址与 Python 中的地址相同，而不需要 C 辅助函数。

    在下面的代码中，可变字符串缓冲区dest具有与返回值相同的地址，并显示了几种检查返回值的 C 地址的方法：

    import ctypes as ct

dll = ct.CDLL('msvcrt')

dll.strcpy.argtypes = ct.c_char_p, ct.c_char_p
dll.strcpy.restype = ct.POINTER(ct.c_char)  # NOT ct.c_char_p to keep pointer
dll.strlen.argtypes = ct.c_char_p,
dll.strlen.restype = ct.c_size_t

dest = ct.create_string_buffer(10)  # writable char buffer
print(f'{ct.addressof(dest) = :#x}')  # its C address
result = dll.strcpy(dest, b'abcdefg')

# Note that for strcpy, returned address is the same as dest address
print(f'{ct.addressof(dest) = :#x}')  # dest array's C address
print(f'{ct.addressof(result.contents) = :#x}')  # result pointer's C address (same as dest)
n = dll.strlen(result)
print(f'{result[:n] = }')  # must slice char* or only prints one character.
print(f'{dest.value = }')  # array has .value (nul-termination) or .raw

dll.strcpy.restype = ct.c_void_p  # alternative, get the pointer address as Python int
result = dll.strcpy(dest, b'abcdefg')
print(f'{result = :#x}')  # same C address as above
p = ct.cast(result, ct.POINTER(ct.c_char))  # Cast afterward
print(f'{ct.addressof(p.contents) = :#x}')  # same C address
n = dll.strlen(p)
print(f'{p[:n] = }')  # must slice char*

p = ct.cast(result, ct.POINTER(ct.c_char * n))  # alternate, pointer to sized array
print(f'{p.contents.value = }')  # don't have to slice, (char*)[n] has known size n

    输出：

    ct.addressof(dest) = 0x1abe80ea398
ct.addressof(dest) = 0x1abe80ea398
ct.addressof(result.contents) = 0x1abe80ea398
result[:n] = b'abcdefg'
dest.value = b'abcdefg'
result = 0x1abe80ea398
ct.addressof(p.contents) = 0x1abe80ea398
p[:n] = b'abcdefg'
p.contents.value = b'abcdefg'
    • 4

相关问题