Windows10下Anaconda环境中如何通过Boost.Python实现Python调用C/C++？

Absolutely, you can fully leverage Boost.Python from a conda-installed Boost in a Windows 10 + Anaconda environment to enable Python calls to C/C++ code. This setup is supported and used in real-world applications—your errors are caused by mismatched library naming, incorrect compiler flags, and misconfigured CMake settings, all of which are fixable.

Let’s break down the fixes for your issues: #

1. Fix manual compilation errors #

The key mistakes in your original commands were:

• Using the wrong Boost library name (conda’s Boost appends Python version numbers to library names, e.g., boost_python38 for Python 3.8, not boost_python)
• Using a Linux-style .so suffix (Windows Python extensions use .pyd for dynamic libraries)
• Missing critical compiler macros to link Boost correctly
• Incorrect library paths for Anaconda’s Python import library

Correct g++ compilation command:

g++ -std=c++11 -shared -DBOOST_PYTHON_DYN_LINK ^
-I C:\App\Anaconda3\include ^
-I C:\App\Anaconda3\Library\include ^
hello.cpp ^
-L C:\App\Anaconda3\Library\lib ^
-L C:\App\Anaconda3\libs ^
-lboost_python38 ^
-lpython38 ^
-o hello.pyd

• BOOST_PYTHON_DYN_LINK: Tells the compiler to link against the dynamic Boost.Python library (conda’s Boost is built as dynamic by default)
• -lboost_python38: Matches the Python 3.8-specific Boost library from conda
• -lpython38: Links against Anaconda’s Python 3.8 import library (found in libs/)
• .pyd: Windows-standard suffix for Python extension modules

2. Fix CMake configuration errors #

Your CMake setup failed due to outdated commands and incorrect Boost component naming. Here’s a corrected, modern CMakeLists.txt:

CMAKE_MINIMUM_REQUIRED(VERSION 3.18)
PROJECT(hello)

# Locate Anaconda's Python 3 development files
find_package(Python3 REQUIRED COMPONENTS Development)

# Locate conda-installed Boost, specifying the Python 3.8 component
find_package(Boost 1.73 REQUIRED COMPONENTS python38)

# Create the Python extension module
Python3_add_library(hello MODULE hello.cpp)

# Link required libraries
target_link_libraries(hello PRIVATE 
    Boost::python38 
    Python3::Python
)

# Copy test script to build directory
FILE(COPY hello.py DESTINATION .)

# Add test (optional)
ADD_TEST(NAME 01-HelloWorld COMMAND ${Python3_EXECUTABLE} hello.py)

Steps to run CMake:

1. Ensure you’re using conda’s installed CMake (run conda activate your environment first)
2. In your BUILD directory, run:

   cmake .. -G "MinGW Makefiles" -DCMAKE_CXX_COMPILER=g++
   mingw32-make
   

• -G "MinGW Makefiles": Tells CMake to generate files compatible with your g++ compiler
• Using conda’s CMake ensures it correctly finds your conda-installed Boost and Python

Critical Additional Notes #

• Use conda’s compiler for compatibility: If you still hit ABI mismatches, install conda’s MinGW compiler with conda install -c conda-forge m2w64-gcc—this ensures your compiler matches the one used to build conda’s Boost libraries.
• Verify Boost library presence: Check C:\App\Anaconda3\Library\lib for libboost_python38.dll.a (MinGW) or boost_python38.lib (MSVC). If missing, reinstall Boost from conda-forge (it has more complete Boost packages):

  conda install -c conda-forge boost-python
  

• Test in Anaconda Prompt: Always run your Python test script from the Anaconda Prompt—it ensures the correct library paths are loaded, avoiding missing DLL errors.

内容的提问来源于stack exchange，提问作者fghoussen