OS X 12.4+XCode13.4编译AI-Feynman报!tapi-tbd错误解决方案

• 运行环境：OS X 12.4 操作系统、XCode 13.4 开发环境
• 编译目标：AI-Feynman 开源项目
• 失败阶段：Fortran扩展模块链接阶段
• 核心报错：链接器ld无法识别x86_64架构下YAML格式tbd文件中的!tapi-tbd类型，编译流程终止

编译阶段执行的命令及完整报错如下：

/Users/davidlaxer/anaconda3/envs/AI-Feynman/bin/x86_64-apple-darwin13.4.0-gfortran -Wall -g -arch x86_64 -Wall -g -undefined dynamic_lookup -bundle build/temp.macosx-10.9-x86_64-3.9/build/src.macosx-10.9-x86_64-3.9/aifeynman/_symbolic_regress1module.o build/temp.macosx-10.9-x86_64-3.9/build/src.macosx-10.9-x86_64-3.9/build/src.macosx-10.9-x86_64-3.9/aifeynman/fortranobject.o build/temp.macosx-10.9-x86_64-3.9/aifeynman/symbolic_regress1.o build/temp.macosx-10.9-x86_64-3.9/build/src.macosx-10.9-x86_64-3.9/aifeynman/_symbolic_regress1-f2pywrappers.o -L/Users/davidlaxer/anaconda3/envs/AI-Feynman/bin/../lib/gcc/x86_64-apple-darwin11.4.2/4.8.5 -L/Users/davidlaxer/anaconda3/envs/AI-Feynman/bin/../lib/gcc/x86_64-apple-darwin11.4.2/4.8.5/../../.. -L/Users/davidlaxer/anaconda3/envs/AI-Feynman/bin/../lib/gcc/x86_64-apple-darwin11.4.2/4.8.5/../../.. -lgfortran -o aifeynman/_symbolic_regress1.cpython-39-darwin.so
ld: unsupported tapi file type '!tapi-tbd' in YAML file '/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/usr/lib/libSystem.tbd' for architecture x86_64
collect2: error: ld returned 1 exit status
error: Command "/Users/davidlaxer/anaconda3/envs/AI-Feynman/bin/x86_64-apple-darwin13.4.0-gfortran -Wall -g -arch x86_64 -Wall -g -undefined dynamic_lookup -bundle build/temp.macosx-10.9-x86_64-3.9/build/src.macosx-10.9-x86_64-3.9/aifeynman/_symbolic_regress1module.o build/temp.macosx-10.9-x86_64-3.9/build/src.macosx-10.9-x86_64-3.9/build/src.macosx-10.9-x86_64-3.9/aifeynman/fortranobject.o build/temp.macosx-10.9-x86_64-3.9/aifeynman/symbolic_regress1.o build/temp.macosx-10.9-x86_64-3.9/build/src.macosx-10.9-x86_64-3.9/aifeynman/_symbolic_regress1-f2pywrappers.o -L/Users/davidlaxer/anaconda3/envs/AI-Feynman/bin/../lib/gcc/x86_64-apple-darwin11.4.2/4.8.5 -L/Users/davidlaxer/anaconda3/envs/AI-Feynman/bin/../lib/gcc/x86_64-apple-darwin11.4.2/4.8.5/../../.. -L/Users/davidlaxer/anaconda3/envs/AI-Feynman/bin/../lib/gcc/x86_64-apple-darwin11.4.2/4.8.5/../../.. -lgfortran -o aifeynman/_symbolic_regress1.cpython-39-darwin.so" failed with exit status 1

Anaconda虚拟环境内预装的gfortran 4.8.5是适配2013年OS X 10.9系统的老旧版本，配套的ld链接器、tapi库解析工具版本过低，无法识别XCode 13.4自带MacOS SDK中采用新格式的.tbd动态库存根文件，最终触发解析错误。

按顺序执行以下操作即可修复编译问题：

1. 进入对应AI-Feynman的conda虚拟环境，替换老旧的编译工具链为适配新系统的版本，执行命令：

   conda activate AI-Feynman
   conda install -c conda-forge gfortran cctools ld64 -y
   

2. 配置编译环境变量，强制编译时使用系统原生的新版链接器、指定正确的SDK路径，执行命令：

   export CC=/usr/bin/clang
   export CXX=/usr/bin/clang++
   export LD=/usr/bin/ld
   export SDKROOT=$(xcrun --sdk macosx --show-sdk-path)
   export LDFLAGS="-Wl,-syslibroot,${SDKROOT}"
   

3. 清理之前编译失败产生的临时缓存文件，重新执行编译安装：

   rm -rf build/
   python setup.py install
   

若执行完上述步骤仍有报错，可直接通过Homebrew安装新版gcc套件，编译前将PATH环境变量中Homebrew的bin目录优先级调至conda路径之前，优先调用brew安装的新版gfortran即可。

内容的提问来源于stack exchange，提问作者dbl001