树莓派5上Pyfmi无法加载示例FMU二进制文件：文件不存在错误

问题描述 #

在树莓派5（运行Raspberry Pi OS）上尝试加载Plecs生成的FMU进行仿真，该功能在Windows桌面环境正常运行，但树莓派上无法加载模型。测试加载PyFMI自带的弹跳球示例FMU时，出现相同错误。

报错信息 #

~ $ python
Python 3.9.18 | packaged by conda-forge | (main, Dec 23 2023, 17:20:25) 
[GCC 12.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from pyfmi import load_fmu
>>> load_fmu("/home/baberg/miniconda3/pkgs/pyfmi-2.13.0-py39hf393ea8_3/lib/python3.9/site-packages/pyfmi/examples/files/FMUs/ME2.0/bouncingBall.fmu")
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "src/pyfmi/fmi.pyx", line 8683, in pyfmi.fmi.load_fmu
  File "src/pyfmi/fmi.pyx", line 7613, in pyfmi.fmi.FMUModelME2.__init__
  File "src/pyfmi/fmi.pyx", line 4135, in pyfmi.fmi.FMUModelBase2.__init__
pyfmi.fmi.InvalidBinaryException: The FMU could not be loaded. Error loading the binary. Could not load the FMU binary: /tmp/baberg/JModelica.org/jm_tmps2vkgivo/binaries/linux64/bouncingBall.so: cannot open shared object file: No such file or directory
>>> 

注：已确认上述报错中提到的.so文件实际存在。

源码追踪 #

追踪PyFMI源码后发现，InvalidBinaryException由status变量返回错误触发，相关代码片段如下：

#Connect the DLL
if _connect_dll:
    self._log_handler.capi_start_callback(self._max_log_size_msg_sent, self._current_log_size)
    status = FMIL2.fmi2_import_create_dllfmu(self._fmu, self._fmu_kind, &self.callBackFunctions)
    self._log_handler.capi_end_callback(self._max_log_size_msg_sent, self._current_log_size)
    if status == FMIL.jm_status_error:
        last_error = pyfmi_util.decode(FMIL2.fmi2_import_get_last_error(self._fmu))
        if enable_logging:
            raise InvalidBinaryException("The FMU could not be loaded. Error loading the binary. " + last_error)
        else:
            raise InvalidBinaryException("The FMU could not be loaded. Error loading the binary. Enable logging for possibly more information.")
    self._allocated_dll = 1

环境信息 #

• Python 3.9.18
• PyFMI 2.13.0
• FMILib 2.4.1
• Cython 3.0.12
• Assimulo 3.7.0
• Conda 25.3.1

排查思路 #

• 检查架构兼容性：树莓派5为ARM64架构，conda-forge提供的PyFMI包可能针对x86_64编译，导致.so文件无法在ARM环境加载。执行file /tmp/baberg/JModelica.org/jm_tmps2vkgivo/binaries/linux64/bouncingBall.so查看文件架构信息。
• 检查动态库依赖：使用ldd /tmp/baberg/JModelica.org/jm_tmps2vkgivo/binaries/linux64/bouncingBall.so命令，查看.so文件的动态依赖是否缺失或版本不匹配，比如GLIBC等系统库的版本是否满足要求。
• 启用详细日志：加载FMU时添加log_level=7参数，即load_fmu("path/to/your.fmu", log_level=7)，获取FMILib加载过程中的详细日志，定位具体失败原因。
• 验证conda包架构：确认conda安装的PyFMI和FMILib是否为ARM64版本。可执行conda list pyfmi查看包的架构信息，若为x86_64，尝试重新安装指定ARM架构的包：conda install -c conda-forge pyfmi --force-reinstall。
• 尝试系统级安装或源码编译：若conda包存在架构问题，可尝试使用Raspberry Pi OS的包管理器安装依赖（如apt install python3-pyfmi，若仓库存在），或从源码编译PyFMI和FMILib，确保编译过程使用树莓派本地的编译器和系统依赖。

内容的提问来源于stack exchange，提问作者user420383