Skip to main content

Testing and debugging Windows builds locally

The first thing that you should know is that you can locally test Windows builds of your packages even if you don't own a Windows machine. Microsoft makes available free trial Windows virtual machines (VMs). If you are unfamiliar with VM systems or have trouble installing Microsoft's VMs, please use a general web search to explore — while these topics are beyond the scope of this documentation, there are ample discussions on them on the broader Internet.

To bootstrap a conda environment, consider miniforge.

Testing using wine​

Some degree of testing and debugging can also be performed without a Windows system, using wine. miniforge works correctly in the wine's cmd shell, and can be used to create and run Conda environments. In fact, sometimes Wine is able to provide more insightful error messages, for example:

wine: Call from 00006FFFFFF999EA to unimplemented function libblas.dll.cdotc_, aborting

It may be necessary to manipulate the WINEDEBUG variable to obtain more debugging logs.

Debugging DLL issues​

When debugging issues related to dynamically-linked libraries (DLLs) failing to load, the following tools can be helpful:

  • Windows Debugger can be used when debugging the dreaded "DLL load failed" errors. For example to debug a numpy import error:

    • "C:\Program Files (x86)\Windows Kits\10\Debuggers\x64\gflags.exe" -i python.exe +sls
    • "C:\Program Files (x86)\Windows Kits\10\Debuggers\x64\cdb.exe" -logo log.txt -g -G -o -xn av python -c "import numpy" The log file saved in log.txt will display information about where the DLLs were loaded from, which DLLs are missing and which symbols are missing from a DLL.

  • Dependency Walker can display a tree diagram of all dependent modules.

  • Dependencies is a more modern replacement for Dependency Walker, with both console and GUI interface. It also works better on Wine.

  • dumpbin tool from MSVC can be used to obtain information about Windows binaries. On Unix, gendef tool can be used instead by installing the conda package.