CI Failures for "Windows 2022" builds

[speth]

Problem description

As noted in #1792, unrelated to the changes in that PR, we're currently getting CI failures for the "Windows 2022" builds. Looking at the logs, it appears that the "Build Cantera" step is completing successfully, but for the separate "Build Tests" step, SCons is re-running the configuration checks and trying to use a version of MinGW that's installed on the GitHub runner.

The "Build Cantera" step's log starts with:

Run scons build system_eigen=y system_yamlcpp=y system_highfive=y logging=debug toolchain=msvc f90_interface=n debug=n  --debug=time -j4
scons: Reading SConscript files ...
SCons 4.7.0 is using the following Python interpreter:
    C:\Users\runneradmin\micromamba\envs\test-env\python.exe (Python 3.12)
Compiling with MSVC version 14.3
Compiling with MSVC toolset 14.3 (default)
Compiling for architecture: amd64
Compiling using the following toolchain(s): ['default']
INFO: Compiling on 'AMD64 Family 25 Model 1 Stepping 1, AuthenticAMD'
INFO: Configuration variables read from 'cantera.conf' and command line:
    f90_interface = 'n'
    system_eigen = 'y'
    system_highfive = 'y'
    system_yamlcpp = 'y'
    debug = False
    logging = 'debug'

But the "Build Tests" step starts with:

Run scons -j4 build-tests --debug=time
scons: Reading SConscript files ...
SCons 4.7.0 is using the following Python interpreter:
    C:\Users\runneradmin\micromamba\envs\test-env\python.exe (Python 3.12)
INFO: Building Cantera from git commit '94219d9'
Compiling for architecture: amd64
Compiling using the following toolchain(s): ['mingw', 'f90']
INFO: Compiling on 'AMD64 Family 25 Model 1 Stepping 1, AuthenticAMD'
INFO: Configuration variables read from 'cantera.conf' and command line:
    f90_interface = 'n'
    system_eigen = 'y'
    system_highfive = 'y'
    system_yamlcpp = 'y'
    debug = False
    logging = 'debug'

It seems like the value of the toolchain setting isn't getting saved in cantera.conf, which is as far as I've gotten so far.

[ischoegl]

It looks like newest actions no longer fail. This may have been a temporary issue on GH? It could be argued that there is an underlying issue as we don’t write variables to cantera.conf though.

[speth]

Yes, I still want to investigate. I've been traveling lately and haven't had access to my Windows machine at home where I was hoping to check on this.