I'm not sure we can be positive that cygpath is always available, can we? Can this be changed to support the case where cygpath isn't available?

Thank you for your review.

The script assumes Cygwin (e.g.: https://github.com/libffi/libffi/blob/master/msvcc.sh#L168-L185).

Your proposal might make sense in order to support WSL. I can't test it, though.

Please check f73c4ea, which works under Cygwin.

The build failure seems unrelated to the changes:

Build execution time has reached the maximum allowed time for your plan (60 minutes).

Could you verify?

Somewhat related... Working with cygwin from cmd, I find issues with improper quoting as illustrated here: java-native-access/jna@112d76d. Specifically, there are relative includes (e.g. ./) which cl.exe is unaware of as well as paths such as C:\Program Files\AdoptOpenJDK\..\include which aren't properly quoted and break command-line invocation.

Your proposal might make sense in order to support WSL

Apparently wslpath is a thing? microsoft/WSL#522 (comment)

MSYS2 would also benefit from such a tool although the MSYS1 documentation states it does it automatically?

"For any executable not dependent on msys-1.0.dll, MSYS will convert POSIX paths that are used as arguments to Win32 paths. This page attempts to document the heuristics MSYS uses to decide which arguments are POSIX paths and how they should be converted based on the source of msys_p2w in path.cc."