It will not work in all cases. For example if the stdio encoding is UTF-8 and the filesystem encoding is Latin1. Or the stdio encoding is CP1251 and the filesystem encoding is UTF-8. I am not also sure that it gives us the result which we want if it doesn't fail.

It is a general and complex issue, and every program which writes file names to stdout is affected.

For now I suggest just use TESTFN_ASCII instead of TESTFN. We will find better solution in future. I hesitate about merging PR 28060 because it can fail also on some non-Windows buildbots with uncommon locale settings.