From a practical standpoint, the observed results remain the same.

I didn't contradict the results. It is indeed based on the shell (and not just in Windows). I just protested your claim that pryrt was wrong.

(As an aside, did you know that Windows programs must parse their own command lines for arguments? That means that somewhere along the process of a cygwin program calling a Windows program, there must be a step that involves generating a Windows-compatible command line from the one actually used. Fun.)

If we're going to help people with the problems they see, I think we should learn as much as we can about the tools they use.

That's why I spoke up. Claiming that running a GNU/Linux program is running a Windows program doesn't help anyone. One should be aware that one is running a program in a virtual box when one does so because there are ramifications.

Thanks for the defense, ++ikegami. :-) I was at least ambiguous: in the modern world of Microsoft pushing PowerShell and de-emphasizing (even taking it out of extended context menus) cmd.exe, I need to train myself to be explicit about testing in cmd (though in this case, the cmd vs ps behavior was effectively the same), so I also ++footpad :-). (Besides, my "in windows" mindset is still often muddied by using gnuwin32 ports to make cmd.exe more usable, and I have to be careful to not include that in pure-windows advice without pointing it out. :-) )

edit: And yes, I understand ikegami's main point wasn't just my defense / validating my ego. ;-) I love the back and forth, with the additional information, and corrections when things are incomplete, ambiguous, or downright wrong. I have learned from this thread, and hope the OP has as well. (When I think I can help the OP, or when I want something clarified to help me learn, or both, are my usual times to participate.)

I do agree that the discussion should have been about shells, not OSs, though. So I'm not really defending you either :)