http://bugs.winehq.org/show_bug.cgi?id=18191
--- Comment #9 from Dmitry Timoshkov dmitry@codeweavers.com 2009-04-26 09:03:36 --- (In reply to comment #7)
In my past testing, it always saves as ISO-8859 (English locale), but if unicode only characters are used, I'm warned that data loss may occur and that I should save as UTF-8.
Most likely what notepad in Windows does is a check whether conversion from unicode to current code page leads to loss then it warns. That's not a "detection", just the WideCharToMultiByte feature.