bug#28255: grep erroneously skips Microsoft UTF-8 text files as being binary

[Paul Eggert]

Simon wrote:
Windows text files can start with a byte order mark of U+FEFF and then
be encoded in UTF-8.  These are skipped as being binary files.

I can't reproduce this problem on Fedora 26 x86-64. Here's how I tried:

$ printf '\357\273\277x\n' >t
$ LC_ALL=C grep x t | od -c
0000000 357 273 277   x  \n
0000005

To help us diagnose the problem, please send a simple, self-contained example, 
and mention your platform.