The ui.editor code calls util.tonativeeol() to normalize line endings
to the platform native format. So, editor files may have CRLF
in them. For some reason, just this one specific test case was
failing on Python 3 on Windows. (The output before this change
was ...\r (esc), but only on the first line of output. I suspect
the test harness output parser was getting confused by the specific
byte sequences in this output or something.
I wanted to hack around this by filtering output via sed to remove
the CR. But check-code says this practice is apparently banned.
So I just made the test conditional.