Junio C Hamano [off-list ref] writes:
I think the way we handled cr-at-eol was suboptimal. We should be able to
link this to crlf attribute (and core.autocrlf configuration) and pretend
as if cr-at-eol was given if a file is subject to the crlf conversion
(iow,. cr-at-eol should be deprecated/removed as a mistake).
I was a moron. I think cr-at-eol was really the best we could do, as it
is to allow CR at the end of the line _in the tracked contents_; iow,
people who would want to use this would not be using crlf attribute at all.
So please strike this part out, but everything else stays the same.