Re: How to turn off rename detection for cherry-pick?

[Junio C Hamano <gitster@pobox.com>]

Jeff King <peff@peff.net> writes:

> though I'm not sure how DIFF_DETECT_COPIES should be handled
> ("recursive" silently downgrades it to DIFF_DETECT_RENAME).

I somehow suspect that we would want to go there to avoid
overloading ourselves (and end-users) with conceptual complexity.

If we detect that the file the side branch modified was copied to
create multiple files on our side (with or without modification on
our end), where would we want to reapply their changes to?  To the
original file only?  To all copies?  The same story goes for the
opposite direction.  How should possibly different changes they made
to their copies be merged back to the sole original file we have
(which we may or may not have modified)?  What if the both sides
made copies of the same file, and now we need to shuffle NxM
combination of changes?  I would expect it would be a whole can of
worms, which we _may_ be able to define concrete rules how we would
handle each of these cases but we may have a hard time to explain in
simple terms to end-users.