On Thu, Jun 10, 2021 at 11:00:59AM -0500, Felipe Contreras wrote:

quoted

I didn't look too deeply at your example, but I suspect it may be
related to the fact that diff3 does not try to minimize the conflicts as
much (and then the recursive merge on top of that piles on extra layers
of confusion).

There's a lot more discussion in this old thread:

  https://lore.kernel.org/git/20130306150548.GC15375@pengutronix.de/ (local)

Geezus. My patches always end up kicking the hornest nest don't they?

Maybe it would make sense to revive the zdiff3 patch and attempt to make
that the default. That would take a lot of time though, so it wasn't as
easy as just flipping a switch from "merge" to "diff3".

I had that patch in my daily build for several years, and I would
occasionally trigger it when seeing an ugly conflict. IIRC, it
segfaulted on me a few times, but I never tracked down the bug. Just a
caution in case anybody wants to resurrect it.

-Peff