replace invalid utf8 sequences by UTF-8 REPLACEMENT CHARACTER (efbfbd)
I still strongly disagree with the git maintainers not to hint people,
to use the only sane default encoding for a distributed project,
which is utf8. I'm tired of hearing filesystem development arguments.
Git is a software offered to merge forth and back across the world
and not to provide a content neutral filesystem.
Btw: I have nothing against the ability to run git in a closed environment,
with a different encoding, that's fine, sure. But that is obviously not
the case for the projects on kernel.org. It's about sane defaults,
nothing else.
You have to make decisions guy, as always in life. The problems to
allow random encoded garbage in commit messages _without_ storing
the encoding, just makes zero sense. Eighter you introduce a per-commit
encoding field, if you insist on this craziness, or you define a default
encoding. Everything else is just lazy and does not solve any problem,
besides that you can claim now, that you are not responsible for the mess
in the repository.
Gitweb shows several commits at once, you allow various encodings committed
to the same repository, without any hint what that garbage from the
individual commits is encoded with. No idea why you don't get
the problem - it's unsolvable. If you merge different peoples work, you
have to speak a common language!