Jeff King wrote:
quoted text
>
> I wrote this after reading just your first patch, and it looks like
> you've made much progress since. It seems like Thomas' patch could just
> get rid of all the recounting entirely, and just pass off the edited
> hunks to "git apply --recount". Which should make his patch much smaller
> and more straightforward.

I think there's no way to split hunks, in the way I currently "help"
with @@ line guessing, using just the --recount feature.  Unless the
editor helps you with adding complete correct @@ lines in the middle
of hunks (Emacs does that).  I don't think it is at all possible to
remove the middle part of a hunk in Johannes' scheme without somehow
figuring out the corresponding @@ line (or at least its old line
number) or editing away every +/- line.

Then again it's not always easy even with my patch: you may have to
manually insert extra context because of the rule against zero lines
of context.  We could guess the context too if we already guessed the
line numbers, but that would mean even more complexity, which you
don't seem to like...

=2D Thomas

=2D-=20
Thomas Rast
trast@student.ethz.ch