[PATCH 0/5] Sequencer: working around historical mistakes

Hi,

As described in the discussion following $gmane/179304/focus=179383, we have decided to handle historical hacks in the sequencer itself. This series that follows is one step in the right direction.

- Part 1/5 makes the gigantic move required to create the sequencer.
If you need an excuse to celebrate, wait till this gets merged :)
- Part 5/5 can be considered as the "ultimate objective" of the
series.  I first wrote this part, and then wrote the other parts to
make tests pass.
- Parts 3/5 and 4/5 are ugly!  Causes heartburn.

Immediate shortcomings of this iteration:
1. No tests yet.  I want to see if it's possible to make this less
ugly first.
2. This series depends on rr/revert-cherry-pick, but doesn't apply to
the current 'next'- sorry, rebasing is a massive pita due to 1/5.

Thanks for reading.

-- Ram

Ramkumar Ramachandra (5):
  sequencer: factor code out of revert builtin
  sequencer: remove CHERRY_PICK_HEAD with sequencer state
  sequencer: sequencer state is useless without todo
  sequencer: handle single commit pick separately
  sequencer: revert d3f4628e

 builtin/revert.c                |  821 +--------------------------------------
 sequencer.c                     |  832 ++++++++++++++++++++++++++++++++++++++-
 sequencer.h                     |   26 ++
 t/t3510-cherry-pick-sequence.sh |   24 --
 4 files changed, 847 insertions(+), 856 deletions(-)