[PATCH 0/9] Fix oversimplified optimization for add_cache_entry().

LT> I really don't want to do this. LT> Can you fix the "optimized" one instead?

JCH> Will do.

I have this list which logically consists of three sets. I am using the new "git rebase" to maintain it and your not taking my patch stream gave me a good opportunity to test it ;-).

  [PATCH 1/9] [RFC] fix date parsing for GIT raw commit timestamp format.
  [PATCH 2/9] git-commit-script: get commit message from an existing one.
  [PATCH 3/9] git-cherry: find commits not merged upstream.
  [PATCH 4/9] git-rebase-script: rebase local commits to new upstream head.

These four are updated "git rebase" I've written while on-road without knowing about your git-patch-id; the updated one uses git-patch-id to identify which are merged and which are not.

To carry earlier commit datetime forward, date.c needs to be fixed (PATCH 1/9, which I sent you separately Friday), and enhancing git-commit-script to be able to specify an existing commit to slurp the commit message (and author & date information) was necessary (well, not strictly necessary, but that is the way I am used to, which is stolen from JIT), which is PATCH 2/9. PATCH 3/9 and PATCH 4/9 use git-patch-id to implement the cherrypick/rebase.

  [PATCH 5/9] Add more tests for read-tree --emu23.
  [PATCH 6/9] git-merge-one-file-script: do not misinterpret rm failure
  [PATCH 7/9] Fix oversimplified optimization for add_cache_entry().

These three are reworked D/F conflict fix. PATCH 5/9 is the same demonstration of what is broken in the current code. The problem fixed by PATCH 6/9 was discovered while I was trying out the fixes. PATCH 7/9 teaches your "optimized" has_file/has_dir implementation to honor stages while they operate.

  [PATCH 8/9] http-pull: documentation updates.
  [PATCH 9/9] Add a bit of developer documentation to pull.h

These are resend.