What's not in git.git

[Junio C Hamano <gitster@pobox.com>]

A list of recent issues discussed without any patch landing in
my tree.

* git-revert <path>

  As discussed, I am in favor of the above command, only when
  <path> cannot be interpreted as a valid commit object name and
  <path> appears in the index, to run "git-checkout -- <path>".

  Also, I am in favor of making "git-revert <commit> -- <path>"
  and "git-cherry-pick <commit> -- <path>" create a commit that
  reverts or cherry-picks the effect of named commit but only
  on named paths, as proposed by Dscho.

  Before the latter materializes, however, I think it is
  possible to implement the former.

  - "git-revert [--] <path>" runs "git-checkout -- <path>"
    without complaining (only when unambiguous if no -- is
    given);

  - "git-cherry-pick [--] <path>" errors out, as it does not
    make any sense.

  - "git-(revert|cherry-pick) <commit> [--] <path>" gives
    "unimplemented yet" error message;

* "Why a merge is more difficult to bisect" document.

  I think the last text with rewording suggestion on the list
  was good.  Should I take that directly, or pull through
  Bruce?

* "Batch mode of git-cat-file" aka git-fast-export.

  I think the idea is sound.

   http://thread.gmane.org/gmane.comp.version-control.git/65053/focus=65057
   http://thread.gmane.org/gmane.comp.version-control.git/62295/focus=62441

* Solaris portability improvements and clean-up.

  Test patches sent for git-compat-util.h and Makefile on
  FLEX_ARRAY and mkdtemp() but they haven't resulted in any
  appliable patch yet.