This is kind of a re-roll of [1] (wow, apparently it took me almost
two months to get done). The goal was, then and now, to teach "git
reset" to work on an unborn branch and to not require a commit when a
tree would do. This time, I also made some tangential improvements
along the way, mostly related to readability and performance.

As usual, the risker patches are towards the end. In particular, I
find it hard to evaluate how risky the last patch is. That last patch
is responsible for much of the improvements in the timing table below,
so it would be nice if it doesn't break things too badly (test pass,
of course). The timings are best-of-five, wall time.

Command                  Before     After
reset (warm)             0.23        0.07
reset -q (warm)          0.23        0.03
reset . (warm)           0.09        0.07
reset -q . (warm)        0.09        0.03
reset --keep (warm)      0.31        0.29
reset --keep -q (warm)   0.31        0.29
reset (cold)             9.74        2.60
reset -q (cold)          9.85        0.37
reset . (cold)           2.66        2.51
reset -q . (cold)        2.59        0.33
reset --keep (cold)      7.58        7.52
reset --keep -q (cold)   7.37        7.21



  [1] http://thread.gmane.org/gmane.comp.version-control.git/210568/focus=210855

Martin von Zweigbergk (19):
  reset $pathspec: no need to discard index
  reset $pathspec: exit with code 0 if successful
  reset.c: pass pathspec around instead of (prefix, argv) pair
  reset: don't allow "git reset -- $pathspec" in bare repo
  reset.c: extract function for parsing arguments
  reset.c: remove unnecessary variable 'i'
  reset.c: extract function for updating {ORIG,}HEAD
  reset.c: share call to die_if_unmerged_cache()
  reset.c: replace switch by if-else
  reset --keep: only write index file once
  reset: avoid redundant error message
  reset.c: move update_index_refresh() call out of read_from_tree()
  reset.c: move lock, write and commit out of update_index_refresh()
  reset [--mixed]: don't write index file twice
  reset.c: finish entire cmd_reset() whether or not pathspec is given
  reset [--mixed] --quiet: don't refresh index
  reset $sha1 $pathspec: require $sha1 only to be treeish
  reset: allow reset on unborn branch
  reset [--mixed]: use diff-based reset whether or not pathspec was
    given

 builtin/reset.c                | 281 +++++++++++++++++++----------------------
 t/t2013-checkout-submodule.sh  |   2 +-
 t/t7102-reset.sh               |  26 +++-
 t/t7106-reset-unborn-branch.sh |  52 ++++++++
 4 files changed, 200 insertions(+), 161 deletions(-)
 create mode 100755 t/t7106-reset-unborn-branch.sh

-- 
1.8.1.rc3.331.g1ef2165

--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to