For some tips on fixing/rewriting commits, see Git Tips.
The plumbing command to create a git commit is git commit-tree:
git commit-tree <tree> [(-p <parent>)...] \
[-S[<keyid>]] [(-m <message>)...] [(-F <file>)...]
Multiple -m arguments specify an ordered list of paragraphs in the
commit message. Stdin is used if -mF are not given or they result
in an empty message. (Use </dev/null) to get an empty message.)
Author and committer name/date information is taken in order from
$GIT_{AUTHOR,COMMITTER}_{NAME,EMAIL,DATE}, user.{name,email} and
$EMAIL. <, > and \n are stripped. Dates are formatted as
@<unix-timestamp> <tzoffset>, ISO-8601 or similar.
Show a raw commit with git cat-file -p REF.
git write-tree creates a tree from an index. The index must be in
a fully merged state.
git stash stores its data as commits marked with the refs/stash ref;
multiple stashes are found through the reflog of refs/stash, e.g.:
$ stash list
stash@{0}: WIP on master: d569aec two
stash@{1}: On master: stash2
The stash consists of several commits 2-3 of which are new:
- head: The current
HEADcommit when you stashed (pre-existing). - untracked: Optional; an orphan (parentless) commit containing a
tree of all untracked files at the time of the stash. This is
created only with the
-u(untracked files) or-a(all files, untracked and ignored) options. - index: the staged commits (in the index) at the time of stash, with parents head and, if present, untracked.
- wc: The working copy's tracked but unstaged files at the time you stashed, with parents head and index.
.----W----.
/ / /
-----H----I U
See also more details on the stash algorithms.