1 Date: Sat, 13 Aug 2005 22:16:02 -0700 (PDT)
2 From: Linus Torvalds <torvalds@osdl.org>
3 To: Steve French <smfrench@austin.rr.com>
4 cc: git@vger.kernel.org
5 Subject: Re: sending changesets from the middle of a git tree
6 Abstract: In this article, Linus demonstrates how a broken commit
7 in a sequence of commits can be removed by rewinding the head and
8 reapplying selected changes.
10 On Sat, 13 Aug 2005, Linus Torvalds wrote:
12 > That's correct. Same things apply: you can move a patch over, and create a
13 > new one with a modified comment, but basically the _old_ commit will be
18 You can entirely _drop_ old branches, so commits may be immutable, but
19 nothing forces you to keep them. Of course, when you drop a commit, you'll
20 always end up dropping all the commits that depended on it, and if you
21 actually got somebody else to pull that commit you can't drop it from
22 _their_ repository, but undoing things is not impossible.
24 For example, let's say that you've made a mess of things: you've committed
25 three commits "old->a->b->c", and you notice that "a" was broken, but you
26 want to save "b" and "c". What you can do is
28 # Create a branch "broken" that is the current code
32 # Reset the main branch to three parents back: this
33 # effectively undoes the three top commits
37 # Check the result visually to make sure you know what's
41 # Re-apply the two top ones from "broken"
43 # First "parent of broken" (aka b):
44 git-diff-tree -p broken^ | git-apply --index
45 git commit --reedit=broken^
47 # Then "top of broken" (aka c):
48 git-diff-tree -p broken | git-apply --index
49 git commit --reedit=broken
51 and you've now re-applied (and possibly edited the comments) the two
52 commits b/c, and commit "a" is basically gone (it still exists in the
53 "broken" branch, of course).
55 Finally, check out the end result again:
57 # Look at the new commit history
60 to see that everything looks sensible.
62 And then, you can just remove the broken branch if you decide you really
65 # remove 'broken' branch
68 # Prune old objects if you're really really sure
71 And yeah, I'm sure there are other ways of doing this. And as usual, the
72 above is totally untested, and I just wrote it down in this email, so if
73 I've done something wrong, you'll have to figure it out on your own ;)
77 To unsubscribe from this list: send the line "unsubscribe git" in
78 the body of a message to majordomo@vger.kernel.org
79 More majordomo info at http://vger.kernel.org/majordomo-info.html