Add dangling objects tips.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Linus Torvalds committed
Jan 25, 2007 at 21:55 UTC
a9eefb3bfc66b7f5dba86ebf57b05e32c33eb9e5
1 file changed
+109
Documentation/howto/dangling-objects.txt
new
+109
@@ -0,0 +1,109 @@
1
+From: Linus Torvalds <torvalds@linux-foundation.org>
2
+Subject: Re: Question about fsck-objects output
3
+Date: Thu, 25 Jan 2007 12:01:06 -0800 (PST)
4
+Message-ID: <Pine.LNX.4.64.0701251144290.25027@woody.linux-foundation.org>
5
+Archived-At: <http://permalink.gmane.org/gmane.comp.version-control.git/37754>
6
+Abstract: Linus describes what dangling objects are, when they
7
+ are left behind, and how to view their relationship with branch
8
+ heads in gitk
9
+
10
+On Thu, 25 Jan 2007, Larry Streepy wrote:
11
+
12
+> Sorry to ask such a basic question, but I can't quite decipher the output of
13
+> fsck-objects. When I run it, I get this:
14
+>
15
+> git fsck-objects
16
+> dangling commit 2213f6d4dd39ca8baebd0427723723e63208521b
17
+> dangling commit f0d4e00196bd5ee54463e9ea7a0f0e8303da767f
18
+> dangling blob 6a6d0b01b3e96d49a8f2c7addd4ef8c3bd1f5761
19
+>
20
+>
21
+> Even after a "repack -a -d" they still exist. The man page has a short
22
+> explanation, but, at least for me, it wasn't fully enlightening. :-)
23
+>
24
+> The man page says that dangling commits could be "root" commits, but since my
25
+> repo started as a clone of another repo, I don't see how I could have any root
26
+> commits. Also, the page doesn't really describe what a dangling blob is.
27
+>
28
+> So, can someone explain what these artifacts are and if they are a problem
29
+> that I should be worried about?
30
+
31
+The most common situation is that you've rebased a branch (or you have
32
+pulled from somebody else who rebased a branch, like the "pu" branch in
33
+the git.git archive itself).
34
+
35
+What happens is that the old head of the original branch still exists, as
36
+does obviously everything it pointed to. The branch pointer itself just
37
+doesn't, since you replaced it with another one.
38
+
39
+However, there are certainly other situations too that cause dangling
40
+objects. For example, the "dangling blob" situation you have tends to be
41
+because you did a "git add" of a file, but then, before you actually
42
+committed it and made it part of the bigger picture, you changed something
43
+else in that file and committed that *updated* thing - the old state that
44
+you added originally ends up not being pointed to by any commit/tree, so
45
+it's now a dangling blob object.
46
+
47
+Similarly, when the "recursive" merge strategy runs, and finds that there
48
+are criss-cross merges and thus more than one merge base (which is fairly
49
+unusual, but it does happen), it will generate one temporary midway tree
50
+(or possibly even more, if you had lots of criss-crossing merges and
51
+more than two merge bases) as a temporary internal merge base, and again,
52
+those are real objects, but the end result will not end up pointing to
53
+them, so they end up "dangling" in your repository.
54
+
55
+Generally, dangling objects aren't anything to worry about. They can even
56
+be very useful: if you screw something up, the dangling objects can be how
57
+you recover your old tree (say, you did a rebase, and realized that you
58
+really didn't want to - you can look at what dangling objects you have,
59
+and decide to reset your head to some old dangling state).
60
+
61
+For commits, the most useful thing to do with dangling objects tends to be
62
+to do a simple
63
+
64
+ gitk <dangling-commit-sha-goes-here> --not --all
65
+
66
+which means exactly what it sounds like: it says that you want to see the
67
+commit history that is described by the dangling commit(s), but you do NOT
68
+want to see the history that is described by all your branches and tags
69
+(which are the things you normally reach). That basically shows you in a
70
+nice way what the danglign commit was (and notice that it might not be
71
+just one commit: we only report the "tip of the line" as being dangling,
72
+but there might be a whole deep and complex commit history that has gotten
73
+dropped - rebasing will do that).
74
+
75
+For blobs and trees, you can't do the same, but you can examine them. You
76
+can just do
77
+
78
+ git show <dangling-blob/tree-sha-goes-here>
79
+
80
+to show what the contents of the blob were (or, for a tree, basically what
81
+the "ls" for that directory was), and that may give you some idea of what
82
+the operation was that left that dangling object.
83
+
84
+Usually, dangling blobs and trees aren't very interesting. They're almost
85
+always the result of either being a half-way mergebase (the blob will
86
+often even have the conflict markers from a merge in it, if you have had
87
+conflicting merges that you fixed up by hand), or simply because you
88
+interrupted a "git fetch" with ^C or something like that, leaving _some_
89
+of the new objects in the object database, but just dangling and useless.
90
+
91
+Anyway, once you are sure that you're not interested in any dangling
92
+state, you can just prune all unreachable objects:
93
+
94
+ git prune
95
+
96
+and they'll be gone. But you should only run "git prune" on a quiescent
97
+repository - it's kind of like doing a filesystem fsck recovery: you don't
98
+want to do that while the filesystem is mounted.
99
+
100
+(The same is true of "git-fsck-objects" itself, btw - but since
101
+git-fsck-objects never actually *changes* the repository, it just reports
102
+on what it found, git-fsck-objects itself is never "dangerous" to run.
103
+Running it while somebody is actually changing the repository can cause
104
+confusing and scary messages, but it won't actually do anything bad. In
105
+contrast, running "git prune" while somebody is actively changing the
106
+repository is a *BAD* idea).
107
+
108
+ Linus
109
+