Tarballing without Git metadata

Question

My source tree contains several directories which are using Git source control, and I need to tarball the whole tree excluding any references to the Git metadata or custom log files.

I thought I'd have a go using a combination of find/egrep/xargs/tar, but somehow the tar file contains the .git directories and the *.log files.

This is what I have:

find -type f . | egrep -v '\.git|\.log' | xargs tar rvf ~/app.tar

Can someone explain my misunderstanding here? Why is tar processing the files that find and egrep are filtering?

I'm open to other techniques as well.

score 107 · Accepted Answer · edited Nov 25 '20 at 13:44

107

You will get a nasty surprise when the number of files increase to more than one xargs command: Then you will first make a tar file of the first files and then overwrite the same tar file with the rest of the files.

GNU tar has the --exclude option which will solve this issue:

tar cvf ~/app.tar --exclude .git --exclude "*.log" .

edited Nov 25 '20 at 13:44

Yuri

4,254
1
29
46

answered Jun 18 '10 at 21:47

Ole Tange

31,768
5
86
104

score 58 · Answer 2 · edited Aug 06 '17 at 01:46

58

You can try directly with the tar option --exclude-vcs:

--exclude-vcs:
          Exclude version control system directories

For example:

tar cvfj nameoffile.tar.bz2 directory/ --exclude-vcs

It works with Git.

edited Aug 06 '17 at 01:46

Peter Mortensen

30,738
21
105
131

answered Feb 20 '13 at 21:49

Ivan

19,560
31
97
141

2

Thanks but "Option --exclude-vcs is not supported" with my version of tar. – zaf Feb 21 '13 at 07:18
5

@zaf yes = GNU tar only. MacOSX ships with bsdtar which doesn't have it. – marko May 22 '13 at 16:04
@Marko Aha, thanks for enlightening me. These details bug me always. – zaf May 23 '13 at 08:32

score 16 · Answer 3 · answered Jun 18 '10 at 12:36

16

Try something like this:

git archive --format=tar -o ~/tarball.tar -v HEAD

Add your .log files and everything else you don't want to be packed to your .gitignore file.

answered Jun 18 '10 at 12:36

jkramer

15,440
5
47
48

nice, is there a way to get a snapshot of code without making a commit tho? I want to archive the current directory even if files are not staged or committed? – Alexander Mills May 03 '19 at 22:32

score 13 · Answer 4 · answered Jul 10 '12 at 01:01

13

To exclude version control system directories:

tar --exclude-vcs

This will exclude svn, git metafiles etc.

answered Jul 10 '12 at 01:01

PeloNZ

460
5
12

score 5 · Answer 5 · edited Aug 06 '17 at 01:46

5

The newer GNU tar has the option to exclude version control directories automatically by using flag --exclude-vcs . This will take care of .git as well.

edited Aug 06 '17 at 01:46

Peter Mortensen

30,738
21
105
131

answered Jan 08 '16 at 16:58

KeshV

626
7
8

score 4 · Answer 6 · edited Aug 06 '17 at 01:45

4

[slap] Bah! The parameters to find were in the wrong order! I didn't see the warnings because they whizzed off the screen. This allowed '.' to pass through egrep which caused tar to slurp up everything.

That will teach me for drowning important messages in verbose debug.

This works:

find . -type f | egrep -v '\.git|\.log' | xargs tar cvf ~/app.tar

edited Aug 06 '17 at 01:45

Peter Mortensen

30,738
21
105
131

answered Jun 18 '10 at 12:36

zaf

22,776
12
65
95

score 3 · Answer 7 · edited Aug 06 '17 at 01:43

3

git-archive may be what you're looking for.

edited Aug 06 '17 at 01:43

Peter Mortensen

30,738
21
105
131

answered Jun 18 '10 at 12:29

Skilldrick

69,215
34
177
229

1

Probably not. Since it's multiple gits in the source tree. – zaf Jun 18 '10 at 12:38

score 3 · Answer 8 · answered Jun 18 '10 at 12:33

3

You could do that without grep. find is powerful

 find . -type f -not \( -iname ".git" -or -iname ".log" \) | xargs ...

answered Jun 18 '10 at 12:33

Jürgen Steinblock

30,746
24
119
189

The problem was my find parameters. See my answer. Thanks for the alternative. – zaf Jun 18 '10 at 12:39

score 2 · Answer 9 · edited Aug 06 '17 at 01:45

2

For doing it from outside the app directory:

tar cvfz app.tar.gz --exclude ".git/*" --exclude ".git" app/

edited Aug 06 '17 at 01:45

Peter Mortensen

30,738
21
105
131

answered Nov 01 '12 at 06:49

Prasanth Jayachandran

550
4
7

score 1 · Answer 10 · answered Aug 19 '17 at 09:10

1

For me, the .gitignore content is what I needed:

tar cvfz $PROJECT.tar.gz --exclude-from=$PROJECT/.gitignore $PROJECT

--exclude-from is reading the file's contents - and exlucde the patterns listed there

answered Aug 19 '17 at 09:10

Ricky Levi

7,298
1
57
65

there's also --exclude-vcs-ignores which reads exclude patterns from the VCS ignore files. – colin lamarre Jan 16 '18 at 22:39

Tarballing without Git metadata

10 Answers10