find files and tar them

[Shivkumar]

Hi,

How do i use find and tar command together ?
I just want to find some files and tar the output in a tar file.

Thanks,
Shiv

[James R. Ferguson · Accepted solution]

Hi Shiv:

If we assume HP-UX (since it is in that community that you ask), then I'd do something like this:

# FILES=$(find /path -xdev -type f -mtime +7)
# tar -cvf /tmp/myarchive ${FILES}

...or whatever your specification is for collecting the files ...

Regards!

...JRF...

[Steven Schweda]

> # FILES=$(find /path -xdev -type f -mtime +7)
> # tar -cvf /tmp/myarchive ${FILES}

And what is the command-line length limit
these days?

And then there are the potential problems in
the funny-file-name class:

dyi # ls -l gtt
total 0
-rw-r--r-- 1 root sys 0 Aug 11 12:59 a a

dyi # FILES=$( find gtt -type f )

dyi # echo ${FILES}
gtt/a a

dyi # tar cvf gtt.tar ${FILES}
tar: cannot stat gtt/a. Not dumped.
tar: cannot stat a. Not dumped.

GNU "tar" has a "--files-from=file-name"
("-T file-name") option which can be helpful
in cases where "${FILES}" would cause
problems. "--files-from=-" or "-T -" works
to get the file names from stdin, so a "find"
pipeline can feed in the file names. For
example:

dyi # find gtt -type f | gtar cfv gtt.tar -T -
gtt/a a

dyi # tar tfv gtt.tar
rw-r--r-- 0/3 0 Aug 11 12:59 2009 gtt/a a


dyi # uname -a
HP-UX dyi B.11.31 U ia64 4235313755 unlimited-user license

dyi # gtar --version
tar (GNU tar) 1.18
[...]

For a variety of reasons, it's often a bad
idea to use file-name wild cards (like "*"),
or equivalent constructs, in a command line.
One of the best reasons to use "find" is to
avoid problems like this, but sloppy use of
"find" can effectively demolish many of its
advantages.

[Andrey Chipalyuk]

You can try -exec option of find command also.
For example:

find ./ -name "..." -exec tar cvf {} \;

[Dennis Handly]

>How do I use find and tar command together?

If you have a trivial number of files, you can use JRF's example. Or you use Steven's gtar example.

Otherwise you don't use tar. Instead you use pax(1):
find /path -xdev -type f -mtime +7 | pax -v -w -f /tmp/myarchive

You can use pax or tar to restore.

>Andrey: You can try -exec option of find command also.

There are lots of reasons this won't work, don't even try.

[Steven Schweda]

> There are lots of reasons this won't work,
> don't even try.

Or, if you're the fellow who made the
suggestion, then try it _before_ posting it.

[James R. Ferguson]

Hi (again):

> Steven: And what is the command-line length limit these days?

Far larger than the days of HP-UX 10.20 and prior:

# getconf ARG_MAX
2048000

That said, on an "un-enhanced" HP-UX system (i.e. without some of the really nice GNU tools), I agree with Dennis --- 'pax' is probably the better choice. Oh, and your point about nasty files with embedded spaces is well taken, of course.

Regards!

...JRF...