Re: [Bug-tar] Problem with multiple archives and --concatenate

[Tim Kientzle]

Paul Fee wrote:
> The documentation for GNU tar suggests that multiple
> archives can be appended onto an existing archive.
>
> e.g.
> tar --concatenate --file=first.tar append1.tar
> append2.tar

If you do this kind of archive manipulation often,
you might find bsdtar's approach to this problem
of interest.  Instead of providing an option to
"concatenate archives", it provides a way to
read entries from one or more existing archives
when creating a new archive.

It does this by recognizing '@archive' as an instruction
to read entries from the specified archive (instead
of treating the archive as a file on disk):

$ tar --version
bsdtar 2.5.5 - libarchive 2.5.5
$ echo start > start
$ echo one > one
$ echo two > two
$ echo three > three
$ tar cf first.tar start
$ tar czf append1.tgz one
$ tar cyf append2.tbz two
$ tar cz --file output.tgz @first.tar  \
   @append1.tgz @append2.tbz three
$ tar tvf output.tgz
-rw-r--r--  0 tim    tim   6 Aug 19 09:20 start
-rw-r--r--  0 tim    tim   4 Aug 19 09:20 one
-rw-r--r--  0 tim    tim   4 Aug 19 09:20 two
-rw-r--r--  0 tim    tim   6 Aug 19 09:20 three

You can combine this with options such as --exclude
and --include to select which entries get carried
through and even use --format to determine the
format of the result:

  tar czf output.tgz --format=pax --exclude=foo/   \
     @one.zip file2 @three.cpio dir4

or use stdio to create an archive-conversion filter:

  tar czf - --include=foo/ --format=ustar @-

There are some limits, of course.  In particular,
converting between dissimilar formats sometimes breaks
hardlinks.

This approach is necessarily a little slower than the
GNU tar --concatenate option, but it gives you a lot
more flexibility.  I thought the folks on this list
might find the approach worth considering.

Cheers,

Tim Kientzle