tar 1.13: multi-volume extract fails

[Phil Edwards]

I've created a multi-tape backup using the --multi-volume and
--listed-incremental=FILE options.  (Also --label="the_tape_label", which
shows up below.  With a more descriptive name, of course.)  Now I'm trying
to restore from those tapes.

The first volume went fine, although instead of restoring

    dir/subdir/file1
    dir/subdir/file2
    ...

it's creating

    huge_timestamp/dir/subdir/file1
    different_huge_timestamp/dir/subdir/file2
    ...

At least I'm guessing they're timestamps.  Whatever they are, the directory
tree is a complete mess.  That's not actually the problem though; I can
reassemble the tree with some trivial shell script.  The problem is when
it tries to move to the second volume:

    ...
    funky_timestamp/dir/subdir/a_50_meg_file
    Prepare volume #2 for /dev/rmt/1n and hit return:
    Reading the_tape_label Volume 2
    gtar: funky_timestamp/dir/subdir/a_50_meg_file is not continued on this 
volume
    Prepare volume #2 for /dev/rmt/1n and hit return:

Well, crap.  So I abort the extract, rewind the second tape, and fire
off the command again.  The manual says that the volumes are independent,
so maybe this'll work.

Nope.  tar prints:

    the_tape_label Volume 2
    Reading the_tape_label Volume 2
    dir/subdir/a_50_meg_file
    gtar: Cannot extract `dir/subdir/a_50_meg_file' -- file is continued from 
another volume
    different_funky_timestamp/dir/subdir/a_different_file
    ...

What am I doing wrong?  How do I get that elusive split-across-tapes file?
(And what the heck are those strings of numbers now strewn all over the
restored directory tree?  The manual didn't mention those.)


Phil

-- 
I would therefore like to posit that computing's central challenge, viz. "How
not to make a mess of it," has /not/ been met.
                                                 - Edsger Dijkstra, 1930-2002