Date: Thu, 29 Jun 1995 08:33:42 -0400 From: Thomas David Rivers <ponds!rivers@dg-rtp.dg.com> To: freebsd-hackers@freefall.cdrom.com Subject: Problem with multi-volume GNU tar files. Message-ID: <199506291233.IAA03589@ponds.UUCP>
next in thread | raw e-mail | index | archive | help
Well - here's another one. I've discovered (while backing up/restoring for a 2.0 -> 2.0.5 using tar) that there is a subtle bug in GNU tar when multivolume requests are made and a sufficiently large file spans a volume. I believe the file has to be large enough so that the remainder (on the next volume) is larger than the block size (which defaults to 10240 bytes.) There is a problem with the count of how much was written to the first volume, and how much is left to write on the second. It turns out the first volume contains more bytes than the 'tar' program "believes", so the headers for the next volume don't contain the correct # of bytes. The symptom of this phenomena is a message (when restoring) of: This volume is out of sequence when, if fact, it is in sequence. I've determined that no bytes are lost; just some confusion about what is where. Judicious use of 'pax' to retrieve the pieces from the tar archives, along with 'dd' to get the right bytes appears to be able to reconstruct the files. I started to look into this; but the volumes I have are 150-meg QIC tapes that take forever to get through... so, I hope to reproduce this with two small floppy images for debugging. Until someone (hopefully me) gets to this, I'd suggest using cpio instead of 'tar' for saving things... - Dave Rivers -
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199506291233.IAA03589>