migur wrote:
>
> This is 5.0.5 box.
> I am trying to accumulate multiple files in a single tar file.
> Files are added daily, so I would need to add them to tar file,
> uniqness of the file names is guarantied.
>
> Here is simple example showing the problem I am running into:
>
> $ tar ncvf A x*
> seek = 0K a x1 548K
> seek = 548K a x2 252K
>
> $ tar nuvf A y*
> tar: buffer size = 10K
> seek = 800K a y1 594K
> seek = 1395K a y2 594K
> seek = 1989K a y3 594K
>
> $ tar tvf A
> tar: blocksize = 20
> rw-rw-r--243/50 560131 Dec 18 13:29 2002 x1
> rw-rw-r--243/50 257334 Dec 18 14:55 2002 x2
> tar: directory checksum error
>
> $ l
> total 10374
> -rw-rw-r-- 1 haa group 2637824 Apr 12 18:22 A
> -rw-rw-r-- 1 haa group 560131 Dec 18 2002 x1
> -rw-rw-r-- 1 haa group 257334 Dec 18 2002 x2
> -rw-rw-r-- 1 haa group 608255 Dec 18 2002 y1
> -rw-rw-r-- 1 haa group 608226 Dec 18 2002 y2
> -rw-rw-r-- 1 haa group 607769 Dec 18 2002 y3
>
> I tried tar "r" instead of "u" with the same outcome: resulting in a
> tar file with directory checksum error.
>
> Any ideas how to do it properly will be appreciated,
> Regards,
> Michael.

Ran into a similar issue with 5.0.2 IFRC. A script would add email
to a tar archive, EDI sessions in reality, that we wanted to keep
track of. Everything worked perfectly for a while, email would
accumulate in the mail file, then would be copied to a unique file
and added to the tar archive before the application read it. To
my surprise I found the archive unreadable after a while. Many
attempts to debug the issue resulted in random failures. I reported
the issue to SCO, who were unable to duplicate the problem.

When you have a bad tar archive, scan through it with "hd" to the
failure location ( which you can determine from the file sizes ).
Do you see some FF or 00 bytes?

Mike

--
Michael Brown

The Kingsway Group