Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!ucsd!ames!sgi!shinobu!odin!anchor!olson
From: olson@anchor.esd.sgi.com (Dave Olson)
Newsgroups: comp.sys.sgi
Subject: Re: The tar -r optione
Message-ID: <1990Dec4.204231.27468@odin.corp.sgi.com>
Date: 4 Dec 90 20:42:31 GMT
References: <9012031527.AA08340@iris.kth.dk>
Sender: news@odin.corp.sgi.com (Net News)
Organization: Silicon Graphics, Inc.  Mountain View, CA
Lines: 35

In <9012031527.AA08340@iris.kth.dk> operator@IRIS.KTH.DK (Martin Liversage) writes:

| We just got a Kennedy 9600 1/2" 9 track tape station interfaced
| through the SCSI bus. I want to be able to create incremental archives
| on a tape. That is, I want to create an archive an later add more
| files to it. Sounds easy, doesn't it? Just use tar -r. The man page
| describes the -r option:
| 
| > Append the named files at the end of the archive. On tape, named files
| > are appended at the end of the last archive on tape. This function is
| > only supported on half-inch nine track tape drives.
| 
| Well it doesn't work (so much for RTFM). Instead tar aborts with an
| error message. I got the GNU tar and it didn't work either. After some
| edge'ing I discovered what tar does (this is just what I think it does
| - I'm no tar wizard).

Please folks: many of us at SGI are willing to help with your problems,
but it can be difficult when you don't say what hardware or OS release
you have.  In this case I have adequate hardware info, but no OS info.
tar in 3.2 had some grungy code that limited -r and -u support to the
xm driver.  This wasn't caught (an oversight on my part) until too late
for 3.2.X, so it is fixed in 3.3.  If you don't have 3.3, this is probably
the problem.  Your summary of how -r and -u work are pretty much correct.

| So while appending to the archive the tape is opened for reading. Tar
| has to do an MTAFILE ioctl before doing the write, but it doesn't.

I'm not sure how you determined this, but if you have 3.2.X, you are correct,
see above.
--

	Dave Olson

Life would be so much easier if we could just look at the source code.