On Thu, Apr 15, 2004 at 04:08:02PM -0700, Joe Rhett wrote:
> > Joe, I'd put debug printf's in the exclude parsing stuff in gnu tar to
> > debug your problem.  By the time you get this far, amanda should be
> > out of the equation.  It just invokes runtar which invokes gnu tar
> > with certain args (as shown in runtar.*.debug).
...
> 
> Again, I've already done this.  AND I've run the exact same command as
> amanda logs that it is running and I don't get those errors.
> 
> Amanda appears to be running a different command than it logs that it 
> is running. 

In an earlier post you indicated you had replaced tar with a perl script
to analyze the call to tar.  At that time you did, or could have, determined
what arguments tar is "actually" called with and compare them with the
debug log.  I'm guessing, since you did not say anything about that, that
you either did not check or they were the same.  What is the actual situation?


In another message this conversation took place:

> On Thu, Apr 15, 2004 at 05:40:28PM -0600, John E Hein wrote:
> > All I've got left for you is to sprinkle fprintf's in amanda/gnu tar
> > code and find out what's really going on, or hack it so you can attach
> > a debugger and step through the code.
> 
> John, you appear to be missing the point in grand style.  TAR is not the
> problem!  TAR does not have this problem when not run by amanda!
> 
> Any code hacking needs to occur within amanda's source, to determine
> exactly what it really is doing.


My style is just as grand.  I'm too missing the point.

Assuming tar is actually called with the logged arguments,
why do you insist it is some amanda code that is the problem?
Oh, I know, because when you run it manually it works.

You may be missing the point that there is a third component
in this equation, the environment.  The environment could be
different when you login and do things manually than when
inet invokes things non-interactively.

I'm not talking about just a few exported variables, but other
settings may be influencing things.  Just as one example, if I recall
correctly, there is a cygwin environment setting that affects some
aspects of cygdrive pathname resolution.  Others affect whether file
openings automatically handle the line ending differences between windows
and unix and still others affect how filename case differences and treated.

If tar is being called with the same arguments you used manually, then
I'd be looking at things influencing the tar run other than amanda.
Once tar is called, amanda has no other involvement in its running.

-- 
Jon H. LaBadie                  jon AT jgcomp DOT com
 JG Computing
 4455 Province Line Road        (609) 252-0159
 Princeton, NJ  08540-4322      (609) 683-7220 (fax)