I'm using 2.4.5p1 on NetBSD with Kerberos encryption and
authentication.

I tried to verify some tapes and found that 'gzip -t' failed on the
restored files.  On investigation, after adding some better
diagnostics to gzip (NetBSD's own), I found that the problem was that
the last 32K block was padded with zeros.

Unflushed dumps in the holding directory have this problem for remote
dumps (krb encrypted), but not local ones.

On an older amanda install, not using krb4, I don't have this problem.

Is anyone else seeing this?

Does GNU gzip ignore trailing NULLs?   The NetBSD implementation goes
back to looking for MAGIC0 after reading the saved length.

-- 
        Greg Troxel <gdt AT ir.bbn DOT com>