Re: [Bug-tar] gnu tar: patch/featureadd

[Paul Eggert]

Sergey Poznyakoff <address@hidden> writes:

> >  tar -DSPc /dev/hdb[1-2] | cdrecord -dev=0,0,0
> 
> is equivalent to
> 
> (dd if=/dev/hdb1; dd if=/dev/hdb2) | tar cSP | cdrecord -dev=0,0,0

I don't see the equivalence here.  "tar cSP" isn't right, since tar
will write "Cowardly refusing to create an empty archive."  I don't
see an easy way to get the proposed functionality with existing tools,
so it appears to me that the suggestion is a good one -- though perhaps
I'm missing something too.

Andrew, would you be willing to assign the copyright to the Free
Software Foundation, so that we could install it in GNU tar?  I can't
yet guarantee that we'll install it, but I know that we can't install
it unless the copyright issues are resolved.

Assuming that problem can be resolved, the proposed patch is
incomplete.  Please submit it in "diff -pu" format, with a proposed
ChangeLog entry for it; see
<http://mail.gnu.org/archive/html/bug-gnulib/2003-07/msg00124.html>
for an example.  Please use the GNU coding standards for code that you
write in the patch.  Please check every system call to see whether it
fails, and report an error if a failure occurs.  You can look at the
current code for how this is done.

Also, let's not use a short option like -D for this.  A long
option is good enough.  We're running out of short options.

Finally, why should block special devices be treated any differently
from character special devices?  Shouldn't the option be called
something like --device-contents instead, and affect both block and
character special devices?