bug#14388: Bug in uname command - reg...

[Bob Proulx]

Linda Walsh wrote:
> Bob Proulx wrote:
> >   $ uname -r
> >   3.2.0-4-amd64
> >   $ uname -v
> >   #1 SMP Debian 3.2.35-2
>
> From that -- it looks quite a bit like it describing version information.
> On mine, I have
> uname -r=3.9.0-Isht-Van
> 
> With Isht-Van being my localversion name
> The uname-v has:
> #6 SMP PREEMPT Wed May 8 17:28:40 PDT 2013
> Compile#, options date.... not sure why debians is different --- 3.2.35-2?
> 
> That's odd looking.

IBM AIX 3.5 returns information like this:

  $ uname -a
  AIX localhost 3 5 00CD7EAF4C00

  $ uname -r
  3

  $ uname -v
  5

HP-UX 11.31 will say something like this:

  $ uname -a
  HP-UX localhost B.11.31 U ia64 3682977672 unlimited-user license

  $ uname -r
  B.11.31

  $ uname -v
  U

Solaris 5.9:

  $ uname -a
  SunOS localhost 5.9 Generic_112233-12 sun4u sparc SUNW,Sun-Fire-15000

  $ uname -r
  5.9

  $ uname -v
  Generic_112233-12

I don't have access to a variety of machines these days so the above
is cutting and pasting from reports sent in that I found on the web.
I may have something slightly wrong in the above.  But it should still
be illustrative that every system interprets what to do with the
fields differently.

> It looks like they are a bit backwards -- but I think I would agree with
> Bob's assessment that it can't be something that is fixed since there
> are tons of programs (including programs IN THE KERNEL (to load modules))
> that use the 'release' to get the linux-version info...
> 
> *doh!*

Yes.  D'oh!  Which is why it isn't as useful as people think it might be.

> > Really the only portable way to use uname(1) is to call it first to
> > see which system name it returns and then after knowing the system
> > type then call it again with whatever options make sense on that
> > system.
>
>       Which is a bit like saying you need an interpreter that's
> system specific to pick out what is relevant.  That's special! ;->

Yes.  That is exactly what I am saying! :-)

Here is some live code from a script designed to run in a an
environment with all of the above.  This isn't general purpose and was
written to work exactly for the cases I needed it to work.  For
example for AIX I would have needed to have it combine -r and -v
together to get to 3.5 but for me knowing "aix" was good enough so I
stopped there.  I am just posting this to give an exapmle of the type
of system specific tests that are needed.  I am sure that if this
script were live working today that it would have variations for
current systems.

Bob

os=unknown
mach=unknown
uname=$(uname)
if [ $? -ne 0 ] || [ -z "$uname" ]; then
  echo "Error: Could not run 'uname'" 1>&2
  exit 1
fi
case $uname in
AIX)
  os=aix
  mach=rs6000
  ;;
HP-UX)
  case $(uname -m) in
  9000/*)
    case $(uname -r) in
      ?.10.*) mach=hppa1.1 ;;
      *)      mach=hppa2.0 ;;
    esac
    ;;
  *) mach=$(uname -m) ;;
  esac
  case $(uname -r) in
    ?.10.*)  os=hpux10 ;;
    *) os=hpux$(uname -r | sed 's/^[AB]\.//') ;;
  esac
  ;;
Linux)
  os=gnulinux
  mach=$(uname -m)
  ;;
esac