Re: Various tests give illusory results

[Ondrej Oprala]

Well, I wanted to satisfy the requirements from comment #6 as much as 
possible
and didn't want to change anything in the existing tests (such as 
passing args to require_root_ )
to make it more automatic. If the check were to act on setuidgid output 
like in the rm test, it'd either
have to be changed in all tests or require_root_ would need additional 
arguments (which are again
changes in the tests). But if it needs to be changed that way, I will, 
of course, redo it.
Cheers,
  Ondrej.

On 08/10/2012 08:53 AM, Bernhard Voelker wrote:
> On 08/09/2012 05:43 PM, Ondrej Oprala wrote:
> > Hi, I think I got a fix for this bugzilla
> > https://bugzilla.redhat.com/show_bug.cgi?id=556358.
> > I added a bit of permission checking to require_root_ so no tests have
> > to be rewriten.
> > Have a nice day :) ,
> >    Ondrej
> Hi Ondrej,
>
> > +setuidgid_has_perm_()
> > +{
> > +
> > +  cat << \EOF > cmds.tmp
> > +  IFS=:
> > +  for DIR in $PATH; do
> > +    test -x $DIR || exit 1
> > +  done
> > +  exit 0
> > +EOF
> > +
> > +  su -s /bin/sh $NON_ROOT_USERNAME < cmds.tmp
> > +
> > +  RET=$?
> > +  return $RET
> > +}
> > +
> just a thought: if setuidgid is part of the test failure,
> then why not using setuidgid in here?
>
> Furthermore: the problem is finding the correct binary, right?
> E.g. in tests/rm/fail-2eperm there is already such a test:
>
>    # Try to ensure that $NON_ROOT_USERNAME can access
>    # the required version of rm.
>    rm_version=$(
>      setuidgid $NON_ROOT_USERNAME env PATH="$PATH" rm --version |
>      sed -n '1s/.* //p'
>    )
>    case $rm_version in
>      $PACKAGE_VERSION) ;;
>      *) skip_ "cannot access just-built rm as user $NON_ROOT_USERNAME";;
>    esac
>
>
> Have a nice day,
> Berny