Re: [PATCH 9/9] new: don't allow new tests in group 'other'

["Darrick J. Wong" <djwong@kernel.org>]

On Thu, Sep 16, 2021 at 09:40:54AM +0300, Amir Goldstein wrote:
> On Thu, Sep 16, 2021 at 2:43 AM Darrick J. Wong <djwong@kernel.org> wrote:
> >
> > From: Darrick J. Wong <djwong@kernel.org>
> >
> > The 'other' group is vaguely defined at best -- other than what?  It's
> > not clear what tests belong in this group, and it has become a dumping
> > ground for random stuff that are classified in other groups.  Don't let
> > people create new other group tests.
> >
> > Signed-off-by: Darrick J. Wong <djwong@kernel.org>
> > ---
> >  new |    7 +++++--
> >  1 file changed, 5 insertions(+), 2 deletions(-)
> >
> >
> > diff --git a/new b/new
> > index 6b7dc5d4..5cf96c50 100755
> > --- a/new
> > +++ b/new
> > @@ -96,9 +96,9 @@ then
> >
> >      while true
> >      do
> > -       echo -n "Add to group(s) [other] (separate by space, ? for list): "
> > +       echo -n "Add to group(s) [auto] (separate by space, ? for list): "
> >         read ans
> > -       [ -z "$ans" ] && ans=other
> > +       [ -z "$ans" ] && ans=auto
> >         if [ "X$ans" = "X?" ]
> >         then
> >             echo $(group_names)
> > @@ -109,6 +109,9 @@ then
> >                 echo "Invalid characters in group(s): $inval"
> >                 echo "Only lower cases, digits and underscore are allowed in groups, separated by space"
> >                 continue
> > +           elif echo "$ans" | grep -q -w "other"; then
> > +               echo "Do not add more tests to group \"other\"."
> > +               continue
> 
> Should we also filter out "other" from group_names(), so it is not listed
> for "?"?

No; there are drawbacks to that, as you point out below.

> With this patch, "other" does not emit a warning when passed in as a script
> command line argument.

Done.

> If we filter "other" from group_names(), then the warning in "expert mode"
> will be a bit confusing (group "other" not defined in documentation).

I will filter it out in the specific case case that the interactive user
specified "?" to list the groups.

> Also, it is not clear to me if this is intentional behavior that interactive
> mode allows non-dcumented groups (with valid chars validation) and
> expert mode does not allow non-documented groups?

Probably not.

> It may be simpler to use the same helper in both modes (is_group_valid)
> to emit the correct warning and either proceed (expert mode) or get
> back to prompt (interactive mode).

This is getting farther afield from where I wanted this thing to go.
Very well, I'll split the ./new cleanups into its own series, but TBH
I've gotten tired of people asking for more and more cleanups out of me.

--D

> Thanks,
> Amir.