Re: df: --inodes versus --portability mode

[Pádraig Brady]

On 09/18/2012 11:54 AM, Bernhard Voelker wrote:
> On 09/18/2012 12:04 PM, Pádraig Brady wrote:
> > On 09/18/2012 10:47 AM, Bernhard Voelker wrote:
> > > For the implementation of the new --output option, and the cleanup
> > > necessarily coming with it, I'd like to fix that so that a -P will
> > > override a previous -i and vice versa:
> > >
> > >     $ src/df -iP | head -n 2
> > >     src/df: warning: option '-P' is overriding the previous mode
> > >     Filesystem     1024-blocks      Used Available Capacity Mounted on
> > >     rootfs            12095032   7515424   3965208      66% /
> > >
> > >     $ src/df -Pi | head -n 2
> > >     src/df: warning: option '-i' is overriding the previous mode
> > >     Filesystem       Inodes   IUsed    IFree IUse% Mounted on
> > >     rootfs           768544  237523   531021   31% /
> > >
> > > WDYT?
> >
> > I don't like order being significant.
>
> Good point.
>
> > du -i is an extension to POSIX, so I think it's fine for -i to override.
>
> s/du/df/ i guess ;-)
>
> > -P essentially meant don't wrap but that's moot now,
>
> That was actually my question: -P means "default mode" in
> the first place, i.e. one could read it as "print block size".
> And now comes -i ...
>
> > but I'd keep it as is for backwards compat.
>
> ... another good point - and I think I should stick with it.
> So as -P does nothing, shouldn't df print a warning that
> it disregards it when -i is in effect?

Well as it used to do something (ensure single lines)
so I was going to suggest not printing a warning,
for a while at least.  Though I now notice that BSD also
implements -i, and that adds the inodes columns _in addition_
to the blocks columns. So it's more likely that users
may be confused about this combination, so I guess a warning
is warranted. Note we couldn't change to the BSD -i behavior
for backward compat reasons.

> I thought that I could get rid of the mashing up of the scale
> factor (human_output_opts, output_block_size) and df's main
> modes DEFAULT_MODE, INODES_MODE, POSIX_MODE and HUMAN_MODE)
> somehow ... because the same question will arise again with
> 'df -P --out=...' ;-/
> IMHO the cleanest solution would be to make the main modes
> -i, -p and the new --o mutually exclusive.

Yes --o should be mutually exclusive with the -i and -P options.

cheers,
Pádraig.