Re: [PATCH] df (without any params) omits the root filesystem line

[Pádraig Brady]

On 01/28/2013 11:00 AM, Bernhard Voelker wrote:
> Hi Padraig,
>
> On 01/27/2013 08:14 PM, Pádraig Brady wrote:
> > On 01/27/2013 05:51 PM, Bernhard Voelker wrote:
> > > Thanks. Pushed:
> > > http://git.sv.gnu.org/cgit/coreutils.git/commit/?id=bb116d35
>
> > Looking at this in detail is still at the end of my TODO list.
>
> Sorry for pushing early.

No worries.

> > udev is a separate file system here, which now seems not displayed?
>
> I've also noticed that.
> This is because devtmpfs is marked as dummy in ME_DUMMY_0 in
> mountlist.c. It has been added in December and came into
> coreutils with the latest gnulib pull:
> http://git.sv.gnu.org/cgit/gnulib.git/commit?id=40f82767

Ah right, I was worried it was an unexpected issue with the
patch causing the change.

As for gnulib though, I'm not 100% sure devtmpfs should be marked
as a dummy file system, as there is storage associated with it.
I.E. you can write normal files to /dev and space will be consumed.

Note df already excluded all these new dummy file systems
by default by checking the size like:

  if (fsu.fsu_blocks == 0 && !show_all_fs && !show_listed_fs)
    return;

> > Also do we need to treat rootfs specially in the code now?
>
> Hmm, as a rootfs mount entry will loose against the real "/" entry
> comes later, the special handling could be removed again ... but
> this would change the behavior for `df -trootfs`, as it wouldn't
> be shown there. On the other side, `df -tdevtmpfs` doesn't show
> anything either, i.e. for dummy file systems the user would always
> have to use the -a option. And this is nothing new.
>
> Therefore, I'm inclined to make df handle rootfs like any other
> dummy FS and remove the special treatment as shown in the attachment.
> WDYT?

Yes I agree this is better and more consistent.

It's a bit weird though that `df -t devtmpfs` won't show
a devtmpfs file system unless -a is specified.
Maybe -t should imply -a?

thanks!
Pádraig.