[Toybox] [PATCH] du.test: skip on Mac.

[Rob Landley]

On 11/22/19 2:38 PM, enh via Toybox wrote:
> None of the current tests are relevant on the Mac because small
> symlinks are inlined into inodes, as are empty directories, so
> everything's using zero blocks.

That's a filesystem issue, not an OS issue. (In theory Linux filesystems can do
this too. I remember murderfs did tricks like this, but nobody uses it anymore
because of the murder part. No idea what butterfs is doing.)

There's a similar issue with test_chattr, getfattr, and setfattr. They were all
kinda ext2-isms which were retained in ext3 but not ext4. (And then _partially_
genericized but there's an API migration that goes along with that.)

So the right thing to do (in the absence of a generic way to test filesystem
capabilities) is to test for a recognized filesystem type that supports them, I
guess?

But although you can "df ." and then grep that partition name in /proc/mounts to
get ext4, "stat -fc %T" says ext3 for ext4, which is sad. And I see the
implementation isn't _checking_ which driver is used by the live mounted
filesystem, instead the blkid tests have been block copied into a second place
so that if they ever _are_ updated, they'll need updating twice. Sigh. No, it's
worse: blkid has an ext4 test now.)

(All this is on the todo list, although the note-to-self in this case is just
"each tests/*.test file that isn't chmod +x has a reason why" and the rest is
just stuff I remember needing to do...)

Rob