[PATCH] Second part of bugfix for bug #13319 (round-tripping ACE's through vfs_fruit).

[Jeremy Allison]

On Fri, Mar 16, 2018 at 07:55:04PM +0100, Ralph Böhme wrote:
> On Fri, Mar 16, 2018 at 11:47:09AM -0700, Jeremy Allison wrote:
> > On Fri, Mar 16, 2018 at 07:34:54PM +0100, Ralph Böhme wrote:
> > > Hi Jeremy,
> > > 
> > > On Thu, Mar 15, 2018 at 03:28:03PM -0700, Jeremy Allison wrote:
> > > > On Thu, Mar 15, 2018 at 06:15:04PM -0400, jim via samba-technical wrote:
> > > > > Should vfs_fruit.c debug message use DEBUG_WARNING instead of DEBUG(1, per
> > > > > replaced code?
> > > > 
> > > > Yes, that's correct, I missed that. Updated patch attached.
> > > 
> > > thanks! Lgtm. Sorry for not catching this in the initial review.
> > 
> > Is that lgtm a RB+ ? Just checking :-).
> 
> Not sure. :)
> 
> > > Once question: why do you add a new share and test-suite for the test? The
> > > option is enabled by default, so we shouldn't need this unless you wanted to
> > > make this explicit in case the default changes.
> > 
> > The default share doesn't have xattr_tdb on the end of
> > the vfs modules. It it gets run on a filesystem that
> > doesn't support native xattrs the test fails (took
> > me a long time to figure out why running it manually
> > worked - ext4, vs running make test failed - tmpfs :-).
> 
> Hm, I guess that is just a mistake on my side and we should add xattr_tdb to all
> fruit shares. I can prepare a patch if you like.

Yeah, but that changes all the existing tests. I really
didn't want to do that (I didn't want to get into a rathole
of debugging stuff that was nothing to do with my changes :-).

> > I didn't want to add that to the "standard" vfs_fruit
> > share just in case this was intentional or caused
> > other tests to fail.
> 
> Would be it possible to just use the existing shares given they use xattr_tdb?
> I'd like to avoid share and test-suite bloat...

Why is adding a new share definition a concern ?

It doesn't slow anything down or use more space (it's
re-using the same share path).

Let's not make adding tests harder than it already is :-).

I prefer to make new tests very isolated, as I've been
burned in the past with "just re-use this existing test"
etc. etc. leading to days debugging why the small change
broke a bunch of unrelated stuff, only to find it wasn't
related at all (aka see the xattr_tdb example :-).

Jeremy.