Re: [PATCH 6/6] xfs: don't run tests that require v4 file systems when not supported

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

On Wed, Apr 10, 2024 at 06:14:02AM +0200, Christoph Hellwig wrote:
> On Tue, Apr 09, 2024 at 08:56:12AM -0700, Darrick J. Wong wrote:
> > > Signed-off-by: Christoph Hellwig <hch@lst.de>
> > > ---
> > >  common/xfs    | 10 ++++++++++
> > >  tests/xfs/002 |  1 +
> > 
> > Looks fine to me.
> > 
> > >  tests/xfs/045 |  1 +
> > 
> > xfs_db can change uuids on v5 filesystems now, so we don't nee the
> > -mcrc=0 in this test.
> 
> Ok, I'll look into that.
> 
> > Looks fine to me.
> > 
> > >  tests/xfs/148 |  2 ++
> > 
> > I wonder if we could rewrite this test to use either the xfs_db write -d
> > command on dirents or attrs directly; or the link/attrset commands,
> > since AFAICT the dir/attr code doesn't itself run namecheck when
> > creating entries/attrs.
> 
> Can I leave that to you? :)

Yes.

> > >  tests/xfs/158 |  1 +
> > >  tests/xfs/160 |  1 +
> > 
> > inobtcount and bigtime are new features, maybe these two tests should
> > lose the clause that checks that we can't upgrade a V4 filesystem?
> 
> I'll take a look.
> 
> > >  tests/xfs/194 |  2 ++
> > 
> > Not sure why this one is fixated on $pagesize/8.  Was that a requirement
> > to induce an error?  Or would this work just as well on a 1k fsblock fs?
> > 
> > (Eric?)
> 
> I can check if it could be made to work on $pagesize/4, but I'll
> need to defer to Eric if that even makes sense.
> 
> > >  tests/xfs/513 |  1 +
> > 
> > I think we should split this into separate tests for V4/V5 options and
> > only _require_xfs_nocrc the one with V4 options, because I wouldn't want
> > to stop testing V5 codepaths simply because someone turned off V4
> > support in the kernle.
> 
> Ok.
> 
> > >  tests/xfs/526 |  1 +
> > 
> > I'm at a loss on this one -- what it does is useful, but there aren't
> > any V5 mkfs options that conflict as nicely as crc=0 does.
> 
> Yes, I tried to look for conflicting options, but I couldn't find
> anything.  Maybe we'll grow some before the v4 support is retired
> for real :)

Well hilariously just yesterday djwong-wtf just grew one now that you
can't format rtgroups=1 without exchange=1 so I guess there's some hope.

--D

--D