On Tue, Jan 16, 2024 at 08:37:21AM -0500, Brian Foster wrote: > On Sun, Jan 14, 2024 at 12:41:20PM +0800, Su Yue wrote: > > mkfs.bcachefs supports force overwrite when option '-f' is given: > > $ mkfs.bcachefs --help | grep force > > -f, --force > > After this commit, MKFS_BCACHEFS_PROG will contains ' -f' so > > we don't have to add '-f' to $MKFS_OPTIONS manually. > > > > It also fixes generic/466 which unsets MKFS_OPTIONS causing > > that test hangs in mfks.bcachefs waiting for confirmation of > > the force overwrite. > > > > This seems mostly reasonable to me, but I'm kind of wondering why I > haven't had to add -f to my MKFS_OPTIONS for bcachefs. Is there > something unique to the test environment that affects this? I.e., I > don't see any issue with generic/466 on a recent Fedora, even if I > pre-format the scratch dev with XFS. I think there's still multiple screwy things going on with blkid probing and superblock wiping - I've been noticing that there's a bunch of ktest tests that don't specify -f and work fine in the CI, but that fail when I loop them locally. It appears wipefs only zeroes out the magic number on our first superblock - but it leaves the layout intact before the first superblock, so we're still able to find our backup superblocks.
