disorderfs: FUSE file-deletion quirk

James Addison jay at jp-hosting.net
Wed Jan 22 13:18:45 UTC 2025


Thanks, Holger,

On Wed, 22 Jan 2025 at 10:06, Holger Levsen <holger at layer-acht.org> wrote:
> [ ... snip ... ]
> On Tue, Jan 21, 2025 at 10:17:35PM +0000, James Addison via rb-general wrote:
> > [...]  Currently my
> > preference is leaning towards suggesting that disorderfs should opt-in
> > by default.
>
> I tend to agree: for fuse it's probably a sensible default to not
> have that feature enabled and for reprotest it's probably sensible
> to expect it's enabled (or rather that a fs behaves as an fs), so
> I think disorderfs should have this enabled and act as a normal fs
> here.

Ok; I'll file a bugreport in disorderfs noting the problem and with a
small patch (and test case) I've developed to change and confirm the
expected behaviour.

During creation of the test case, I uncovered an important detail,
likely explaining why we haven't encountered this problem more often:
the problem only occurs for files that are currently open (e.g. for
reading/writing) when they are removed.

Regards,
James


More information about the rb-general mailing list