Using vfs_fruit alongside vfs_recycle

[Isaac Boukris]

Hi Ralph,

On Fri, Mar 6, 2020 at 12:25 PM Ralph Boehme <slow at samba.org> wrote:
>
> Am 3/6/20 um 12:16 PM schrieb Ralph Boehme via samba-technical:
> > Am 3/6/20 um 12:02 PM schrieb Isaac Boukris via samba-technical:
> >> Hi,
> >>
> >> The manpage of vfs_fruit states that "This module is not stackable
> >> other than described in this manpage.", is this still correct?
> >
> > well, everything is is simply not tested systematically. Many
> > combinations may work, with some you may run into issues...
> >
> >> I'm looking into a bug report which I can reproduce on master, where
> >> usage of vfs_fruit alongside vfs_recycle causes an error log at level
> >> 0 upon deleting a file, failing to find "._filename". So I wonder if
> >> this configuration is unsupported or is there a logical bug here.
> >
> > Iirc I've never worked with systems where vfs_recycle was configured
> > with fruit, so I haven't seen this before.
> >
> >> [2020/03/06 12:28:48.627691,  0]
> >> ../../source3/modules/vfs_recycle.c:239(recycle_get_file_size)
> >>   recycle: stat for ./._test.txt returned No such file or directory
> >
> > Hm, maybe just a too high log level in vfs_recycle?
>
> Yeah, looking at the codepath in vfs_recycle that leads up to this
> message it is going via unlink -> recycle_get_file_size. vfs_recycle is
> assuming that any path passed to unlink must exist, otherwise it logs
> the error.

You can see debug level 10, in the redhat bug i linked above (comment
13). Couldn't follow where the "._" prefix comes from.

> Attached patch should fix it. Can you file a bugreport?

https://bugzilla.samba.org/show_bug.cgi?id=14316

Thanks!