Re: FUSE not work.

On Fri, Jul 19, 2013 at 09:42:41PM -0700, Kevin Oberman wrote:
> On Fri, Jul 19, 2013 at 6:22 PM, Pedro Giffuni <pfg_at_freebsd.org> wrote:
> 
> > On 19.07.2013 12:07, Alexander Panyushkin wrote:
> >
> >> 19.07.2013 19:42, Pedro Giffuni пишет:
> >>
> >>> (re-posting since the original response didn't make it through)
> >>>
> >>> On 19.07.2013 10:29, Alexander Panyushkin wrote:
> >>>
> >>>> 19.07.2013 17:36, Pedro Giffuni пишет:
> >>>>
> >>>>> Hello;
> >>>>>
> >>>>> The internal data structures in the FUSE kernel module have been
> >>>>> updated to be more compatible with MacFUSE and the linux FUSE.
> >>>>>
> >>>>> This basically means you have to rebuild your fuse modules
> >>>>> (including NTFS).
> >>>>>
> >>>>> We should not really update the FUSE_KERNEL_MINOR_VERSION yet.
> >>>>> I will add a note to UPDATING.
> >>>>>
> >>>>> Pedro.
> >>>>>
> >>>>
> >>>>
> >>>>
> >>>> Unfortunately the port is not  build in FreeBSD 10
> >>>>
> >>>>
> >>>> ===>>> Starting build for sysutils/fusefs-kmod <<<===
> >>>>
> >>>>
> >>> fusefs-kmod is not needed on -current. I meant you have to rebuild
> >>> and reinstall fusefs-* ; in particular fusesfs-libs and ntfs.
> >>>
> >>> Since this is a ports issue I will not update the UPDATING file.
> >>>
> >>> Developers that need to know should use  __FreeBSD_version
> >>> 1000038 as a reference.
> >>>
> >>>
> >> I rebuild sysutils/fusefs-ntfs with Clang and GCC, but this has no effect.
> >>
> >> sysctl kern.osreldate
> >> kern.osreldate: 1000037
> >>
> >> ntfs-3g /dev/ad10s1 /mnt/ntfs_0
> >> mount
> >> ...
> >> /dev/fuse on /mnt/ntfs_0 (fusefs, local, synchronous)
> >>
> >> ls -la /mnt/ntfs_0
> >> unique: 18, opcode: GETATTR (3), nodeid: 1, insize: 40, pid: 47053
> >> getattr /
> >>    unique: 18, success, outsize: 112
> >> fuse: writing device: Invalid argument
> >> ls: /mnt/ntfs_0: Input/output error
> >>
> >> fuse is at least partly broken in -current. I discovered that if you
> mount two devices that use fuse, when you umount any of them, ll are marked
> as not mounted and disappear from df(1) or mount(8) output, but only one is
> actually cleanly unmounted and, if it has been written to, will likely be
> corrupt. This applies to at least ntfs.
> 
> ATM no one is supporting fuse in current, so this may become a serious
> issue before long.

This is fixed now.

btw this problem is/was not related to fuse on current but rather to
fusefs-libs.

Bapt