Before this change, the VFS layer did not properly handle unicode normalization,
which caused problems particularly for users of macOS. While attempts were made
to handle it with various `-o modules=iconv` combinations, this was an imperfect
solution, as no one combination allowed both NFC and NFD content to
simultaneously be both visible and editable via Finder.
After this change, the VFS supports `--no-unicode-normalization` (default `false`)
via the existing `--vfs-case-insensitive` logic, which is extended to apply to both
case insensitivity and unicode normalization form.
This change also adds an additional flag, `--vfs-block-norm-dupes`, to address a
probably rare but potentially possible scenario where a directory contains
multiple duplicate filenames after applying case and unicode normalization
settings. In such a scenario, this flag (disabled by default) hides the
duplicates. This comes with a performance tradeoff, as rclone will have to scan
the entire directory for duplicates when listing a directory. For this reason,
it is recommended to leave this disabled if not needed. However, macOS users may
wish to consider using it, as otherwise, if a remote directory contains both NFC
and NFD versions of the same filename, an odd situation will occur: both
versions of the file will be visible in the mount, and both will appear to be
editable, however, editing either version will actually result in only the NFD
version getting edited under the hood. `--vfs-block-norm-dupes` prevents this
confusion by detecting this scenario, hiding the duplicates, and logging an
error, similar to how this is handled in `rclone sync`.
Before this change it wasn't possible to see where transfers were
going from and to in core/stats and core/transferred.
When use in rclone mount in particular this made interpreting the
stats very hard.
Before this change the VFS cache could get into a state where when an
object was updated remotely, the fingerprint of the item was correct
for the new object but the data in the VFS cache was for the old
object.
This fixes the problem by updating the fingerprint of the item at the
point we remove the stale data. The empty cache item now represents
the new item even though it has no data in.
This stops the fallback code for an empty fingerprint running (used
when we are writing items to the cache instead of reading them) which
was causing the problem.
Fixes#6053
See: https://forum.rclone.org/t/cached-webdav-mount-fingerprints-get-nuked-on-ls/43974/
Before this change if a backend can't upload 0 length files and
`--vfs-cache-mode writes` was in use then the writeback logic would
try to upload the 0 length file forever.
This change causes it to exit on the first failure to upload.
Summary:
In cases where cmount is not available in macOS, we alias nfsmount to mount command and transparently start the NFS server and mount it to the target dir.
The NFS server is started on localhost on a random port so it is reasonably secure.
Test Plan:
```
go run rclone.go mount --http-url https://beta.rclone.org :http: nfs-test
```
Added mount tests:
```
go test ./cmd/nfsmount
```
billy defines a common file system interface that is used in multiple go packages.
vfs.Handle implements billy.File mostly, only two methods needed to be added to
make it compliant.
An interface check is added as well.
This is a preliminary work for adding serve nfs command.
A subtle bug where dir modification time is not updated when the dir already exists
in the cache. It is only noticeable when some clients use dir modification time to
invalidate cache.
Name() method was originally left out and defaulted to the base
class which always returns empty. This trigerred incorrect behavior
in serve nfs where it relied on the Name() of the interafce to figure
out what file it was modifying.
This method is copied from RWFileHandle struct.
Added extra assert in the tests.
This almost 100% backwards compatible. The only difference being that
in the rc options/get output CacheMode will be output as strings
instead of integers. This is a lot more convenient for the user. They
still accept integer inputs though so the fallout from this should be
minimal.
This adds an additional parameter to the creation of each flag. This
specifies one or more flag groups. This **must** be set for global
flags and **must not** be set for local flags.
This causes flags.md to be built with sections to aid comprehension
and it causes the documentation pages for each command (and the
`--help`) to be built showing the flags groups as specified in the
`groups` annotation on the command.
See: https://forum.rclone.org/t/make-docs-for-mortals-not-only-rclone-gurus/39476/
This changes hasVirtual to an atomic struct variable that's updated on
add or delete from the virtual map.
This keeps it up to date and avoids deadlocks.
Signed-off-by: Anagh Kumar Baranwal <6824881+darthShadow@users.noreply.github.com>
This empties the directory cache after twice the directory cache
period to release memory.
Signed-off-by: Anagh Kumar Baranwal <6824881+darthShadow@users.noreply.github.com>
Before this change if the VFS took more than 5 to initialise (which
can happen if there is a lot of files or a lot of files which need
uploading) the backend was dropped out of the cache before the VFS was
fully created.
This was noticeable in the dropbox backend where the batcher Shutdown
too soon and prevented further uploads.
This fixes the problem by Pinning backends before the VFS cache is
created.
https://forum.rclone.org/t/if-more-than-251-elements-in-the-que-to-upload-fails-with-batcher-is-shutting-down/38076/2
Before this fix, when a write to a read only directory failed, rclone
would leav spurious directory entries in the directory.
This confuses `rclone serve webdav` into giving this error
http: superfluous response.WriteHeader
This fixes the VFS layer to remove any directory entries where the
file creation did not succeed.
Fixes#5702
This error happened on a restart of the VFS with files to upload into
a new directory on a bucket based backend. Rclone was assuming that
directories created before the restart would still exist, but this is
a bad assumption for bucket based backends which don't really have
directories.
This change creates the pretend directory and thus the directory cache
if the parent directory does not exist when adding a virtual on a
backend which can't have empty directories.
See: https://forum.rclone.org/t/that-pesky-failed-to-reload-error-message/34527
Before this change if a file was uploaded to a backend which didn't
support modtimes, the time of the file read after the upload had
completed would change to the time the file was uploaded on the
backend.
When using `--vfs-cache-mode writes` or `full` this time would be
different by the `--vfs-write-back` delay which would cause
applications to think the file had been modified.
This changes uses the last modification time read by the OS as a
virtual modtime for backends which don't support setting modtimes. It
does not change the modtime to that actually uploaded.
This means that as long as the file remains in the directory cache it
will have the expected modtime.
See: https://forum.rclone.org/t/saving-files-causes-wrong-modified-time-to-be-set-for-a-few-seconds-on-webdav-mount-with-bitrix24/36451
Before this change, if you renamed a directory containg files yet to
be uploaded then deleted the directory the files would still be
uploaded.
This fixes the problem by changing the directory path in all the file
objects in a directory when it is renamed. This wasn't necessary until
we introduced virtual files and directories which lived beyond the
directory flush mechanism.
Fixes#6809
Before this fix, opening a file with `O_CREATE|O_RDONLY` caused an IO error to
be returned when using `--vfs-cache-mode off` or `--vfs-cache-mode writes`.
This was because the file was opened with read intent, but the `O_CREATE`
implies write intent to create the file even though the file is opened
`O_RDONLY`.
This fix sets write intent for the file if `O_CREATE` is passed in which fixes
the problem for all the VFS cache modes.
It also extends the exhaustive open flags testing to `--vfs-cache-mode writes`
as well as `--vfs-cache-mode full` which would have caught this problem.
See: https://forum.rclone.org/t/i-o-error-trashing-file-on-sftp-mount/34317/
Before this patch a deadlock could occur if the cache cleaner was
running when an object upload finished.
This fixes the problem by delaying marking the object as clean until
we have notified the VFS layer. This means that the cache cleaner
won't consider the object until **after** the VFS layer has been
notified, thus avoiding the deadlock.
See: https://forum.rclone.org/t/rclone-mount-deadlock-when-dir-cache-time-strikes/33486/
Before this change we truncated files in the backing store regardless
of whether we needed to or not.
After, we check to see if the file is the right size and don't
truncate if it is.
Apparently Windows Defender likes to check executables each time they
are modified, and truncating a file to its existing size is enough to
trigger the Windows Defender scan. This was causing a big slowdown for
operations which opened and closed the file a lot, such as looking at
properties on an executable.
See: https://forum.rclone.org/t/for-mount-sftp-why-right-click-on-exe-file-is-so-slow-until-it-freezes/33830
Before this change if we copied files of unknown size, then they lost
their metadata.
This was particularly noticeable using --s3-decompress.
This change adds metadata to Rcat and RcatSized and changes Copy to
pass the metadata in when it calls Rcat for an unknown sized input.
Fixes#6546
Before this change the VFS cache cleaner would loop indefinitely while
the cache was above quota. This used up all the CPU.
This fix prevents the cache cleaner from looping. It will be kicked on
ENOSPACE and run in its scheduled time otherwise so this should be
sufficient.
See: https://forum.rclone.org/t/vfs-keeps-checking-same-files/32120
