diff-sync.md

Since the reMarkable 2.10 release, there is a new sync method available: It collects only differences instead of uploading the whole document. So it takes less time to upload/download large modified documents, and it handles edition conflicts.

!!! note This feature is available since rmfakecloud v0.0.5.

In order to use this feature, you'll need to update your user to activate it:

rmfakecloud setuser -u ddvk -a -s

The -a is to let/set the user admin: as the tool will remove the admin permission if not set.

You'll then need to reconnect on your device to apply the settings, and a full resync will automatically begin.

Deal with file lost

There was an issue with rmapi affecting both official cloud and rmfakecloud, where files were randomly lost.

The files were not deleted on rmfakecloud side, it was possible to relink them in the real document tree.

Two utility have been designed to help to relink:

• history2git15: it creates a git repository of the states of the tree, creating a commit for each generation/tree modification.
• relinkfile15: given an user and an root index name (it's sha256 hash as seen in .root.history or as commit subject of history2git15), it'll relink all files given as argument in their state at the time of the given root index.

!!! warning Before using those utilities, please shutdown your rmfakecloud instance to avoid concurrent modifications and possible corruptions.

history2git15

Before using history2git15, you need to install git on your system.

This utility takes as argument a path to a .root.history file. It'll create an history directory, which will contain the git repository.

./history2git15 /var/lib/rmfakecloud/users/ddvk/sync/.root.history

As it can takes a large amount of time, you can limit to the latest modifications with the -tail option:

./history2git15 -tail 20 /var/lib/rmfakecloud/users/ddvk/sync/.root.history

In this example only the last 20 modifications of the tree will be saved as a git repository.

After a successfull run, go to /var/lib/rmfakecloud/users/ddvk/sync/history and use git to explore the differences referenced in the two files:

• doctree: this is a human readable hiereachy of the directories and files, along with the date of the last modification.
• tree: this is the whole tree state with all metadata accessible, in JSON format.

relinkfile15

!!! warning Be sure that rmfakecloud is stopped before using this command. Bad things will happen if some devices performs synchronization while relinkfile15 works.

This utility can relink in the root tree a non-deleted file.

DATA_DIR=/var/lib/rmfakecloud ./relink15 -user ddvk -root-hash 1c0ee6fb7fde7d09dd25b954dd9f23f950d9e25f1fbc661ca18aebf40bb14a00 "Notebook 42" "My calendar.pdf"

DATA_DIR is the same option used by rmfakecloud. If you don't use it with rmfakecloud, make sure you are in the same directory as when you start rmfakecloud: you should have a data/ directory.

The -user option is the registration address used (or the name of the directory inside data/users).

The -root-hash is the name of the file to use as the old root index.

The rest of the arguments given to the command line are the name of the file as given by history2git15 in doctree (or as visibleName in tree). This doesn't handle directories hierarchy at the time of writing: it only matches the filename. You need to make sure the parent directory still exists (as we said the hierarchy is handle by metadata, not by indexes).

How does it work?

All files are saved with a sha256 name in the user sync directory. File can be raw pages, document metadata, directory metadata, indexes, ...

Each file is saved along with a generation number. Each modification increment this generation number.

Root tree file

All files in the tree are indexed in a root index. The index references all the index to others files and directories.

The hierarchy is given by each file metadata. The root index file doesn't handle file hierarchy.

Current root file in use

It is possible to retrieve the current root index in use by consulting the content of the file root. It contains the name of the file containing the root index.

History and generation

Previous root indexes are kept in the directory and can be listed with the file .root.history. Each line corresponds to a given generation (line 1 = generation 1, ...) and contains the date of the modification and the corresponding filename.

File indexes

Each file, as seen on the tablet, is split in several parts (metadata, raw pages, cached transcripted text, content, ...). File indexes store the references of each sub-file.

Files on the tablet are referenced by an UUID. On rmfakecloud files parts are also stores with sha256 names.