Contents
Preface
•People
•Mailing Lists and Bug Reporting
•Development Status
•Copying
•Acknowledgements
Installation
•Downloading Unison
•Running Unison
•Upgrading
•Building Unison from Scratch
Unix
Windows
Installation Options
Tutorial
•Preliminaries
•Local Usage
•Remote Usage
•Remote Shell Method
•Socket Method
•Using Unison for All Your Files
•Using Unison to Synchronize More Than Two Machines
•Going Further
Basic Concepts
•Roots
•Paths
•What is an Update?
•What is a Conflict?
•Reconciliation
•Invariants
•Caveats and Shortcomings
Reference Guide
•Running Unison
•The .unison Directory
•Archive Files
•Preferences
•Profiles
•Sample Profiles
A Minimal Profile
A Basic Profile
A Power-User Profile
•Keeping Backups
•Merging Conflicting Versions
•The User Interface
•Exit code
•Path specification
•Ignoring Paths
•Symbolic Links
•Permissions
•Cross-Platform Synchronization
•Slow Links
•Fast Update Detection
•Mount Points and Removable Media
•Click-starting Unison
Installing Ssh
•Unix
•Windows
Changes in Version 2.27.57
Unison is a file-synchronization tool for Unix and Windows. It allows
two replicas of a collection of files and directories to be stored on
different hosts (or different disks on the same host), modified
separately, and then brought up to date by propagating the changes in
each replica to the other.
Unison
shares a number of features with tools such as configuration
management packages (
CVS,
PRCS,
etc.),
distributed filesystems
(
Coda,
etc.),
uni-directional mirroring utilities
(
rsync,
etc.),
and other synchronizers
(
Intellisync,
Reconcile,
etc).
However, there are several points where it differs:
-
Unison runs on both Windows (95, 98, NT, 2k, and XP) and Unix (OSX, Solaris,
Linux, etc.) systems. Moreover, Unison works across
platforms, allowing you to synchronize a Windows laptop with a
Unix server, for example.
- Unlike a distributed filesystem, Unison is a user-level program:
there is no need to modify the kernel or to have
superuser privileges on either host.
- Unlike simple mirroring or backup utilities, Unison can deal
with updates to both replicas of a distributed directory structure.
Updates that do not conflict are propagated automatically.
Conflicting updates are detected and displayed.
- Unison works between any pair of machines connected to the
internet, communicating over either a direct socket link or
tunneling over an encrypted ssh connection.
It is careful with network bandwidth, and runs well over slow links
such as PPP connections. Transfers of small updates to large files are
optimized using a compression protocol similar to rsync.
- Unison has a clear and precise specification, described
below.
- Unison is resilient to failure. It is careful to leave the
replicas and its own private structures in a sensible state at all
times, even in case of abnormal termination or communication
failures.
- Unison is free; full source code is available under the GNU
Public License.
Preface
People
Benjamin Pierce leads the
Unison project.
The current version of Unison was designed and implemented by
Trevor Jim,
Benjamin Pierce,
and
Jérôme Vouillon,
with
Alan Schmitt,
Malo Denielou,
Zhe Yang,
Sylvain Gommier, and
Matthieu Goulay.
The Mac user interface was started by Trevor Jim and enormously improved by
Ben Willmore.
Our implementation of the
rsync
protocol was built by
Norman Ramsey
and Sylvain Gommier. It is is based on
Andrew Tridgell's
thesis work
and inspired by his
rsync
utility.
The mirroring and merging functionality was implemented by
Sylvain Roy, improved by Malo Denielou, and improved yet further by
Stéphane Lescuyer.
Jacques Garrigue
contributed the original Gtk version of the user
interface; the Gtk2 version was built by Stephen Tse.
Sundar Balasubramaniam helped build a prototype implementation of
an earlier synchronizer in Java.
Insik Shin
and
Insup Lee contributed design
ideas to this implementation.
Cedric Fournet
contributed to an even earlier prototype.
Mailing Lists and Bug Reporting
Mailing Lists:
Moderated mailing lists are available for bug reporting, announcements
of new versions, discussions among users, and discussions among
developers. See
http://www.cis.upenn.edu/~bcpierce/unison/lists.html for more
information.
Development Status
Unison is no longer under active development as a research
project. (Our research efforts are now focused on a follow-on
project called Harmony, described at
http://www.cis.upenn.edu/~bcpierce/harmony.)
At this point, there is no one whose job it is to maintain Unison,
fix bugs, or answer questions.
However, the original developers are all still using Unison daily. It
will continue to be maintained and supported for the foreseeable future,
and we will occasionally release new versions with bug fixes, small
improvements, and contributed patches.
Reports of bugs affecting correctness or safety are of interest to many
people and will generally get high priority. Other bug reports will be
looked at as time permits. Bugs should be reported to the users list at
unison-users@yahoogroups.com.
Feature requests are welcome, but will probably just be added to the
ever-growing todo list. They should also be sent to
unison-users@yahoogroups.com.
Patches are even more welcome. They should be sent to
unison-hackers@lists.seas.upenn.edu.
(Since safety and robustness are Unison's most important properties,
patches will be held to high standards of clear design and clean coding.)
If you want to contribute to Unison, start by downloading the developer
tarball from the download page. For some details on how the code is
organized, etc., see the file
CONTRIB.
Copying
Unison is free software. You are free to change and redistribute it
under the terms of the GNU General Public License. Please see the
file COPYING in the Unison distribution for more information.
Acknowledgements
Work on Unison has been supported by the National Science Foundation
under grants CCR-9701826 and ITR-0113226,
Principles and Practice of
Synchronization, and by University of Pennsylvania's Institute for
Research in Cognitive Science (IRCS).
Installation
Unison is designed to be easy to install. The following sequence of
steps should get you a fully working installation in a few minutes. If
you run into trouble, you may find the suggestions on the
Frequently Asked
Questions page helpful. Pre-built binaries are available for a
variety of platforms.
Unison can be used with either of two user interfaces:
-
a simple textual interface, suitable for dumb terminals (and
running from scripts), and
- a more sophisticated grapical interface, based on Gtk2.
You will need to install a copy of Unison on every machine that you
want to synchronize. However, you only need the version with a
graphical user interface (if you want a GUI at all) on the machine
where you're actually going to display the interface (the
client
machine). Other machines that you synchronize with can get along just
fine with the textual version.
Downloading Unison
The Unison download site lives under
http://www.cis.upenn.edu/~bcpierce/unison.
If a pre-built binary of Unison is available for the client machine's
architecture, just download it and put it somewhere in your search
path (if you're going to invoke it from the command line) or on your
desktop (if you'll be click-starting it).
The executable file for the graphical version (with a name including
gtkui) actually provides
both interfaces: the graphical one
appears by default, while the textual interface can be selected by including
-ui text on the command line. The
textui executable
provides just the textual interface.
If you don't see a pre-built executable for your architecture, you'll
need to build it yourself. See the
Building Unison section.
There are also a small number of contributed ports to other
architectures that are not maintained by us. See the
Contributed
Ports page to check what's available.
Check to make sure that what you have downloaded is really executable.
Either click-start it, or type
unison -version at the command
line.
Unison can be used in three different modes: with different directories on a
single machine, with a remote machine over a direct socket connection, or
with a remote machine using
ssh for authentication and secure
transfer. If you intend to use the last option, you may need to install
ssh; see the
Installing Ssh section.
Running Unison
Once you've got Unison installed on at least one system, read
the
Tutorial section of the user manual (or type
unison -doc
tutorial) for instructions on how to get started.
Upgrading
Upgrading to a new version of Unison is as simple as throwing away the old
binary and installing the new one.
Before upgrading, it is a good idea to run the
old version one last
time, to make sure all your replicas are completely synchronized. A new
version of Unison will sometimes introduce a different format for the
archive files used to remember information about the previous state of the
replicas. In this case, the old archive will be ignored (not deleted — if
you roll back to the previous version of Unison, you will find the old
archives intact), which means that any differences between the replicas will
show up as conflicts that need to be resolved manually.
Building Unison from Scratch
If a pre-built image is not available, you will need to compile it from
scratch; the sources are available from the same place as the binaries.
In principle, Unison should work on any platform to which OCaml has been
ported and on which the
Unix module is fully implemented. It has
been tested on many flavors of Windows (98, NT, 2000, XP) and Unix (OS X,
Solaris, Linux, FreeBSD), and on both 32- and 64-bit architectures.
Unix
You'll need the Objective Caml compiler (version 3.07 or later), which is
available from
http://caml.inria.fr. Building and installing OCaml
on Unix systems is very straightforward; just follow the instructions in the
distribution. You'll probably want to build the native-code compiler in
addition to the bytecode compiler, as Unison runs much faster when compiled
to native code, but this is not absolutely necessary.
(Quick start: on many systems, the following sequence of commands will
get you a working and installed compiler: first do
make world opt,
then
su to root and do
make install.)
You'll also need the GNU
make utility, standard on many Unix
systems. (Type
make –version to check that you've got the
GNU version.)
Once you've got OCaml installed, grab a copy of the Unison sources,
unzip and untar them, change to the new
unison directory, and
type “
make UISTYLE=text.”
The result should be an executable file called
unison.
Type
./unison to make sure the program is executable. You
should get back a usage message.
If you want to build the graphical user interface, you will need to install
two additional things:
Now build unison. If your search paths are set up correctly, simply typing
make
again should build a
unison executable with a Gtk2 graphical
interface. (In previous releases of Unison, it was necessary to add
UISTYLE=gtk2 to the 'make' command above. This requirement has been
removed: the makefile should detect automatically when lablgtk2 is
present and set this flag automatically.)
Put the
unison executable somewhere in your search path, either
by adding the Unison directory to your PATH variable or by copying the
executable to some standard directory where executables are stored.
Windows
Although the binary distribution should work on any version of Windows,
some people may want to build Unison from scratch on those systems too.
Bytecode version:
The simpler but slower compilation option
to build a Unison executable is to build a bytecode version. You need
first install Windows version of the OCaml compiler (version 3.07 or
later, available from
http://caml.inria.fr). Then grab a copy
of Unison sources and type
make NATIVE=false
to compile the bytecode. The result should be an executable file called
unison.exe.
Native version:
Building a more efficient, native version of
Unison on Windows requires a little more work. See the file
INSTALL.win32 in the source code distribution.
Installation Options
The
Makefile in the distribution includes several switches that
can be used to control how Unison is built. Here are the most useful
ones:
-
Building with
NATIVE=true uses the native-code OCaml
compiler, yielding an executable that will run quite a bit faster. We use
this for building distribution versions.
- Building with
make DEBUGGING=true generates debugging
symbols.
- Building with
make STATIC=true generates a (mostly)
statically linked executable. We use this for building distribution
versions, for portability.
Tutorial
Preliminaries
Unison can be used with either of two user interfaces:
-
a straightforward textual interface and
- a more sophisticated graphical interface
The textual interface is more convenient for running from scripts and
works on dumb terminals; the graphical interface is better for most
interactive use. For this tutorial, you can use either. If you are running
Unison from the command line, just typing
unison
will select either the text or the graphical interface, depending on which
has been selected as default when the executable you are running was
built. You can force the text interface even if graphical is the default by
adding
-ui text.
The other command-line arguments to both versions are identical.
The graphical version can also be run directly by clicking on its icon, but
this may require a little set-up (see the
Click-starting
Unison section). For this tutorial, we assume that you're starting it from the
command line.
Unison can synchronize files and directories on a single machine, or
between two machines on a network. (The same program runs on both
machines; the only difference is which one is responsible for
displaying the user interface.) If you're only interested in a
single-machine setup, then let's call that machine the
client. If
you're synchronizing two machines, let's call them
client and
server.
Local Usage
Let's get the client machine set up first and see how to synchronize
two directories on a single machine.
Follow the instructions in the
Installation section to either
download or build an executable version of Unison, and install it
somewhere on your search path. (If you just want to use the textual user
interface, download the appropriate textui binary. If you just want to
the graphical interface—or if you will use both interfaces [the gtkui
binary actually has both compiled in]—then download the gtkui binary.)
Create a small test directory
a.tmp containing a couple of files
and/or subdirectories, e.g.,
mkdir a.tmp
touch a.tmp/a a.tmp/b
mkdir a.tmp/d
touch a.tmp/d/f
Copy this directory to b.tmp:
cp -r a.tmp b.tmp
Now try synchronizing
a.tmp and
b.tmp. (Since they are
identical, synchronizing them won't propagate any changes, but Unison
will remember the current state of both directories so that it will be
able to tell next time what has changed.) Type:
unison a.tmp b.tmp
Textual Interface:-
You should see a message notifying you that all the files are actually
equal and then get returned to the command line.
Graphical Interface:-
You should get a big empty window with a message at the bottom
notifying you that all files are identical. Choose the Exit item from
the File menu to get back to the command line.
Next, make some changes in a.tmp and/or b.tmp. For example:
rm a.tmp/a
echo "Hello" > a.tmp/b
echo "Hello" > b.tmp/b
date > b.tmp/c
echo "Hi there" > a.tmp/d/h
echo "Hello there" > b.tmp/d/h
Run Unison again:
unison a.tmp b.tmp
This time, the user interface will display only the files that have
changed. If a file has been modified in just one
replica, then it will be displayed with an arrow indicating the
direction that the change needs to be propagated. For example,
<--- new file c [f]
indicates that the file
c has been modified only in the second
replica, and that the default action is therefore to propagate the new
version to the first replica. To
follw Unison's recommendation,
press the “f” at the prompt.
If both replicas are modified and their contents are different, then
the changes are in conflict:
<-?-> is displayed to indicate
that Unison needs guidance on which replica should override the
other.
new file <-?-> new file d/h []
By default, neither version will be propagated and both
replicas will remain as they are.
If both replicas have been modified but their new contents are the same
(as with the file
b), then no propagation is necessary and
nothing is shown. Unison simply notes that the file is up to date.
These display conventions are used by both versions of the user
interface. The only difference lies in the way in which Unison's
default actions are either accepted or overriden by the user.
Textual Interface:-
The status of each modified file is displayed, in turn.
When the copies of a file in the two replicas are not identical, the
user interface will ask for instructions as to how to propagate the
change. If some default action is indicated (by an arrow), you can
simply press Return to go on to the next changed file. If you want to
do something different with this file, press “
<” or “>” to force
the change to be propagated from right to left or from left to right,
or else press “/” to skip this file and leave both replicas alone.
When it reaches the end of the list of modified files, Unison will ask
you one more time whether it should proceed with the updates that have
been selected.
When Unison stops to wait for input from the user, pressing “?”
will always give a list of possible responses and their meanings.
Graphical Interface:-
The main window shows all the files that have been modified in either
a.tmp or b.tmp. To override a default action (or to select
an action in the case when there is no default), first select the file, either
by clicking on its name or by using the up- and down-arrow keys. Then
press either the left-arrow or “
<” key (to cause the version in b.tmp to
propagate to a.tmp) or the right-arrow or “>” key (which makes the a.tmp
version override b.tmp).
Every keyboard command can also be invoked from the menus at the top
of the user interface. (Conversely, each menu item is annotated with
its keyboard equivalent, if it has one.)
When you are satisfied with the directions for the propagation of changes
as shown in the main window, click the “Go” button to set them in
motion. A check sign will be displayed next to each filename
when the file has been dealt with.
Remote Usage
Next, we'll get Unison set up to synchronize replicas on two different
machines.
Follow the instructions in the Installation section to download or
build an executable version of Unison on the server machine, and
install it somewhere on your search path. (It doesn't matter whether
you install the textual or graphical version, since the copy of Unison on
the server doesn't need to display any user interface at all.)
It is important that the version of Unison installed on the server
machine is the same as the version of Unison on the client machine.
But some flexibility on the version of Unison at the client side can
be achieved by using the
-addversionno option; see
the
Preferences section.
Now there is a decision to be made. Unison provides two methods for
communicating between the client and the server:
-
Remote shell method: To use this method, you must have
some way of invoking remote commands on the server from the client's
command line, using a facility such as
ssh.
This method is more convenient (since there is no need to manually
start a “unison server” process on the server) and also more
secure (especially if you use ssh).
- Socket method: This method requires only that you can get
TCP packets from the client to the server and back. A draconian
firewall can prevent this, but otherwise it should work anywhere.
Decide which of these you want to try, and continue with
the
Remote Shell Method section or
the
Socket Method section, as appropriate.
Remote Shell Method
The standard remote shell facility on Unix systems is
ssh, which provides the
same functionality as the older
rsh but much better security. Ssh is available from
ftp://ftp.cs.hut.fi/pub/ssh/; up-to-date binaries for some
architectures can also be found at
ftp://ftp.faqs.org/ssh/contrib. See section
A.2
for installation instructions for the Windows version.
Running
ssh requires some coordination between the client and server
machines to establish that the client is allowed to invoke commands on
the server; please refer to the or
ssh documentation
for information on how to set this up. The examples in this section
use
ssh, but you can substitute
rsh for
ssh if
you wish.
First, test that we can invoke Unison on the server from the client.
Typing
ssh remotehostname unison -version
should print the same version information as running
unison -version
locally on the client. If remote execution fails, then either
something is wrong with your ssh setup (e.g., “permission denied”)
or else the search path that's being used when executing commands on
the server doesn't contain the
unison executable (e.g.,
“command not found”).
Create a test directory
a.tmp in your home directory on the client
machine.
Test that the local unison client can start and connect to the
remote server. Type
unison -testServer a.tmp ssh://remotehostname/a.tmp
Now cd to your home directory and type:
unison a.tmp ssh://remotehostname/a.tmp
The result should be that the entire directory
a.tmp is propagated
from the client to your home directory on the server.
After finishing the first synchronization, change a few files and try
synchronizing again. You should see similar results as in the local
case.
If your user name on the server is not the same as on the client, you
need to specify it on the command line:
unison a.tmp ssh://username@remotehostname/a.tmp
Notes:
Socket Method
Warning: The socket method is
insecure: not only are the texts of your changes transmitted over
the network in unprotected form, it is also possible for anyone in
the world to connect to the server process and read out the contents
of your filesystem! (Of course, to do this they must understand the
protocol that Unison uses to communicate between client and server,
but all they need for this is a copy of the Unison sources.) The socket
method is provided only for expert users with specific needs; everyone
else should use the ssh method.
To run Unison over a socket connection, you must start a Unison
daemon process on the server. This process runs continuously,
waiting for connections over a given socket from client machines
running Unison and processing their requests in turn.
To start the daemon, type
unison -socket NNNN
on the server machine, where
NNNN is the socket number that the
daemon should listen on for connections from clients. (
NNNN can
be any large number that is not being used by some other program; if
NNNN is already in use, Unison will exit with an error
message.) Note that paths specified by the client will be interpreted
relative to the directory in which you start the server process; this
behavior is different from the ssh case, where the path is relative to
your home directory on the server.
Create a test directory
a.tmp in your home directory on the
client machine. Now type:
unison a.tmp socket://remotehostname:NNNN/a.tmp
The result should be that the entire directory
a.tmp is
propagated from the client to the server (
a.tmp will be
created on the server in the directory that the server was started
from).
After finishing the first synchronization, change a few files and try
synchronizing again. You should see similar results as in the local
case.
Since the socket method is not used by many people, its functionality is
rather limited. For example, the server can only deal with one client at a
time.
Using Unison for All Your Files
Once you are comfortable with the basic operation of Unison, you may
find yourself wanting to use it regularly to synchronize your commonly
used files. There are several possible ways of going about this:
-
Synchronize your whole home directory, using the Ignore facility
(see the Ignore section)
to avoid synchronizing temporary files and things that only belong on
one host.
- Create a subdirectory called shared (or current, or
whatever) in your home directory on each host, and put all the files
you want to synchronize into this directory.
- Create a subdirectory called shared (or current, or
whatever) in your home directory on each host, and put links to
all the files you want to synchronize into this directory. Use the
follow preference (see the Symbolic Links section) to make
Unison treat these links as transparent.
- Make your home directory the root of the synchronization, but
tell Unison to synchronize only some of the files and subdirectories
within it on any given run. This can be accomplished by using the -path switch
on the command line:
unison /home/username ssh://remotehost//home/username -path shared
The -path option can be used as many times as needed, to
synchronize several files or subdirectories:
unison /home/username ssh://remotehost//home/username \
-path shared \
-path pub \
-path .netscape/bookmarks.html
These -path arguments can also be put in your preference file.
See the Preferences section for an example.
Most people find that they only need to maintain a profile (or
profiles) on one of the hosts that they synchronize, since Unison is
always initiated from this host. (For example, if you're
synchronizing a laptop with a fileserver, you'll probably always run
Unison on the laptop.) This is a bit different from the usual
situation with asymmetric mirroring programs like
rdist, where
the mirroring operation typically needs to be initiated from the
machine with the most recent changes. the
Profile section
covers the syntax of Unison profiles, together with some sample profiles.
Some tips on improving Unison's performance can be found on the
Frequently
Asked Questions page.
Using Unison to Synchronize More Than Two Machines
Unison is designed for synchronizing pairs of replicas. However, it is
possible to use it to keep larger groups of machines in sync by performing
multiple pairwise synchronizations.
If you need to do this, the most reliable way to set things up is to
organize the machines into a “star topology,” with one machine designated
as the “hub” and the rest as “spokes,” and with each spoke machine
synchronizing only with the hub. The big advantage of the star topology is
that it eliminates the possibility of confusing “spurious conflicts”
arising from the fact that a separate archive is maintained by Unison for
every pair of hosts that it synchronizes.
Going Further
On-line documentation for the various features of Unison
can be obtained either by typing
unison -doc topics
at the command line, or by selecting the Help menu in the graphical
user interface.
The on-line information and the printed manual are essentially identical.
If you use Unison regularly, you should subscribe to one of the mailing
lists, to receive announcements of new versions. See
the
Mailing Lists section.
Basic Concepts
To understand how Unison works, it is necessary to discuss a few
straightforward concepts.
These concepts are developed more rigorously and at more length in a number
of papers, available at
http://www.cis.upenn.edu/~bcpierce/papers.
But the informal presentation here should be enough for most users.
Roots
A replica's
root tells Unison where to find a set of files to be
synchronized, either on the local machine or on a remote host.
For example,
relative/path/of/root
specifies a local root relative to the directory where Unison is
started, while
/absolute/path/of/root
specifies a root relative to the top of the local filesystem,
independent of where Unison is running. Remote roots can begin with
ssh://,
rsh://
to indicate that the remote server should be started with rsh or ssh:
ssh://remotehost//absolute/path/of/root
rsh://user@remotehost/relative/path/of/root
If the remote server is already running (in the socket mode), then the syntax
socket://remotehost:portnum//absolute/path/of/root
socket://remotehost:portnum/relative/path/of/root
is used to specify the hostname and the port that the client Unison should
use to contact it.
The syntax for roots is based on that of URIs (described in RFC 2396).
The full grammar is:
replica ::= [protocol:]//[user@][host][:port][/path]
| path
protocol ::= file
| socket
| ssh
| rsh
user ::= [-_a-zA-Z0-9]+
host ::= [-_a-zA-Z0-9.]+
port ::= [0-9]+
When
path is given without any protocol prefix, the protocol is
assumed to be
file:. Under Windows, it is possible to
synchronize with a remote directory using the
file: protocol over
the Windows Network Neighborhood. For example,
unison foo //host/drive/bar
synchronizes the local directory
foo with the directory
drive:\bar on the machine
host, provided that
host
is accessible via Network Neighborhood. When the
file: protocol
is used in this way, there is no need for a Unison server to be running
on the remote host. However, running Unison this way is only a good
idea if the remote host is reached by a very fast network connection,
since the full contents of every file in the remote replica will have to
be transferred to the local machine to detect updates.
The names of roots are
canonized by Unison before it uses them
to compute the names of the corresponding archive files, so
//saul//home/bcpierce/common and
//saul.cis.upenn.edu/common
will be recognized as the same replica under different names.
Paths
A
path refers to a point
within a set of files being
synchronized; it is specified relative to the root of the replica.
Formally, a path is just a sequence of names, separated by
/.
Note that the path separator character is always a forward slash, no
matter what operating system Unison is running on. Forward slashes
are converted to backslashes as necessary when paths are converted to
filenames in the local filesystem on a particular host.
(For example, suppose that we run Unison on a Windows system, synchronizing
the local root
c:\pierce with the root
ssh://saul.cis.upenn.edu/home/bcpierce on a Unix server. Then
the path
current/todo.txt refers to the file
c:\pierce\current\todo.txt on the client and
/home/bcpierce/current/todo.txt on the server.)
The empty path (i.e., the empty sequence of names) denotes the whole
replica. Unison displays the empty path as “
[root].”
If
p is a path and
q is a path beginning with
p, then
q is said to be a
descendant of
p. (Each path is also a
descendant of itself.)
What is an Update?
The
contents of a path
p in a particular replica could be a
file, a directory, a symbolic link, or absent (if
p does not
refer to anything at all in that replica). More specifically:
-
If
p refers to an ordinary file, then the
contents of p are the actual contents of this file (a string of bytes)
plus the current permission bits of the file.
- If
p refers to a symbolic link, then the contents of p
are just the string specifying where the link points.
- If
p refers to a directory, then the
contents of p are just the token “DIRECTORY” plus the current
permission bits of the directory.
- If
p does not refer to anything in this replica, then the
contents of p are the token “ABSENT.”
Unison keeps a record of the contents of each path after each
successful synchronization of that path (i.e., it remembers the
contents at the last moment when they were the same in the two
replicas).
We say that a path is
updated (in some replica) if its current
contents are different from its contents the last time it was successfully
synchronized. Note that whether a path is updated has nothing to do with
its last modification time—Unison considers only the contents when
determining whether an update has occurred. This means that touching a file
without changing its contents will
not be recognized as an update. A
file can even be changed several times and then changed back to its original
contents; as long as Unison is only run at the end of this process, no
update will be recognized.
What Unison actually calculates is a close approximation to this
definition; see the
Caveats and Shortcomings section.
What is a Conflict?
A path is said to be
conflicting if the following conditions all hold:
-
it has been updated in one replica,
- it or any of its descendants has been updated in the other
replica,
and
- its contents in the two replicas are not identical.
Reconciliation
Unison operates in several distinct stages:
-
On each host, it compares its archive file (which records
the state of each path in the replica when it was last synchronized)
with the current contents of the replica, to determine which paths
have been updated.
- It checks for “false conflicts” — paths that have been
updated on both replicas, but whose current values are identical.
These paths are silently marked as synchronized in the archive files
in both replicas.
- It displays all the updated paths to the user. For updates that
do not conflict, it suggests a default action (propagating the new
contents from the updated replica to the other). Conflicting updates
are just displayed. The user is given an opportunity to examine the
current state of affairs, change the default actions for
nonconflicting updates, and choose actions for conflicting updates.
- It performs the selected actions, one at a time. Each action is
performed by first transferring the new contents to a temporary file
on the receiving host, then atomically moving them into place.
- It updates its archive files to reflect the new state of the
replicas.
Invariants
Given the importance and delicacy of the job that it performs, it is
important to understand both what a synchronizer does under normal
conditions and what can happen under unusual conditions such as system
crashes and communication failures.
Unison is careful to protect both its internal state and the state of
the replicas at every point in this process. Specifically, the
following guarantees are enforced:
-
At every moment, each path in each replica has either (1) its original contents (i.e., no change at all has been made to this
path), or (2) its correct final contents (i.e., the value that the
user expected to be propagated from the other replica).
- At every moment, the information stored on disk about Unison's
private state can be either (1) unchanged, or (2) updated to reflect
those paths that have been successfully synchronized.
The upshot is that it is safe to interrupt Unison at any time, either
manually or accidentally. [Caveat: the above is
almost true there
are occasionally brief periods where it is not (and, because of
shortcoming of the Posix filesystem API, cannot be); in particular, when
it is copying a file onto a directory or vice versa, it must first move
the original contents out of the way. If Unison gets
interrupted during one of these periods, some manual cleanup may be
required. In this case, a file called
DANGER.README will be left
in your home directory, containing information about the operation that
was interrupted. The next time you try to run Unison, it will notice this
file and warn you about it.]
If an interruption happens while it is propagating updates, then there
may be some paths for which an update has been propagated but which
have not been marked as synchronized in Unison's archives. This is no
problem: the next time Unison runs, it will detect changes to these
paths in both replicas, notice that the contents are now equal, and
mark the paths as successfully updated when it writes back its private
state at the end of this run.
If Unison is interrupted, it may sometimes leave temporary working files
(with suffix
.tmp) in the replicas. It is safe to delete these
files. Also, if the
backups flag is set, Unison will
leave around old versions of files that it overwrites, with names like
file.0.unison.bak. These can be deleted safely when they are no
longer wanted.
Unison is not bothered by clock skew between the different hosts on
which it is running. It only performs comparisons between timestamps
obtained from the same host, and the only assumption it makes about
them is that the clock on each system always runs forward.
If Unison finds that its archive files have been deleted (or that the
archive format has changed and they cannot be read, or that they don't
exist because this is the first run of Unison on these particular
roots), it takes a conservative approach: it behaves as though the
replicas had both been completely empty at the point of the last
synchronization. The effect of this is that, on the first run, files
that exist in only one replica will be propagated to the other, while
files that exist in both replicas but are unequal will be marked as
conflicting.
Touching a file without changing its contents should never affect whether or
not Unison does an update. (When running with the fastcheck preference set
to true—the default on Unix systems—Unison uses file modtimes for a
quick first pass to tell which files have definitely not changed; then, for
each file that might have changed, it computes a fingerprint of the file's
contents and compares it against the last-synchronized contents. Also, the
-times option allows you to synchronize file times, but it does not
cause identical files to be changed; Unison will only modify the file
times.)
It is safe to “brainwash” Unison by deleting its archive files
on both replicas. The next time it runs, it will assume that
all the files it sees in the replicas are new.
It is safe to modify files while Unison is working. If Unison
discovers that it has propagated an out-of-date change, or that the
file it is updating has changed on the target replica, it will signal
a failure for that file. Run Unison again to propagate the latest
change.
Changes to the ignore patterns from the user interface (e.g., using
the `i' key) are immediately reflected in the current profile.
Caveats and Shortcomings
Here are some things to be careful of when using Unison.
-
In the interests of speed, the update detection algorithm may
(depending on which OS architecture that you run Unison on)
actually use an approximation to the definition given in
the What is an Update? section.
In particular, the Unix
implementation does not compare the actual contents of files to their
previous contents, but simply looks at each file's inode number and
modtime; if neither of these have changed, then it concludes that the
file has not been changed.
Under normal circumstances, this approximation is safe, in the sense
that it may sometimes detect “false updates” will never miss a real
one. However, it is possible to fool it, for example by using
retouch to change a file's modtime back to a time in the past.
- If you synchronize between a single-user filesystem and a shared
Unix server, you should pay attention to your permission bits: by
default, Unison will synchronize permissions verbatim, which may leave
group-writable files on the server that could be written over by a lot of
people.
You can control this by setting your umask on both computers to
something like 022, masking out the “world write” and “group write”
permission bits.
Unison does not synchronize the setuid and setgid bits, for
security.
- The graphical user interface is single-threaded. This
means that if Unison is performing some long-running operation, the
display will not be repainted until it finishes. We recommend not
trying to do anything with the user interface while Unison is in the
middle of detecting changes or propagating files.
- Unison does not understand hard links.
- It is important to be a little careful when renaming directories
containing “ignore”d files.
For example, suppose Unison is synchronizing directory A between the two
machines called the “local” and the “remote” machine; suppose directory
A contains a subdirectory D; and suppose D on the local machine contains a
file or subdirectory P that matches an ignore directive in the profile used
to synchronize. Thus path A/D/P exists on the local machine but not on the
remote machine.
If D is renamed to D' on the remote machine, and this change is
propagated to the local machine, all such files or subdirectories P
will be deleted. This is because Unison sees the rename as a delete and a
separate create: it deletes the old directory (including the ignored files)
and creates a new one (not including the ignored files, since they
are completely invisible to it).
Reference Guide
This section covers the features of Unison in detail.
Running Unison
There are several ways to start Unison.
-
Typing “unison profile” on the command line. Unison
will look for a file profile.prf in the
.unison
directory. If this file does not specify a pair of roots, Unison will
prompt for them and add them to the information specified by the profile.
- Typing “unison profile root1 root2” on the command
line.
In this case, Unison will use profile, which should not contain
any root directives.
- Typing “unison root1 root2” on the command line. This
has the same effect as typing “unison default root1 root2.”
- Typing just “unison” (or invoking Unison by clicking on
a desktop icon). In this case, Unison will ask for the profile to use
for synchronization (or create a new one, if necessary).
The .unison Directory
Unison stores a variety of information in a private directory on each
host. If the environment variable
UNISON is defined, then its
value will be used as the name of this directory. If
UNISON is
not defined, then the name of the directory depends on which
operating system you are using. In Unix, the default is to use
$HOME/.unison.
In Windows, if the environment variable
USERPROFILE is defined, then the directory will be
$USERPROFILE\.unison;
otherwise if
HOME is defined, it will be
$HOME\.unison;
otherwise, it will be
c:\.unison.
The archive file for each replica is found in the
.unison
directory on that replica's host. Profiles (described below) are
always taken from the
.unison directory on the client host.
Note that Unison maintains a completely different set of archive files
for each pair of roots.
We do not recommend synchronizing the whole
.unison directory, as this
will involve frequent propagation of large archive files. It should be safe
to do it, though, if you really want to. Synchronizing just the profile
files in the
.unison directory is definitely OK.
Archive Files
The name of the archive file on each replica is calculated from
-
the canonical names of all the hosts (short names like
saul are converted into full addresses like saul.cis.upenn.edu),
- the paths to the replicas on all the hosts (again, relative
pathnames, symbolic links, etc. are converted into full, absolute paths), and
- an internal version number that is changed whenever a new Unison
release changes the format of the information stored in the archive.
This method should work well for most users. However, it is occasionally
useful to change the way archive names are generated. Unison provides
two ways of doing this.
The function that finds the canonical hostname of the local host (which
is used, for example, in calculating the name of the archive file used to
remember which files have been synchronized) normally uses the
gethostname operating system call. However, if the environment
variable
UNISONLOCALHOSTNAME is set, its value will be used
instead. This makes it easier to use Unison in situations where a
machine's name changes frequently (e.g., because it is a laptop and gets
moved around a lot).
A more powerful way of changing archive names is provided by the
rootalias preference. The preference file may contain any number of
lines of the form:
rootalias = //hostnameA//path-to-replicaA -> //hostnameB//path-to-replicaB
When calculating the name of the archive files for a given pair of roots,
Unison replaces any root that matches the left-hand side of any rootalias
rule by the corresponding right-hand side.
So, if you need to relocate a root on one of the hosts, you can add a
rule of the form:
rootalias = //new-hostname//new-path -> //old-hostname//old-path
Warning: The
rootalias option is dangerous and should only
be used if you are sure you know what you're doing. In particular, it
should only be used if you are positive that either (1) both the original
root and the new alias refer to the same set of files, or (2) the files
have been relocated so that the original name is now invalid and will
never be used again. (If the original root and the alias refer to
different sets of files, Unison's update detector could get confused.)
After introducing a new
rootalias, it is a good idea to run Unison
a few times interactively (with the
batch flag off, etc.) and
carefully check that things look reasonable—in particular, that update
detection is working as expected.
Preferences
Many details of Unison's behavior are configurable by user-settable
“preferences.”
Some preferences are boolean-valued; these are often called
flags.
Others take numeric or string arguments, indicated in the preferences
list by
n or
xxx. Most of the string preferences can be
given several times; the arguments are accumulated into a list
internally.
There are two ways to set the values of preferences: temporarily, by
providing command-line arguments to a particular run of Unison, or
permanently, by adding commands to a
profile in the
.unison
directory on the client host. The order of preferences (either on the
command line or in preference files) is not significant. On the command
line, preferences and other arguments (the profile name and roots) can be
intermixed in any order.
To set the value of a preference
p from the command line, add an
argument
-p (for a boolean flag) or
-p n or
-p xxx (for
a numeric or string preference) anywhere on the command line. To set a
boolean flag to
false on the command line, use
-p=false.
Here are all the preferences supported by Unison. This list can be
obtained by typing
unison -help.
Usage: unison [options]
or unison root1 root2 [options]
or unison profilename [options]
Options:
-addprefsto xxx file to add new prefs to
-addversionno add version number to name of unison executable on server
-auto automatically accept default actions
-backup xxx add a pattern to the backup list
-backupcurrent xxx add a pattern to the backupcurrent list
-backupcurrentnot xxx add a pattern to the backupcurrentnot list
-backupdir xxx Directory for storing centralized backups
-backuplocation xxx where backups are stored ('local' or 'central')
-backupnot xxx add a pattern to the backupnot list
-backupprefix xxx prefix for the names of backup files
-backups keep backup copies of all files (see also 'backup')
-backupsuffix xxx a suffix to be added to names of backup files
-batch batch mode: ask no questions at all
-confirmbigdeletes request confirmation for whole-replica deletes
-confirmmerge ask for confirmation before commiting results of a merge
-contactquietly Suppress the 'contacting server' message during startup
-debug xxx debug module xxx ('all' -> everything, 'verbose' -> more)
-doc xxx show documentation ('-doc topics' lists topics)
-dumbtty do not try to change terminal settings in text UI
-fastcheck xxx do fast update detection (`true', `false', or `default')
-follow xxx add a pattern to the follow list
-force xxx force changes from this replica to the other
-forcepartial xxx add a pattern to the forcepartial list
-group synchronize group
-height n height (in lines) of main window in graphical interface
-host xxx bind the socket to this host name in server socket mode
-ignore xxx add a pattern to the ignore list
-ignorecase xxx ignore upper/lowercase in filenames (`true', `false', or `default')
-ignorelocks ignore locks left over from previous run (dangerous!)
-ignorenot xxx add a pattern to the ignorenot list
-immutable xxx add a pattern to the immutable list
-immutablenot xxx add a pattern to the immutablenot list
-key xxx define a keyboard shortcut for this profile (in some UIs)
-killserver kill server when done (even when using sockets)
-label xxx provide a descriptive string label for this profile
-log record actions in file specified by logfile preference
-logfile xxx Log file name
-maxbackups n number of backed up versions of a file
-maxthreads n maximum number of simultaneous file transfers
-merge xxx add a pattern to the merge list
-mountpoint xxx abort if this path does not exist
-numericids don't map uid/gid values by user/group names
-owner synchronize owner
-path xxx path to synchronize
-perms n part of the permissions which is synchronized
-prefer xxx choose this replica's version for conflicting changes
-preferpartial xxx add a pattern to the preferpartial list
-pretendwin Use creation times for detecting updates
-repeat xxx synchronize repeatedly (text interface only)
-retry n re-try failed synchronizations N times (text interface only)
-root xxx root of a replica
-rootalias xxx Register alias for canonical root names
-rsrc xxx synchronize resource forks and HFS meta-data (`true', `false', or `default')
-rsync activate the rsync transfer mode
-selftest run internal tests and exit
-servercmd xxx name of unison executable on remote server
-showarchive show name of archive and 'true names' (for rootalias) of roots
-silent print nothing (except error messages)
-socket xxx act as a server on a socket
-sortbysize list changed files by size, not name
-sortfirst xxx add a pattern to the sortfirst list
-sortlast xxx add a pattern to the sortlast list
-sortnewfirst list new before changed files
-sshargs xxx other arguments (if any) for remote shell command
-sshcmd xxx path to the ssh executable
-terse suppress status messages
-testserver exit immediately after the connection to the server
-times synchronize modification times
-ui xxx select user interface ('text' or 'graphic'); command-line only
-version print version and exit
-xferbycopying optimize transfers using local copies, if possible
Here, in more detail, are what they do. Many are discussed in even greater
detail in other sections of the manual.
-
addprefsto xxx
-
By default, new preferences added by Unison (e.g., new
ignore clauses) will be appended to whatever preference file Unison was told to load at the beginning of the run. Setting the preference addprefsto filename makes Unison add new preferences to the file named filename instead.
- addversionno
-
When this flag is set to true, Unison will use unison-currentversionnumber instead of just
unison as the remote server command. This allows multiple binaries for different versions of unison to coexist conveniently on the same server: whichever version is run on the client, the same version will be selected on the server.
- auto
-
When set to true, this flag causes the user interface to skip asking for confirmations on non-conflicting changes. (More precisely, when the user interface is done setting the propagation direction for one entry and is about to move to the next, it will skip over all non-conflicting entries and go directly to the next conflict.)
- backup xxx
-
Including the preference -backup pathspec causes Unison to keep backup files for each path that matches pathspec. These backup files are kept in the directory specified by the
backuplocation preference. The backups are named according to the backupprefix and backupsuffix preferences. The number of versions that are kept is determined by the maxbackups preference.
The syntax of pathspec is described in the Path Specification section.
- backupcurrent xxx
-
Including the preference -backupcurrent pathspec causes Unison to keep a backup of the current version of every file matching pathspec. This file will be saved as a backup with version number 000. Such backups can be used as inputs to external merging programs, for instance. See the documentatation for the
merge preference. For more details, see the Merging Conflicting Versions section.
The syntax of pathspec is described in the Path Specification section.
- backupcurrentnot xxx
-
Exceptions to
backupcurrent, like the ignorenot preference.
- backupdir xxx
-
If this preference is set, Unison will use it as the name of the directory used to store backup files specified by the backup preference, when backuplocation is set to
central. It is checked after the UNISONBACKUPDIR environment variable.
- backuplocation xxx
-
This preference determines whether backups should be kept locally, near the original files, or in a central directory specified by the backupdir preference. If set to
local, backups will be kept in the same directory as the original files, and if set to central, backupdir will be used instead.
- backupnot xxx
-
The values of this preference specify paths or individual files or regular expressions that should not be backed up, even if the backup preference selects them—i.e., it selectively overrides backup. The same caveats apply here as with ignore and t ignorenot.
- backupprefix xxx
-
When a backup for a file
NAME is created, it is stored in a directory specified by backuplocation, in a file called backupprefixNAMEbackupsuffix. backupprefix can include a directory name (causing Unison to keep all backup files for a given directory in a subdirectory with this name), and both backupprefix and backupsuffix can contain the string$VERSION, which will be replaced by the age of the backup (1 for the most recent, 2 for the second most recent, and so on...). This keyword is ignored if it appears in a directory name in the prefix; if it does not appear anywhere in the prefix or the suffix, it will be automatically placed at the beginning of the suffix.
- backups
-
Setting this flag to true is equivalent to setting backuplocation to local and backup to
Name *.
- backupsuffix xxx
-
See backupprefix for full documentation.
- batch
-
When this is set to true, the user interface will ask no questions at all. Non-conflicting changes will be propagated; conflicts will be skipped.
- confirmbigdeletes
-
When this is set to true, Unison will request an extra confirmation if it appears that the entire replica has been deleted, before propagating the change. If the batch flag is also set, synchronization will be aborted. When the path preference is used, the same confirmation will be requested for top-level paths. (At the moment, this flag only affects the text user interface.) See also the mountpoint preference.
- confirmmerge
-
Setting this preference causes both the text and graphical interfaces to ask the user if the results of a merge command may be commited to the replica or not. Since the merge command works on temporary files, the user can then cancel all the effects of applying the merge if it turns out that the result is not satisfactory. In batch-mode, this preference has no effect.
- contactquietly
-
If this flag is set, Unison will skip displaying the `Contacting server' message (which some users find annoying) during startup.
- debug xxx
-
This preference is used to make Unison print various sorts of information about what it is doing internally on the standard error stream. It can be used many times, each time with the name of a module for which debugging information should be printed. Possible arguments for
debug can be found by looking for calls to Util.debug in the sources (using, e.g., grep). Setting -debug all causes information from all modules to be printed (this mode of usage is the first one to try, if you are trying to understand something that Unison seems to be doing wrong); -debug verbose turns on some additional debugging output from some modules (e.g., it will show exactly what bytes are being sent across the network).
- diff xxx
-
This preference can be used to control the name and command-line arguments of the system utility used to generate displays of file differences. The default is `
diff -u CURRENT2 CURRENT1'. If the value of this preference contains the substrings CURRENT1 and CURRENT2, these will be replaced by the names of the files to be diffed. If not, the two filenames will be appended to the command. In both cases, the filenames are suitably quoted.
- doc xxx
-
The command-line argument -doc secname causes unison to display section secname of the manual on the standard output and then exit. Use
-doc all to display the whole manual, which includes exactly the same information as the printed and HTML manuals, modulo formatting. Use -doc topics to obtain a list of the names of the various sections that can be printed.
- dumbtty
-
When set to
true, this flag makes the text mode user interface avoid trying to change any of the terminal settings. (Normally, Unison puts the terminal in `raw mode', so that it can do things like overwriting the current line.) This is useful, for example, when Unison runs in a shell inside of Emacs.
When dumbtty is set, commands to the user interface need to be followed by a carriage return before Unison will execute them. (When it is off, Unison recognizes keystrokes as soon as they are typed.)
This preference has no effect on the graphical user interface.
- dumparchives
-
When this preference is set, Unison will create a file unison.dump on each host, containing a text summary of the archive, immediately after loading it.
- fastcheck xxx
-
When this preference is set to
true, Unison will use the modification time and length of a file as a
`pseudo inode number'
when scanning replicas for updates, instead of reading the full contents of every file. Under Windows, this may cause Unison to miss propagating an update if the modification time and length of the file are both unchanged by the update. However, Unison will never overwrite such an update with a change from the other replica, since it always does a safe check for updates just before propagating a change. Thus, it is reasonable to use this switch under Windows most of the time and occasionally run Unison once with fastcheck set to false, if you are worried that Unison may have overlooked an update. The default value of the preference is auto, which causes Unison to use fast checking on Unix replicas (where it is safe) and slow checking on Windows replicas. For backward compatibility, yes, no, and default can be used in place of true, false, and auto. See the Fast Checking section for more information.
- follow xxx
-
Including the preference -follow pathspec causes Unison to treat symbolic links matching pathspec as `invisible' and behave as if the object pointed to by the link had appeared literally at this position in the replica. See the Symbolic Links section for more details. The syntax of pathspec> is described in the Path Specification section.
- force xxx
-
Including the preference -force root causes Unison to resolve all differences (even non-conflicting changes) in favor of root. This effectively changes Unison from a synchronizer into a mirroring utility.
You can also specify -force newer (or -force older) to force Unison to choose the file with the later (earlier) modtime. In this case, the -times preference must also be enabled.
This preference is overridden by the forcepartial preference.
This preference should be used only if you are sure you know what you are doing!
- forcepartial xxx
-
Including the preference forcepartial PATHSPEC -> root causes Unison to resolve all differences (even non-conflicting changes) in favor of root for the files in PATHSPEC (see the Path Specification section for more information). This effectively changes Unison from a synchronizer into a mirroring utility.
You can also specify forcepartial PATHSPEC -> newer (or forcepartial PATHSPEC older) to force Unison to choose the file with the later (earlier) modtime. In this case, the -times preference must also be enabled.
This preference should be used only if you are sure you know what you are doing!
- group
-
When this flag is set to
true, the group attributes of the files are synchronized. Whether the group names or the group identifiers are synchronizeddepends on the preference numerids.
- height n
-
Used to set the height (in lines) of the main window in the graphical user interface.
- ignore xxx
-
Including the preference -ignore pathspec causes Unison to completely ignore paths that match pathspec (as well as their children). This is useful for avoiding synchronizing temporary files, object files, etc. The syntax of pathspec is described in the Path Specification section, and further details on ignoring paths is found in the Ignoring Paths section.
- ignorecase xxx
-
When set to true, this flag causes Unison to treat filenames as case insensitive—i.e., files in the two replicas whose names differ in (upper- and lower-case) `spelling' are treated as the same file. When the flag is set to false, Unison will treat all filenames as case sensitive. Ordinarily, when the flag is set to t default, filenames are automatically taken to be case-insensitive if either host is running Windows or OSX. In rare circumstances it is useful to set the flag manually (e.g. when running Unison on a Unix system with a FAT [Windows] volume mounted).
- ignorelocks
-
When this preference is set, Unison will ignore any lock files that may have been left over from a previous run of Unison that was interrupted while reading or writing archive files; by default, when Unison sees these lock files it will stop and request manualintervention. This option should be set only if you are positive that no other instance of Unison might be concurrently accessing the same archive files (e.g., because there was only one instance of unison running and it has just crashed or you have just killed it). It is probably not a good idea to set this option in a profile: it is intended for command-line use.
- ignorenot xxx
-
This preference overrides the preference ignore.
It gives a list of patterns
(in the same format as
ignore) for paths that should definitely not be ignored,
whether or not they happen to match one of the ignore patterns.
Note that the semantics of ignore and ignorenot is a
little counter-intuitive. When detecting updates, Unison examines
paths in depth-first order, starting from the roots of the replicas
and working downwards. Before examining each path, it checks whether
it matches ignore and does not match ignorenot; in this case
it skips this path and all its descendants. This means that,
if some parent of a given path matches an ignore pattern, then
it will be skipped even if the path itself matches an ignorenot
pattern. In particular, putting ignore = Path * in your profile
and then using t ignorenot to select particular paths to be
synchronized will not work. Instead, you should use the path
preference to choose particular paths to synchronize.
- immutable xxx
-
This preference specifies paths for directories whose children are all immutable files — i.e., once a file has been created, its contents never changes. When scanning for updates, Unison does not check whether these files have been modified; this can speed update detection significantly (in particular, for mail directories).
- immutablenot xxx
-
This preference overrides immutable.
- key xxx
-
Used in a profile to define a numeric key (0-9) that can be used in the graphical user interface to switch immediately to this profile.
- killserver
-
When set to
true, this flag causes Unison to kill the remote server process when the synchronization is finished. This behavior is the default for ssh connections, so this preference is not normally needed when running over ssh; it is provided so that socket-mode servers can be killed off after a single run of Unison, rather than waiting to accept future connections. (Some users prefer to start a remote socket server for each run of Unison, rather than leaving one running all the time.)
- label xxx
-
Used in a profile to provide a descriptive string documenting its settings. (This is useful for users that switch between several profiles, especially using the `fast switch' feature of the graphical user interface.)
- log
-
When this flag is set, Unison will log all changes to the filesystems
on a file.
- logfile xxx
-
By default, logging messages will be appended to the file
unison.log in your HOME directory. Set this preference if
you prefer another file.
- maxbackups n
-
This preference specifies the number of backup versions that will be kept by unison, for each path that matches the predicate
backup. The default is 2.
- maxthreads n
-
This preference controls how much concurrency is allowed during the transport phase. Normally, it should be set reasonably high (default is 20) to maximize performance, but when Unison is used over a low-bandwidth link it may be helpful to set it lower (e.g. to 1) so that Unison doesn't soak up all the available bandwidth.
- merge xxx
-
This preference can be used to run a merge program which will create a new version for each of the files and the backup, with the last backup and the both replicas. Setting the merge preference for a path will also cause this path to be backed up, just like t backup. The syntax of pathspec>cmd is described in the Path Specification section, and further details on Merging functions are present in the Merging files section.
- mountpoint xxx
-
Including the preference -mountpoint PATH causes Unison to double-check, at the end of update detection, that PATH exists and abort if it does not. This is useful when Unison is used to synchronize removable media. This preference can be given more than once. See the Mount Points section.
- numericids
-
When this flag is set to
true, groups and users are synchronized numerically, rather than by name.
The special uid 0 and the special group 0 are never mapped via user/group names even if this preference is not set.
- owner
-
When this flag is set to
true, the owner attributes of the files are synchronized. Whether the owner names or the owner identifiers are synchronizeddepends on the preference extttnumerids.
- path xxx
-
When no
path preference is given, Unison will simply synchronize the two entire replicas, beginning from the given pair of roots. If one or more path preferences are given, then Unison will synchronize only these paths and their children. (This is useful for doing a fast sync of just one directory, for example.) Note that path preferences are intepreted literally—they are not regular expressions.
- perms n
-
The integer value of this preference is a mask indicating which permission bits should be synchronized. It is set by default to 0o1777: all bits but the set-uid and set-gid bits are synchronised (synchronizing theses latter bits can be a security hazard). If you want to synchronize all bits, you can set the value of this preference to −1.
- prefer xxx
-
Including the preference -prefer root causes Unison always to resolve conflicts in favor of root, rather than asking for guidance from the user. (The syntax of root is the same as for the
root preference, plus the special values newer and older.)
This preference is overridden by the preferpartial preference.
This preference should be used only if you are sure you know what you are doing!
- preferpartial xxx
-
Including the preference preferpartial PATHSPEC -> root causes Unison always to resolve conflicts in favor of root, rather than asking for guidance from the user, for the files in PATHSPEC (see the Path Specification section for more information). (The syntax of root is the same as for the
root preference, plus the special values newer and older.)
This preference should be used only if you are sure you know what you are doing!
- pretendwin
-
When set to true, this preference makes Unison use Windows-style fast update detection (using file creation times as “pseudo-inode-numbers”), even when running on a Unix system. This switch should be used with care, as it is less safe than the standard update detection method, but it can be useful for synchronizing VFAT filesystems (which do not support inode numbers) mounted on Unix systems. The fastcheck option should also be set to true.
- repeat xxx
-
Setting this preference causes the text-mode interface to synchronize repeatedly, rather than doing it just once and stopping. If the argument is a number, Unison will pause for that many seconds before beginning again.
- retry n
-
Setting this preference causes the text-mode interface to try again to synchronize updated paths where synchronization fails. Each such path will be tried N times.
- root xxx
-
Each use of this preference names the root of one of the replicas for Unison to synchronize. Exactly two roots are needed, so normal modes of usage are either to give two values for
root in the profile, or to give no values in the profile and provide two on the command line. Details of the syntax of roots can be found in the Roots section.
The two roots can be given in either order; Unison will sort them into a canonical order before doing anything else. It also tries to `canonize' the machine names and paths that appear in the roots, so that, if Unison is invoked later with a slightly different name for the same root, it will be able to locate the correct archives.
- rootalias xxx
-
When calculating the name of the archive files for a given pair of roots, Unison replaces any roots matching the left-hand side of any rootalias rule by the corresponding right-hand side.
- rshargs xxx
-
The string value of this preference will be passed as additional arguments (besides the host name and the name of the Unison executable on the remote system) to the
rsh command used to invoke the remote server.
- rshcmd xxx
-
This preference can be used to explicitly set the name of the rsh executable (e.g., giving a full path name), if necessary.
- rsrc xxx
-
When set to true, this flag causes Unison to synchronize resource forks and HFS meta-data. On filesystems that do not natively support resource forks, this data is stored in Carbon-compatible ._ AppleDouble files. When the flag is set to false, Unison will not synchronize these data. Ordinarily, the flag is set to default, and these data are
automatically synchronized if either host is running OSX. In rare circumstances it is useful to set the flag manually.
- rsync
-
Unison uses the 'rsync algorithm' for 'diffs-only' transfer of updates to large files. Setting this flag to false makes Unison use whole-file transfers instead. Under normal circumstances, there is no reason to do this, but if you are having trouble with repeated 'rsync failure' errors, setting it to false should permit you to synchronize the offending files.
- selftest
-
Run internal tests and exit. This option is mostly for developers and must be used carefully: in particular, it will delete the contents of both roots, so that it can install its own files for testing. This flag only makes sense on the command line. When it is provided, no preference file is read: all preferences must be specified on thecommand line. Also, since the self-test procedure involves overwriting the roots and backup directory, the names of the roots and of the backupdir preference must include the string "test" or else the tests will be aborted. (If these are not given on the command line, dummy subdirectories in the current directory will be created automatically.)
- servercmd xxx
-
This preference can be used to explicitly set the name of the Unison executable on the remote server (e.g., giving a full path name), if necessary.
- showarchive
-
When this preference is set, Unison will print out the 'true names'of the roots, in the same form as is expected by the rootaliaspreference.
- silent
-
When this preference is set to true, the textual user interface will print nothing at all, except in the case of errors. Setting silent to true automatically sets the batch preference to true.
- sortbysize
-
When this flag is set, the user interface will list changed files by size (smallest first) rather than by name. This is useful, for example, for synchronizing over slow links, since it puts very large files at the end of the list where they will not prevent smaller files from being transferred quickly.
This preference (as well as the other sorting flags, but not the sorting preferences that require patterns as arguments) can be set interactively and temporarily using the 'Sort' menu in the graphical user interface.
- sortfirst xxx
-
Each argument to sortfirst is a pattern pathspec, which describes a set of paths. Files matching any of these patterns will be listed first in the user interface. The syntax of pathspec is described in the Path Specification section.
- sortlast xxx
-
Similar to
sortfirst, except that files matching one of these patterns will be listed at the very end.
- sortnewfirst
-
When this flag is set, the user interface will list newly created files before all others. This is useful, for example, for checking that newly created files are not `junk', i.e., ones that should be ignored or deleted rather than synchronized.
- sshargs xxx
-
The string value of this preference will be passed as additional arguments (besides the host name and the name of the Unison executable on the remote system) to the
ssh command used to invoke the remote server.
- sshcmd xxx
-
This preference can be used to explicitly set the name of the ssh executable (e.g., giving a full path name), if necessary.
- sshversion xxx
-
This preference can be used to control which version of ssh should be used to connect to the server. Legal values are 1 and 2, which will cause unison to try to use
ssh1 orssh2 instead of just ssh to invoke ssh. The default value is empty, which will make unison use whatever version of ssh is installed as the default `ssh' command.
- terse
-
When this preference is set to true, the user interface will not print status messages.
- testserver
-
Setting this flag on the command line causes Unison to attempt to connect to the remote server and, if successful, print a message and immediately exit. Useful for debugging installation problems. Should not be set in preference files.
- times
-
When this flag is set to
true, file modification times (but not directory modtimes) are propagated.
- ui xxx
-
This preference selects either the graphical or the textual user interface. Legal values are
graphic or text.
Because this option is processed specially during Unison's start-up sequence, it can only be used on the command line. In preference files it has no effect.
If the Unison executable was compiled with only a textual interface, this option has no effect. (The pre-compiled binaries are all compiled with both interfaces available.)
- version
-
Print the current version number and exit. (This option only makes sense on the command line.)
- xferbycopying
-
When this preference is set, Unison will try to avoid transferring file contents across the network by recognizing when a file with the required contents already exists in the target replica. This usually allows file moves to be propagated very quickly. The default value istrue.
Profiles
A
profile is a text file that specifies permanent settings for
roots, paths, ignore patterns, and other preferences, so that they do
not need to be typed at the command line every time Unison is run.
Profiles should reside in the
.unison directory on the client
machine. If Unison is started with just one argument
name on
the command line, it looks for a profile called
name.prf in
the
.unison directory. If it is started with no arguments, it
scans the
.unison directory for files whose names end in
.prf and offers a menu (provided that the Unison executable is compiled with the graphical user interface). If a file named
default.prf is
found, its settings will be offered as the default choices.
To set the value of a preference
p permanently, add to the
appropriate profile a line of the form
p = true
for a boolean flag or
p = <value>
for a preference of any other type.
Whitespaces around
p and
xxx are ignored.
A profile may also include blank lines and lines beginning
with
#; both are ignored.
When Unison starts, it first reads the profile and then the command
line, so command-line options will override settings from the
profile.
Profiles may also include lines of the form
include
name, which will cause the file
name (or
name.prf, if
name does not exist in the
.unison directory) to be read at the point, and included as if
its contents, instead of the
include line, was part of the
profile. Include lines allows settings common to several profiles to
be stored in one place.
A profile may include a preference `
label = desc' to
provide a description of the options selected in this profile. The
string
desc is listed along with the profile name in the profile
selection dialog, and displayed in the top-right corner of the main
Unison window in the graphical user interface.
The graphical user-interface also supports one-key shortcuts for commonly
used profiles. If a profile contains a preference of the form
`
key = n', where
n is a single digit, then
pressing this digit key will cause Unison to immediately switch to
this profile and begin synchronization again from scratch. In this
case, all actions that have been selected for a set of changes
currently being displayed will be discarded.
Sample Profiles
A Minimal Profile
Here is a very minimal profile file, such as might be found in
.unison/default.prf:
# Roots of the synchronization
root = /home/bcpierce
root = ssh://saul//home/bcpierce
# Paths to synchronize
path = current
path = common
path = .netscape/bookmarks.html
A Basic Profile
Here is a more sophisticated profile, illustrating some other useful
features.
# Roots of the synchronization
root = /home/bcpierce
root = ssh://saul//home/bcpierce
# Paths to synchronize
path = current
path = common
path = .netscape/bookmarks.html
# Some regexps specifying names and paths to ignore
ignore = Name temp.*
ignore = Name *~
ignore = Name .*~
ignore = Path */pilot/backup/Archive_*
ignore = Name *.o
ignore = Name *.tmp
# Window height
height = 37
# Keep a backup copy of every file in a central location
backuplocation = central
backupdir = /home/bcpierce/backups
backup = Name *
backupprefix = $VERSION.
backupsuffix =
# Use this command for displaying diffs
diff = diff -y -W 79 --suppress-common-lines
# Log actions to the terminal
log = true
A Power-User Profile
When Unison is used with large replicas, it is often convenient to be
able to synchronize just a part of the replicas on a given run (this
saves the time of detecting updates in the other parts). This can be
accomplished by splitting up the profile into several parts — a common
part containing most of the preference settings, plus one “top-level”
file for each set of paths that need to be synchronized. (The
include mechanism can also be used to allow the same set of preference
settings to be used with different roots.)
The collection
of profiles implementing this scheme might look as follows.
The file
default.prf is empty except for an
include
directive:
# Include the contents of the file common
include common
Note that the name of the common file is
common, not
common.prf; this prevents Unison from offering
common as one of
the list of profiles in the opening dialog (in the graphical UI).
The file
common contains the real preferences:
# Roots of the synchronization
root = /home/bcpierce
root = ssh://saul//home/bcpierce
# (... other preferences ...)
# If any new preferences are added by Unison (e.g. 'ignore'
# preferences added via the graphical UI), then store them in the
# file 'common' rathen than in the top-level preference file
addprefsto = common
# Names and paths to ignore:
ignore = Name temp.*
ignore = Name *~
ignore = Name .*~
ignore = Path */pilot/backup/Archive_*
ignore = Name *.o
ignore = Name *.tmp
Note that there are no
path preferences in
common. This
means that, when we invoke Unison with the default profile (e.g., by
typing '
unison default' or just '
unison' on the command
line), the whole replicas will be synchronized. (If we
never want
to synchronize the whole replicas, then
default.prf would instead
include settings for all the paths that are usually synchronized.)
To synchronize just part of the replicas, Unison is invoked with an
alternate preference file—e.g., doing '
unison workingset', where the
preference file
workingset.prf contains
path = current/papers
path = Mail/inbox
path = Mail/drafts
include common
causes Unison to synchronize just the listed subdirectories.
The
key preference can be used in combination with the graphical UI
to quickly switch between different sets of paths. For example, if the
file
mail.prf contains
path = Mail
batch = true
key = 2
include common
then pressing 2 will cause Unison to look for updates in the
Mail
subdirectory and (because the
batch flag is set) immediately
propagate any that it finds.
Keeping Backups
When Unison overwrites a file or directory by propagating a new version from
the other replica, it can keep the old version around as a backup. There
are several preferences that control precisely where these backups are
stored and how they are named.
To enable backups, you must give one or more
backup preferences.
Each of these has the form
backup = <pathspec>
where
<pathspec> has the same form as for the
ignore
preference. For example,
backup = Name *
causes Unison to keep backups of
all files and directories. The
backupnot preference can be used to give a few exceptions: it
specifies which files and directories should
not be backed up, even if
they match the
backup pathspec.
It is important to note that the
pathspec is matched against the path
that is being updated by Unison, not its descendants. For example, if you
set
backup = Name *.txt and then delete a whole directory named
foo containing some text files, these files will not be backed up
because Unison will just check that
foo does not match
*.txt.
Similarly, if the directory itself happened to be called
foo.txt,
then the whole directory and all the files in it will be backed up,
regardless of their names.
Backup files can be stored either
centrally or
locally. This
behavior is controlled by the preference
backuplocation, whose value
must be either
central or
local. (The default is
central.)
When backups are stored locally, they are kept in the same
directory as the original.
When backups are stored centrally, the directory used to hold them is
controlled by the preference
backupdir and the
environment variable
UNISONBACKUPDIR. (The environment variable is
checked first.) If neither of these are set, then the directory
.unison/backup in the user's home directory is used.
The preference
maxbackups controls how many previous versions of
each file are kept (including the current version).
By default, backup files are named
.bak.VERSION.FILENAME,
where
FILENAME is the original filename and
VERSION is the
backup number (1 for the most recent, 2 for the next most recent,
etc.). This can be changed by setting the preferences
backupprefix
and/or
backupsuffix. If desired,
backupprefix may include a
directory prefix; this can be used with
backuplocation = local to put all
backup files for each directory into a single subdirectory. For example, setting
backuplocation = local
backupprefix = .unison/$VERSION.
backupsuffix =
will put all backups in a local subdirectory named
.unison. Also,
note that the string
$VERSION in either
backupprefix or
backupsuffix (it must appear in one or the other) is replaced by
the version number. This can be used, for example, to ensure that backup
files retain the same extension as the originals.
For backward compatibility, the
backups preference is also supported.
It simply means
backup = Name * and
backuplocation = local.
Merging Conflicting Versions
Unison can invoke external programs to merge conflicting versions of a file.
The preference
merge controls this process.
The
merge preference may be given once or several times in a
preference file (it can also be given on the command line, of course, but
this tends to be awkward because of the spaces and special characters
involved). Each instance of the preference looks like this:
merge = <PATHSPEC> -> <MERGECMD>
The
<PATHSPEC> here has exactly the same format as for the
ignore preference (see the
Path specification section). For example,
using “
Name *.txt” as the
<PATHSPEC> tells Unison that this
command should be used whenever a file with extension
.txt needs to
be merged.
Many external merging programs require as inputs not just the two files that
need to be merged, but also a file containing the
last synchronized
version. You can ask Unison to keep a copy of the last synchronized
version for some files using the
backupcurrent preference. This
preference is used in exactly the same way as
backup and its meaning
is similar, except that it causes backups to be kept of the
current
contents of each file after it has been synchronized by Unison, rather than
the
previous contents that Unison overwrote. These backups are kept
on
both replicas in the same place as ordinary backup files—i.e.
according to the
backuplocation and
backupdir preferences.
They are named like the original files if
backupslocation is set to
'central' and otherwise, Unison uses the
backupprefix and
backupsuffix preferences and assumes a version number 000 for these
backups.
The
<MERGECMD> part of the preference specifies what external command
should be invoked to merge files at paths matching the
<PATHSPEC>.
Within this string, several special substrings are recognized; these will be
substituted with appropriate values before invoking a sub-shell to execute
the command.
-
CURRENT1 is replaced by the name of (a temporary copy of)
the local variant of the file.
CURRENT2 is replaced by the name of a temporary
file, into which the contents of the remote variant of the file have
been transferred by Unison prior to performing the merge.
CURRENTARCH is replaced by the name of the backed up copy
of the original version of the file (i.e., the file saved by Unison
if the current filename matches the path specifications for the
backupcurrent preference, as explained above), if one exists.
If no archive exists and CURRENTARCH appears in the
merge command, then an error is signalled.
CURRENTARCHOPT is replaced by the name of the backed up copy
of the original version of the file (i.e., its state at the end of
the last successful run of Unison), if one exists, or the empty
string if no archive exists.
NEW is replaced by the name of a temporary file
that Unison expects to be written by the merge program when it
finishes, giving the desired new contents of the file.
PATH is replaced by the path (relative to the roots of
the replicas) of the file being merged.
NEW1 and NEW2 are replaced by the names of temporary files
that Unison expects to be written by the merge program when it
is only able to partially merge the originals; in this case, NEW1
will be written back to the local replica and NEW2 to the remote
replica; NEWARCH, if present, will be used as the “last common
state” of the replicas. (These three options are provided for
later compatibility with the Harmony data synchronizer.)
To accomodate the wide variety of programs that users might want to use for
merging, Unison checks for several possible situations when the merge
program exits:
-
If the merge program exits with a non-zero status, then merge is
considered to have failed and the replicas are not changed.
- If the file
NEW has been created, it is written back to both
replicas (and stored in the backup directory). Similarly, if just the
file NEW1 has been created, it is written back to both
replicas.
- If neither
NEW nor NEW1 have been created, then Unison
examines the temporary files CURRENT1 and CURRENT2 that
were given as inputs to the merge program. If either has been changed (or
both have been changed in identical ways), then its new contents are written
back to both replicas. If either CURRENT1 or CURRENT2 has
been deleted, then the contents of the other are written back to
both replicas.
- If the files
NEW1, NEW2, and NEWARCH have all
been created, they are written back to the local replica, remote replica,
and backup directory, respectively. If the files NEW1, NEW2 have
been created, but NEWARCH has not, then these files are written back to the
local replica and remote replica, respectively. Also, if NEW1 and
NEW2 have identical contents, then the same contents are stored as
a backup (if the backupcurrent preference is set for this path) to
reflect the fact that the path is currently in sync.
- If
NEW1 and NEW2 (resp. CURRENT1 and
CURRENT2) are created (resp. overwritten) with different contents
but the merge command did not fail (i.e., it exited with status code 0),
then we copy NEW1 (resp. CURRENT1) to the other replica and
to the archive.
This behavior is a design choice made to handle the case where a merge
command only synchronizes some specific contents between two files,
skipping some irrelevant information (order between entries, for
instance). We assume that, if the merge command exits normally, then the
two resulting files are “as good as equal.” (The reason we copy one on
top of the other is to avoid Unison detecting that the files are unequal
the next time it is run and trying again to merge them when, in fact, the
merge program has already made them as similar as it is able to.)
If the
confirmmerge preference is set and Unison is not run in
batch mode, then Unison will always ask for confirmation before
actually committing the results of the merge to the replicas.
A large number of external merging programs are available.
For example, on Unix systems setting the
merge preference to
merge = Name *.txt -> diff3 CURRENT1 CURRENTARCH CURRENT2 -m > NEW
will tell Unison to use the external
diff3 program for merging.
Alternatively, users of
emacs may find the following settings convenient:
merge = Name *.txt -> emacs -q --eval '(ediff-merge-files-with-ancestor
"CURRENT1" "CURRENT2" "CURRENTARCH" nil "NEW")'
(These commands are displayed here on two lines to avoid running off the
edge of the page. In your preference file, each command should be written on a
single line.)
Users running emacs under windows may find something like this useful:
merge = Name * -> C:\Progra~1\Emacs\emacs\bin\emacs.exe -q --eval
"(ediff-files """CURRENT1""" """CURRENT2""")"
Users running Mac OS X (you may need the Developer Tools installed to get
the
opendiff utility) may prefer
merge = Name *.txt -> opendiff CURRENT1 CURRENT2 -ancestor CURRENTARCH -merge NEW
Here is a slightly more involved hack. The
opendiff program can
operate either with or without an archive file. A merge command of this
form
merge = Name *.txt ->
if [ CURRENTARCHOPTx = x ];
then opendiff CURRENT1 CURRENT2 -merge NEW;
else opendiff CURRENT1 CURRENT2 -ancestor CURRENTARCHOPT -merge NEW;
fi
(still all on one line in the preference file!) will test whether an archive
file exists and use the appropriate variant of the arguments to
opendiff.
Ordinarily, external merge programs are only invoked when Unison is
not running in batch mode. To specify an external merge program that
should be used no matter the setting of the
batch flag, use the
mergebatch preference instead of
merge.
Please post suggestions for other useful values of the
merge preference to the unison-users mailing list—we'd like
to give several examples here.
The User Interface
Both the textual and the graphical user interfaces are intended to be
mostly self-explanatory. Here are just a few tricks:
-
By default, when running on Unix the textual user interface will
try to put the terminal into the “raw mode” so that it reads the input a
character at a time rather than a line at a time. (This means you can
type just the single keystroke “
>” to tell Unison to
propagate a file from left to right, rather than “> Enter.”)
There are some situations, though, where this will not work — for
example, when Unison is running in a shell window inside Emacs.
Setting the dumbtty preference will force Unison to leave the
terminal alone and process input a line at a time.
Exit code
When running in the textual mode, Unison returns an exit status, which
describes whether, and at which level, the synchronization was successful.
The exit status could be useful when Unison is invoked from a script.
Currently, there are four possible values for the exit status:
-
0: successful synchronization; everything is up-to-date now.
- 1: some files were skipped, but all file transfers were successful.
- 2: non-fatal failures occurred during file transfer.
- 3: a fatal error occurred, or the execution was interrupted.
The graphical interface does not return any useful information through the
exit status.
Path specification
Several Unison preferences (e.g.,
ignore/
ignorenot,
follow,
sortfirst/
sortlast,
backup,
merge, etc.)
specify individual paths or sets of paths. These preferences share a
common syntax based on regular-expressions. Each preference
is associated with a list of path patterns; the paths specified are those
that match any one of the path pattern.
-
Pattern preferences can be given on the command line,
or, more often, stored in profiles, using the same syntax as other preferences.
For example, a profile line of the form
ignore = pattern
adds pattern to the list of patterns to be ignored.
- Each pattern can have one of three forms. The most
general form is a Posix extended regular expression introduced by the
keyword
Regex. (The collating sequences and character classes of
full Posix regexps are not currently supported).
Regex regexp
For convenience, two other styles of pattern are also recognized:
Name name
matches any path in which the last component matches name, while
Path path
matches exactly the path path.
The name and path arguments of the latter forms of
patterns are not regular expressions. Instead,
standard “globbing” conventions can be used in name and
path:
-
a
* matches any sequence of characters not including /
(and not beginning with ., when used at the beginning of a
name)
- a
? matches any single character except / (and leading
.)
[xyz] matches any character from the set {x,
y, z }
{a,bb,ccc} matches any one of a, bb, or
ccc.
- The path separator in path patterns is always the
forward-slash character “/” — even when the client or server is
running under Windows, where the normal separator character is a
backslash. This makes it possible to use the same set of path
patterns for both Unix and Windows file systems.
Some examples of path patterns appear in the
Ignoring
Paths section.