Contributed by &a.jdp;. Introduction

CVSup is a software package for distributing and updating source trees from a master CVS repository on a remote server host. The FreeBSD sources are maintained in a CVS repository on a central development machine in California. With CVSup, FreeBSD users can easily keep their own source trees up to date.

CVSup uses the so-called "pull" model of updating. Under the pull model, each client asks the server for updates, if and when they are wanted. The server waits passively for update requests from its clients. Thus all updates are instigated by the client. The server never sends unsolicited updates. Users must either run the CVSup client manually to get an update, or they must set up a cron job to run it automatically on a regular basis.

The term "CVSup", capitalized just so, refers to the entire software package. Its main components are the client "cvsup" which runs on each user's machine, and the server "cvsupd" which runs at each of the FreeBSD mirror sites.

As you read the FreeBSD documentation and mailing lists, you may see references to . Sup was the predecessor to CVSup, and it served a similar purpose. CVSup is in used in much the same way as sup and, in fact, uses configuration files which are backward-compatible with sup's. Sup is no longer used in the FreeBSD project, however, because CVSup is both faster and more flexible. Installation

The easiest way to install CVSup if you are running FreeBSD 2.2 or later is to use either from the FreeBSD or the corresponding , depending on whether you prefer to roll your own or not.

If you are running FreeBSD-2.1.6, you unfortunately cannot use the binary package versions due to the fact that it requires a version of the C library that did not yet exist in FreeBSD-2.1.6. You can easily use , however, just as with FreeBSD 2.2. Simply unpack the tar file, cd to the cvsup subdirectory and type "make install"

Because CVSup is written in , both the package and the port require that the Modula-3 runtime libraries be installed. These are available as the port and the package. If you follow the same directions as for cvsup, these libraries will be compiled and/or installed automatically when you install the CVSup port or package.

The Modula-3 libraries are rather large, and fetching and compiling them is not an instantaneous process. For that reason, a third option is provided. You can get statically linked FreeBSD executables for CVSup from: (client). (server).

Most users will need only the client. These executables are entirely self-contained, and they will run on any version of FreeBSD from FreeBSD-2.1.0 to FreeBSD-current.

In summary, your options for installing CVSup are: FreeBSD-2.2 or later: static binary, port, or package FreeBSD-2.1.6: static binary or port FreeBSD-2.1.5 or earlier: static binary Configuration

CVSup's operation is controlled by a configuration file called the "supfile". Beginning with FreeBSD-2.2, there are some sample supfiles in the directory . These examples are also available from if you are on a pre-2.2 system.

The information in a supfile answers the following questions for cvsup:

In the following sections, we will construct a typical supfile by answering each of these questions in turn. First, we describe the overall structure of a supfile.

A supfile is a text file. Comments begin with "#" and extend to the end of the line. Lines that are blank and lines that contain only comments are ignored.

Each remaining line describes a set of files that the user wishes to receive. The line begins with the name of a "collection", a logical grouping of files defined by the server. The name of the collection tells the server which files you want. After the collection name come zero or more fields, separated by white space. These fields answer the questions listed above. There are two types of fields: flag fields and value fields. A flag field consists of a keyword standing alone, e.g., "delete" or "compress". A value field also begins with a keyword, but the keyword is followed without intervening white space by "=" and a second word. For example, "release=cvs" is a value field.

A supfile typically specifies more than one collection to receive. One way to structure a supfile is to specify all of the relevant fields explicitly for each collection. However, that tends to make the supfile lines quite long, and it is inconvenient because most fields are the same for all of the collections in a supfile. CVSup provides a defaulting mechanism to avoid these problems. Lines beginning with the special pseudo-collection name "*default" can be used to set flags and values which will be used as defaults for the subsequent collections in the supfile. A default value can be overridden for an individual collection, by specifying a different value with the collection itself. Defaults can also be changed or augmented in mid-supfile by additional "*default" lines.

With this background, we will now proceed to construct a supfile for receiving and updating the main source tree of . Which files do you want to receive?

As with sup, the files available via CVSup are organized into named groups called "collections". The collections making up the FreeBSD source tree are described in . In this example, we wish to receive the entire main source tree for the FreeBSD system. There is a single large collection "src-all" which will give us all of that, except the export-controlled cryptography support. Let us assume for this example that we are in the USA or Canada. Then we can get the cryptography code with two additional collections, "src-eBones" and "src-secure". As a first step toward constructing our supfile, we simply list these collections, one per line: src-all src-eBones src-secure

Which version(s) of them do you want?

With CVSup, you can receive virtually any version of the sources that ever existed. That is possible because the cvsupd server works directly from the CVS repository, which contains all of the versions. You specify which one of them you want using the "tag=" and "date=" value fields.

The "tag=" field names a symbolic tag in the repository. There are two kinds of tags, revision tags and branch tags. A revision tag refers to a specific revision. Its meaning stays the same from day to day. A branch tag, on the other hand, refers to the latest revision on a given line of development, at any given time. Because a branch tag does not refer to a specific revision, it may mean something different tomorrow than it means today.

Here are the branch tags that users might be interested in:

Here are the revision tags that users might be interested in:

Be very careful to type the tag name exactly as shown. CVSup cannot distinguish between valid and invalid tags. If you misspell the tag, CVSup will behave as though you had specified a valid tag which happens to refer to no files at all. It will delete your existing sources in that case.

When you specify a branch tag, you normally receive the latest versions of the files on that line of development. If you wish to receive some past version, you can do so by specifying a date with the "date=" value field. The cvsup(1) manual page explains how to do that.

For our example, we wish to receive FreeBSD-current. We add this line at the beginning of our supfile: *default tag=.

There is an important special case that comes into play if you specify neither a "tag=" field nor a "date=" field. In that case, you receive the actual RCS files directly from the server's CVS repository, rather than receiving a particular version. Developers generally prefer this mode of operation. By maintaining a copy of the repository itself on their systems, they gain the ability to browse the revision histories and examine past versions of files. This gain is achieved at a large cost in terms of disk space, however.

Where do you want to get them from?

This one is easy. We use the "host=" field to tell cvsup to get its updates from the primary FreeBSD distribution site, "cvsup.FreeBSD.org": *default host=cvsup.FreeBSD.org

On any particular run of cvsup, we can override this setting on the command line, with "-h hostname".

Where do you want to put them on your own machine?

The "prefix=" field tells cvsup where to put the files it receives. In this example, we will put the source files directly into our main source tree, "/usr/src". The "src" directory is already implicit in the collections we have chosen to receive, so this is the correct specification: *default prefix=/usr

Where should cvsup maintain its status files?

The cvsup client maintains certain status files in what is called the "base" directory. These files help CVSup to work more efficiently, by keeping track of which updates you have already received. We will use the standard base directory, "/usr/local/etc/cvsup": *default base=/usr/local/etc/cvsup

This setting is used by default if it is not specified in the supfile, so we actually do not need the above line.

If your base directory does not already exist, now would be a good time to create it. The cvsup client will refuse to run if the base directory does not exist.

Miscellaneous supfile settings:

There is one more line of boiler plate that normally needs to be present in the supfile: *default release=cvs delete use-rel-suffix compress

"release=cvs" indicates that the server should get its information out of the main FreeBSD CVS repository. This is virtually always the case, but there are other possibilities which are beyond the scope of this discussion.

"delete" gives CVSup permission to delete files. You should always specify this, so that CVSup can keep your source tree fully up to date. CVSup is careful to delete only those files for which it is responsible. Any extra files you happen to have will be left strictly alone.

"use-rel-suffix" is ... arcane. If you really want to know about it, see the cvsup(1) manual page. Otherwise, just specify it and do not worry about it.

"compress" enables the use of gzip-style compression on the communication channel. If your network link is T1 speed or faster, you probably should not use compression. Otherwise, it helps substantially.

Putting it all together:

Here is the entire supfile for our example: *default tag=. *default host=cvsup.FreeBSD.org *default prefix=/usr *default base=/usr/local/etc/cvsup *default release=cvs delete use-rel-suffix compress src-all src-eBones src-secure Running CVSup

You are now ready to try an update. The command line for doing this is quite simple: cvsup supfile

where "supfile" is of course the name of the supfile you have just created. Assuming you are running under X11, cvsup will display a GUI window with some buttons to do the usual things. Press the "go" button, and watch it run.

Since you are updating your actual "/usr/src" tree in this example, you will need to run the program as root so that cvsup has the permissions it needs to update your files. Having just created your configuration file, and having never used this program before, that might understandably make you nervous. There is an easy way to do a trial run without touching your precious files. Just create an empty directory somewhere convenient, and name it as an extra argument on the command line: mkdir /var/tmp/dest cvsup supfile /var/tmp/dest

The directory you specify will be used as the destination directory for all file updates. CVSup will examine your usual files in "/usr/src", but it will not modify or delete any of them. Any file updates will instead land in "/var/tmp/dest/usr/src". CVSup will also leave its base directory status files untouched when run this way. The new versions of those files will be written into the specified directory. As long as you have read access to "/usr/src", you do not even need to be root to perform this kind of trial run.

If you are not running X11 or if you just do not like GUIs, you should add a couple of options to the command line when you run cvsup: cvsup -g -L 2 supfile

The "-g" tells cvsup not to use its GUI. This is automatic if you are not running X11, but otherwise you have to specify it.

The "-L 2" tells cvsup to print out the details of all the file updates it is doing. There are three levels of verbosity, from "-L 0" to "-L 2". The default is 0, which means total silence except for error messages.

There are plenty of other options available. For a brief list of them, type "cvsup -H". For more detailed descriptions, see the manual page.

Once you are satisfied with the way updates are working, you can arrange for regular runs of cvsup using cron(8). Obviously, you should not let cvsup use its GUI when running it from cron. Announcements, Questions, and Bug Reports

Most FreeBSD-related discussion of CVSup takes place on the &a.hackers;. New versions of the software are announced there, as well as on the &a.announce;.

Questions and bug reports should be addressed to the author of the program at .