Fossil

Fossil Quick Start Guide
Login

This is a guide to help you get started using the Fossil Distributed Version Control System quickly and painlessly.

Installing

Fossil is a single self-contained C program. You need to either download a precompiled binary or compile it yourself from sources. Install Fossil by putting the fossil binary someplace on your $PATH.

You can test that Fossil is present and working like this: 

fossil version
This is fossil version 2.25 [8f798279d5] 2024-11-06 12:59:09 UTC

General Work Flow

Fossil works with repository files and check-out directories using a workflow like this:

Fossil can be entirely driven from the command line. Many features can also be conveniently accessed from the built-in web user interface.

The following sections give a brief overview of these operations.

Starting A New Project

To start a new project with Fossil, create a new empty repository: 

fossil init repository-filename

You can name the database anything you like, and you can place it anywhere in the filesystem. The .fossil extension is traditional, but it is only required if you are going to use the fossil server DIRECTORY feature.

Next, do something along the lines of: 

mkdir -p ~/src/project/trunk
cd ~/src/project/trunk
fossil open repository-filename
fossil add foo.c bar.h qux.md
fossil commit

If your project directory already exists, obviating the mkdir step, you will instead need to add the --force flag to the open command to authorize Fossil to open the repo into a non-empty checkout directory. (This is to avoid accidental opens into, for example, your home directory.)

The convention of naming your checkout directory after a long-lived branch name like "trunk" is in support of Fossil's ability to have as many open checkouts as you like. This author frequently has additional checkout directories named ../release, ../scratch, etc. The release directory is open to the branch of the same name, while the scratch directory is used when disturbing one of the other long-lived checkout directories is undesireable, as when performing a bisect operation.

Cloning An Existing Repository

Most fossil operations interact with a repository that is on the local disk drive, not on a remote system. Hence, before accessing a remote repository it is necessary to make a local copy of that repository, a process called "cloning".

This is done as follows: 

fossil clone URL repository-filename

The URL specifies the fossil repository you want to clone. The repository-filename is the new local filename into which the cloned repository will be written. For example, to clone the source code of Fossil itself: 

fossil clone https://fossil-scm.org/ myclone.fossil

If your logged-in username is 'exampleuser', you should see output something like this: 

Round-trips: 8   Artifacts sent: 0  received: 39421
Clone done, sent: 2424  received: 42965725  ip: 10.10.10.0
Rebuilding repository meta-data...
100% complete...
Extra delta compression... 
Vacuuming the database... 
project-id: 94259BB9F186226D80E49D1FA2DB29F935CCA0333
server-id:  016595e9043054038a9ea9bc526d7f33f7ac0e42
admin-user: exampleuser (intial remote-access password is "yoWgDR42iv")>

This exampleuser will be used by Fossil as the author of commits when you checkin changes to the repository. It is also used by Fossil when you make your repository available to others using the built-in server mode by running fossil server and will also be used when running fossil ui to view the repository through the Fossil UI. See the quick start topic for setting up a server for more details.

If the remote repository requires a login, include a userid in the URL like this: 

fossil clone https://remoteuserid@www.example.org/ myclone.fossil

You will be prompted separately for the password. Use "%HH" escapes for special characters in the userid. For example "/" would be replaced by "%2F" meaning that a userid of "Projects/Budget" would become "Projects%2FBudget")

If you are behind a restrictive firewall, you might need to specify an HTTP proxy.

A Fossil repository is a single disk file. Instead of cloning, you can just make a copy of the repository file (for example, using "scp"). Note, however, that the repository file contains auxiliary information above and beyond the versioned files, including some sensitive information such as password hashes and email addresses. If you want to share Fossil repositories directly by copying, consider running the fossil scrub command to remove sensitive information before transmitting the file.

Importing From Another Version Control System

Rather than start a new project, or clone an existing Fossil project, you might prefer to import an existing Git project into Fossil using the fossil import command.

You can even decide to export your project back into git using the fossil git command, which is how the Fossil project maintains its public GitHub mirror. There is no limit to the number of times a tree can be imported and exported between Fossil and git.

The Git fast-export format has become a popular way to move files between version management systems, including from Mercurial. Fossil can also import Subversion projects directly.

Checking Out A Local Tree

To work on a project in fossil, you need to check out a local copy of the source tree. Create the directory you want to be the root of your tree, cd into that directory, and then: 

fossil open repository-filename

For example: 

fossil open ../myclone.fossil
    BUILD.txt
    COPYRIGHT-BSD2.txt
    README.md
      ︙

This leaves you with the newest version of the tree checked out. From anywhere underneath the root of your local tree, you can type commands like the following to find out the status of your local tree: 

fossil info
fossil status
fossil changes
fossil diff
fossil timeline
fossil ls
fossil branch

If you created a new repository using "fossil init" some commands will not produce much output.

Note that Fossil allows you to make multiple check-outs in separate directories from the same repository. This enables you, for example, to do builds from multiple branches or versions at the same time without having to generate extra clones.

To switch a checkout between different versions and branches, use:< 

fossil update
fossil checkout

update honors the "autosync" option and does a "soft" switch, merging any local changes into the target version, whereas checkout does not automatically sync and does a "hard" switch, overwriting local changes if told to do so.

Making and Committing Changes

To add new files to your project or remove existing ones, use these commands: 

fossil add file...
fossil rm file...
fossil addremove file...

The command: 

fossil changes

lists files that have changed since the last commit to the repository. For example, if you edit the file "README.md": 

fossil changes
EDITED     README.md

To see exactly what change was made you can use the command fossil diff: 

fossil diff
Index: README.md
============================================================
--- README.md
+++ README.md
@@ -1,5 +1,6 @@
+Made some changes to the project
# Original text

"fossil diff" shows the difference between your tree on disk now and as the tree was when you last committed changes. If you haven't committed yet, then it shows the difference relative to the tip-of-trunk commit in the repository, being what you get when you "fossil open" a repository without specifying a version, populating the working directory.

To see the most recent changes made to the repository by other users, use "fossil timeline" to find out the most recent commit, and then "fossil diff" between that commit and the current tree: 

fossil timeline
=== 2021-03-28 === 
03:18:54 [ad75dfa4a0] *CURRENT* Added details to frobnicate command (user: user-one tags: trunk) 
=== 2021-03-27 === 
23:58:05 [ab975c6632] Update README.md. (user: user-two tags: trunk) 
     ⋮ 


fossil diff --from current --to ab975c6632
Index: frobnicate.c
============================================================
--- frobnicate.c
+++ frobnicate.c
@@ -1,10 +1,11 @@
+/* made a change to the source file */
# Original text

"current" is an alias for the checkout version, so the command "fossil diff --from ad75dfa4a0 --to ab975c6632" gives identical results.

To commit your changes to a local-only repository: 

fossil commit     (... Fossil will start your editor, if defined)
# Enter a commit message for this check-in. Lines beginning with # are ignored.
#
# user: exampleuser
# tags: trunk
#
# EDITED     README.md
Edited file to add description of code changes
New_Version: 7b9a416ced4a69a60589dde1aedd1a30fde8eec3528d265dbeed5135530440ab

You will be prompted for check-in comments using whatever editor is specified by your VISUAL or EDITOR environment variable. If none is specified Fossil uses line-editing in the terminal.

To commit your changes to a repository that was cloned from a remote repository, you give the same command, but the results are different. Fossil defaults to autosync mode, a single-stage commit that sends all changes committed to the local repository immediately on to the remote parent repository. This only works if you have write permission to the remote respository.

Naming of Files, Checkins, and Branches

Fossil deals with information artifacts. This Quickstart document only deals with files and collections of files, but be aware there are also tickets, wiki pages and more. Every artifact in Fossil has a universally-unique hash id, and may also have a human-readable name.

The following are all equivalent ways of identifying a Fossil file, checkin or branch artifact:

A special convenience branch is "trunk", which is Fossil's default branch name for the first checkin, and the default for any time a branch name is needed but not specified.

This will get you started on identifying checkins. The Checkin Names document is a complete reference, including how timestamps can also be used.

Accessing Your Local Repository's Web User Interface

After you create a new repository, you usually want to do some local configuration. This is most easily accomplished by firing up the Fossil UI: 

fossil ui repository-filename

You can shorten that to just fossil ui if you are inside a checked-out local tree.

This command starts an internal web server, after which Fossil automatically launches your default browser, pointed at itself, presenting a special view of the repository, its web user interface.

You may override Fossil's logic for selecting the default browser so: 

fossil setting web-browser path-to-web-browser

When launched this way, Fossil binds its internal web server to the IP loopback address, 127.0.0.1, which it treats specially, bypassing all user controls, effectively giving visitors the all-powerful Setup capabliity.

Why is that a good idea, you ask? Because it is a safe presumption that only someone with direct file access to the repository database file could be using the resulting web interface. Anyone who can modify the repo DB directly could give themselves any and all access with a SQL query, or even by direct file manipulation; no amount of access control matters to such a user.

(Contrast the many other ways of setting Fossil up as an HTTP server, where the repo DB is on the other side of the HTTP server wall, inaccessible by all means other than Fossil's own mediation. For this reason, the "localhost bypasses access control" policy does not apply to these other interfaces. That is a very good thing, since without this difference in policy, it would be unsafe to bind a fossil server instance to localhost on a high-numbered port and then reverse-proxy it out to the world via HTTPS, a practice this author does engage in, with confidence.)

Once you are finished configuring Fossil, you may safely Control-C out of the fossil ui command to shut down this privileged built-in web server. Moreover, you may by grace of SQLite do this at any time: all changes are either committed durably to the repo DB or rolled back, in their totality. This includes configuration changes.

Sharing Changes

When autosync is turned off, the changes you commit are only on your local repository. To share those changes with other repositories, do: 

fossil push URL

Where URL is the http: URL of the server repository you want to share your changes with. If you omit the URL argument, fossil will use whatever server you most recently synced with.

The push command only sends your changes to others. To Receive changes from others, use pull. Or go both ways at once using sync: 

fossil pull URL
fossil sync URL

When you pull in changes from others, they go into your repository, not into your checked-out local tree. To get the changes into your local tree, use update: 

fossil update VERSION

The VERSION can be the name of a branch or tag or any abbreviation to the 40-character artifact identifier for a particular check-in, or it can be a date/time stamp. (more info) If you omit the VERSION, then fossil moves you to the latest version of the branch you are currently on.

The default behavior is for autosync to be turned on. That means that a pull automatically occurs when you run update and a push happens automatically after you commit. So in normal practice, the push, pull, and sync commands are rarely used. But it is important to know about them, all the same. 

fossil checkout VERSION

Is similar to update except that it does not honor the autosync setting, nor does it merge in local changes - it prefers to overwrite them and fails if local changes exist unless the --force flag is used.

Branching And Merging

Use the --branch option to the commit command to start a new branch at the point of need. (Contrast git.)

To merge two branches back together, first update to the branch you want to merge into. Then do a merge of the other branch that you want to incorporate the changes from. For example, to merge "featureX" changes into "trunk" do this: 

fossil update trunk
fossil merge featureX
# make sure the merge didn't break anything...
fossil commit

The argument to the merge command can be any of the version identifier forms that work for update. (more info.) The merge command has options to cherry-pick individual changes, or to back out individual changes, if you don't want to do a full merge.

The merge command puts all changes in your working check-out. No changes are made to the repository. You must run commit separately to add the merge changes into your repository to make them persistent and so that your coworkers can see them. But before you do that, you will normally want to run a few tests to verify that the merge didn't cause logic breaks in your code.

The same branch can be merged multiple times without trouble. Fossil automatically keeps up with things and avoids conflicts when doing multiple merges. So even if you have merged the featureX branch into trunk previously, you can do so again and Fossil will automatically know to pull in only those changes that have occurred since the previous merge.

If a merge or update doesn't work out (perhaps something breaks or there are many merge conflicts) then you back up using: 

fossil undo

This will back out the changes that the merge or update made to the working checkout. There is also a redo command if you undo by mistake. Undo and redo only work for changes that have not yet been checked in using commit and there is only a single level of undo/redo.

Setting Up A Server

In addition to the inward-facing fossil ui mode covered above, Fossil can also act as an outward-facing web server: 

fossil server repository-filename

Just as with fossil ui, you may omit the repository-filename parameter when running this from within an open check-out.

Unlike fossil ui mode, Fossil binds to all network interfaces by default in this mode, and it enforces the configured role-based access controls. Further, because it is meant to provide external web service, it doesn't try to launch a local web browser pointing to a "Fossil UI" presentation; external visitors see your repository's configured home page instead.

To serve varying needs, there are additional ways to serve a Fossil repo to external users:

…along with several other options.

We recommend that you read the Benefits of a Fossil Server article, because you might need to do this and not yet know it.

HTTP Proxies

If you are behind a restrictive firewall that requires you to use an HTTP proxy to reach the internet, then you can configure the proxy in three different ways. You can tell fossil about your proxy using a command-line option on commands that use the network, sync, clone, push, and pull. 

fossil clone URL --proxy Proxy-URL

It is annoying to have to type in the proxy URL every time you sync your project, though, so you can make the proxy configuration persistent using the setting command: 

fossil setting proxy Proxy-URL

Or, you can set the "http_proxy" environment variable: 

export http_proxy=Proxy-URL

To stop using the proxy, do: 

fossil setting proxy off

Or unset the environment variable. The fossil setting for the HTTP proxy takes precedence over the environment variable and the command-line option overrides both. If you have a persistent proxy setting that you want to override for a one-time sync, that is easily done on the command-line. For example, to sync with a co-worker's repository on your LAN, you might type: 

fossil sync http://192.168.1.36:8080/ --proxy off