gitolite/doc/gitolite.conf.mkd

# F=conf the access control file `gitolite.conf`

Gitolite has an advanced access control language that is designed to be
powerful but easy to use.  Other objectives were that it should be even easier
to read, review and audit the rules, and it should scale to thousands of repos
and users.  There was also, in the author's mind, a desperate need to create
something as different as possible from the brain-dead, nausea-inducing
"Windows INI" style syntax that some other popular tools seem to favour.

This document describes the syntax and semantics of the access control rules
and other configuration directives in the `gitolite.conf` file.

## #syntax syntax

In general, everything is **space separated**; there are no commas,
semicolons, etc., in the syntax.

**Comments** are in the usual shell-ish style.

**User names** and **repo names** are as simple as possible; they must start
with an alphanumeric, but after that they can also contain `.`, `_`, or `-`.

Usernames can optionally be followed by an `@` and a domainname containing at
least one `.` (this allows you to use an email address as someone's username).
Reponames can contain `/` characters (this allows you to put your repos in a
tree-structure for convenience)

### continuation lines

There are no continuation lines -- gitolite does not process C-style
backslash-escaped newlines as anything special.  However, the section on
"groups" will tell you how you can break up large lists of names in a group
definition into multiple lines.

### include files

Gitolite allows you to break up the configuration into multiple files and
include them in the main file for convenience.

    include     "foo.conf"

will include the contents of the file "foo.conf" from the same directory as
the main config file.  You can also use an absolute path if you like, although
in the interests of cloning the admin-repo sanely you should avoid doing this!

You can also use a glob, as in:

    include     "*.conf"

which will include all the ".conf" files from the directory in which the main
config file exists.

Files that have been already processed once are skipped, with a warning.

<font color="gray">Advanced users: `subconf`, a command that is very closely
related to `include`, is documented [here][subconf].</font>

## F=bac basic access control

Here's a very basic set of rules:

    repo    gitolite-admin
            RW+     =   sitaram

    repo    testing
            RW+     =   @all

    repo    gitolite simplicity
            RW+     =   sitaram dilbert
            RW      =   alice ashok
            R       =   wally

It should be easy to guess what most of this means:

  * `R` means "read" permission
  * `RW` means "read and write", but no rewind
  * `RW+` means "read and write", with rewind allowed

A "rewind" is more often called a "non-fast forward push"; see git docs for
what that is.  The `+` was chosen because it is part of the "git push" syntax
for non-ff pushes.

Note that *tags* are generally considered "write once", so they are treated
specially by gitolite: you need rewind permission (`RW+`) to *overwrite* a
tag with a new value.

In a later section you'll see some more advanced permissions.

### how rules are matched

It's important to understand that there're two levels at which access control
happens.  Please see [this][2levels] for details, especially about the first level
check.  Much of the complexity applies only to the second level check, so that
is all we will be discussing here.  This check is done by the update hook, and
determines whether the push succeeds or fails.

For basic permissions like this, matching is simple.  Gitolite already knows:

  * the user
  * the repo
  * the branch or tag ("ref") being pushed
  * whether it is a normal (ff) push or a rewind (non-ff) push.

Gitolite goes down the list of rules matching the user, repo, and the ref.
The first matching rule that has the permission you're looking for (`W` or
`+`), results in success.  A fallthrough results in failure.

### branches, tags, and specifying "refex"es

One of the original goals of gitolite was to allow access control at the
branch/tag (aka "ref") level.  The git source code contains a sample update
hook that has the following in it:

    # from Documentation/howto/update-hook-example.txt:

    refs/heads/master	    junio
    +refs/heads/pu          junio
    refs/heads/cogito$      pasky
    refs/heads/bw/.*        linus
    refs/heads/tmp/.*       .*
    refs/tags/v[0-9].*      junio

If you did this in gitolite, this is what the equivalents would be:

    repo    git
            RW      master$             =   junio   # line 1
            RW+     pu$                 =   junio   # line 2
            RW      cogito$             =   pasky   # line 3
            RW      bw/                 =   linus   # line 4
            RW      tmp/                =   @all    # line 5
            RW      refs/tags/v[0-9]    =   junio   # line 6

The following points will help you understand these rules.  (Git recap:
branches and tags together are called "ref"s in git.  A branch ref usually
looks like "refs/heads/foo", while a tag ref looks like "refs/tags/bar")

  * the general syntax of a paragraph of rules is:

        # start line:
        repo [one or more repos and/or repo groups]
        # followed by one or more permissions lines:
        [permission] [zero or more refexes] = [one or more users]

  * a **refex** is a *perl regex* that matches a ref.  When you try to push a
    commit to a branch or a tag, that "ref" is matched against the refex part
    of each rule.

  * if the refex does not start with `refs/`, gitolite assumes a prefix of
    `refs/heads/`.  This is useful because *branch* matching is the most
    common case, as you can see this applies to lines 1 through 5 here.

  * if no refex appears, the rule applies to all refs in that repo (as if you
    had specified `refs/.*` as the refex).

  * refexes are prefix-matched (they are internally anchored with `^` before
    being used).  This means only the beginning of the actual ref needs to
    match the refex, unless the refex has an explicit `$` meta-character at
    the end (like the first 3 lines in our example do).

    Line 5, for instance, allows anyone to push a branch inside the "tmp/"
    namespace, while line 6 provides the ability to push version tags; "v1",
    "v1.0", "v2.0rc1", all match the criterion specified by `v[0-9]` because
    this is a prefix match only.

### groups

Gitolite allows you to define **groups** of repos. users, or even refexes.  A
group is semantically (but *not* syntactically) like a `#define` in C.  Here
is an example of each kind:

    @oss_repos  =   gitolite linux git perl rakudo entrans vkc
    @staff      =   sitaram some_dev another-dev
    @important  =   master$ QA_done refs/tags/v[0-9]

The syntax of a group definition is simply:

    @groupname = [one or more names]

A group can *accumulate* values.  For example:

    @staff      =   sitaram some_dev another-dev
    @staff      =   au.thor

is the same as

    @staff      =   sitaram some_dev another-dev au.thor

This is more convenient than continuation lines, because it allows you to add
to a group anywhere.  Many people generate their gitolite.conf itself from
some *other* database, and it is very useful to be able to do this sort of
thing.

Groups can include other groups, and the included group will be expanded to
whatever value it *currently* has:

    @staff      =   sitaram some_dev another-dev    # line 1
    @staff      =   au.thor                         # line 2
    @interns    =   indy james                      # line 3
    @alldevs    =   bob @interns @staff             # line 4

"@alldevs" expands to 7 names now.  However, remember that the config file is
parsed in a single-pass, so later *additions* to a group name cannot affect
earlier *uses* of it.  If you moved line 2 to the end, "@alldevs" would only
have 6 names in it.

#### the special `@all` group

There's a special group called `@all` that includes all authenticated users
when used as a username; you've seen examples of it earlier.

Advanced users: also see the entry for `GL_ALL_INCLUDES_SPECIAL` in the
documentation for [`~/.gitolite.rc`][rc].

When used as a reponame, it includes all repos.

### F=rpr_ side note: "R" permissions for refs

You can control "read" access only at the repo level, not at the branch level.
For example, this **won't** limit Wally to reading only the master branch:

    repo foo
        R master    =   wally       # WILL NOT DO WHAT YOU THINK IT DOES!!

and this **won't** prevent him from reading it:

    repo foo
        - master    =   wally       # WILL NOT DO WHAT YOU THINK IT DOES!!
        R           =   wally

This (inability to distinguish one ref from another during a read operation)
is a git issue, not a gitolite issue.

There are 3 ways around this, though:

  * switch to gerrit, which has its own git stack, its own sshd, and God knows
    what else.  All written in Java, the COBOL of the internet era ;-)
  * bug the git people to add this feature in ;-)
  * use a separate repo for Wally.

Using separate repos is not that hard with gitolite.  Here's how to maintain a
[partial copy][partialcopy] of the main repo and keep it synced (while not
allowing the secret branches into it).

## F=aac advanced access control

The previous section is sufficient for most common needs, but gitolite can go
a lot further than that.

### #deny "deny" rules

#### warnings and required reading

Gitolite performs access checks at 2 levels.  The first check is performed for
both read *and* write operations, while the second one happens only for write
operations.

**Required reading**: [this section][2levels] of the documentation.

**Warning**: When using deny rules, the order of your rules matters, where
earlier it did not.  If you're just starting to add a deny rule to an existing
ruleset, it's a good idea to review the entire ruleset once, to make sure
you're doing it right.

#### "deny" rules for refs in a repo

You can use "deny" rules for the second check, to prevent people pushing
branches or tags that they should not be allowed to.

Take a look at the following snippet, which *seems* to say that "bruce" can
write versioned tags (anything containing `refs/tags/v[0-9]`), but the other
staffers can't:

        @staff = bruce whitfield martin
                [... and later ...]
        RW refs/tags/v[0-9]     = bruce
        RW refs/tags            = @staff

But that's not how the matching works.  As long as any refex matches the
refname being updated, it's a "yes".  Since the second refex (which says
"anything containing `refs/tags`") is a superset of the first one, it lets
anyone on `@staff` create versioned tags, not just Bruce.

So how do we say "these people can create any tags except tags that look like
this pattern"?

One way to fix this is to allow "deny" rules.  We make a small addition to the
permissions syntax, and define a more rigorous, ordered, interpretation.

Let's recap the **existing semantics**:

>   The first matching refex that has the permission you're looking for (`W`
>   or `+`), results in success.  A fallthrough results in failure.

Here are the **new semantics**, with changes from the "main" one in bold:

>   The first matching refex that has the permission you're looking for (`W`
>   or `+`) **or a minus (`-`)**, results in success **or failure,
>   respectively**.  A fallthrough **also** results in failure.

So the example we started with becomes, if you use "deny" rules:

        RW refs/tags/v[0-9]     = bruce
        -  refs/tags/v[0-9]     = @staff
        RW refs/tags            = @staff

And here's how it works:

  * for non-version tags, only the 3rd rule matches, so anyone on staff can
    push them
  * for version tags by bruce, the first rule matches so he can push them
  * for version tags by staffers *other than bruce*, the second rule matches
    before the third one, and it has a `-` as the permission, so the push
    fails

#### "deny" rules for the entire repo

The previous section described deny rules for the second check, which is a
fairly common need.  However, gitolite does not process deny rules for the
first check -- it's usually simple enough to make sure your config file does
not allow the particular user/repo combindation at all.

But there's one case where this becomes cumbersome: when you want to create
exceptions to uses of `@all`.

For example, if you want gitweb to show `@all` repos except the special
'gitolite-admin' repo, you must manually (and laboriously) maintain a list of
all your repos (except gitolite-admin) and use that in place of `@all`.  Oh
joy...

But now you can do this:

    repo gitolite-admin
        -   =   gitweb daemon
        [... other access rules ...]
        config gitolite-options.deny-repo = 1

    repo @all
        R   =   gitweb daemon

Here are some notes on how/why this works:

  * The order is important -- the deny rule must come first.

  * The 'config' line is what tells git to behave differently, i.e., apply
    deny rules for the first check also.

  * Since there is no "ref" to match against the refexes in the rules,
    gitolite just ignores the refexes, and simply looks at the permission (R,
    RW, "-", etc) and the user list.

### #rwcd creating and deleting branches

Since the beginning of gitolite, `RW` gave the ability, not only to update,
but to *create* a branch (that matched the refex).  Similarly, `RW+` meant
being able to not only rewind, but also delete a ref.  Conceptually, a rewind
is almost the same as a delete+push (the only difference I can see is if you
had core.logAllRefUpdates set, which is *not* a default setting).

However, there seem to be cases where it is useful to distinguish these cases.
Arguments can be made on all sides if you're dealing with new users, so
gitolite supports that.

We'll look at the delete/rewind case in detail first:

  * if the rules for a repo do not contain a `D` anywhere, then `RW+` will
    allow both rewind and delete operations.  Apart from being more convenient
    if you don't need this separation, this also ensures backward
    compatibility for setups created before this separation feature was added
    to gitolite).

  * if, however, *any* of the rules for a repo contains a `D` (example: `RWD`,
    `RW+D`, etc) then `RW+` by itself will permit only a rewind, not a delete

The same thing applies to create/push, where if you have permissions like
`RWC` or `RW+C` anywhere in that repo, a simple `RW` or `RW+` can no longer
*create* a new ref.

You can combine the `C` and `D` also.  Thus, the set of permissions you now
know about are, in regex syntax: `R|RW+?C?D?`.  See a later section for the
full set of permissions possible.

Some usage hints:

  * if you find that `RW+` no longer allows creation/deletion but you can't
    see a `C`/`D` permission in the rules, remember that gitolite allows a
    repo config to be specified in multiple places for convenience, included
    delegated or included files.  Be sure to search everywhere :)

  * a quick way to make this the default for *all* your repos is:

        repo @all
            RWCD    dummy-branch    =   foo

    where foo can be either the administrator, or if you can ignore the
    warning message when you push, a non-existant user.

## summary: permissions

The full set of permissions, in regex syntax: `-|R|RW+?C?D?`.  This expands to
one of `-`, `R`, `RW`, `RW+`, `RWC`, `RW+C`, `RWD`, `RW+D`, `RWCD`, or
`RW+CD`.  and by now you know what they all mean.

[Side note: There is one more very important permissions to be dealt with --
the standalone `C`, which is not really a "ref" level permission and can be
found in doc/wildcard-repositories.mkd.]

## F=confother__ other tips

### personal branches

Gitolite lets you define a "personal" or "scratch" namespace prefix for each
developer.  See [here][pers] for details.

### #ruleaccum rule accumulation

(Also see [this][ruleaccum2] for a different example that may be more
intuitive for some people).

Gitolite lets you specify access rules for a repo in bits and pieces, and
accumulates them in the same sequence they were given.  This is very
convenient.  Let's say you have a mix of open source and closed source
projects, and "bosses" should have read access to all projects, and everyone
should have read access to open source projects.  Assuming the appropriate
group definitions, this would work:

    # all bosses have read access to all projects
    repo @open @closed @topsecret
        R   =   @bosses

    # everyone has read access to "open" projects
    repo @open
        R   =   @bosses @devs @interns

If you notice that `@bosses` are given read access to `@open` via both rules,
don't worry that this causes some duplication or inefficiency.  It doesn't :-)

Elsewhere in the file, you would specify access for individual repos (like RW,
RW+, etc).  Gitolite combines all of these access rules, maintaining the
textual order in which they occur, when authorising a push.

And although this example used groups, you can use reponames as well, or mix
and match them.  You can even distribute rulesets across multiple "include"
files if you wish.

Just remember that if you use [deny rules][deny] anywhere then the *order of the
rules matters*!

This feature also helps people who generate their gitolite.conf itself from
some *other* database -- it allows them much more flexibility in how they
generate rules.

### #gwd specifying gitweb and daemon access

Gitolite allows you to specify access for git-daemon and gitweb.  This is a
feature that I personally do not use (corporate environments don't like
unauthenticated access of any kind to any repo!), but someone wanted it, so
here goes.

Gitolite has two pre-defined, "special", usernames: `daemon` and `gitweb`.

To make a repo or repo group accessible via "git daemon", just give read
permission to the special user "daemon".  Similarly, give read permission to
`gitweb` to allow the gitweb CGI to show the repo.  Something like this:

    repo    foo bar baz
        R   =   gitweb daemon

This gives you a quick way to offer multiple repos up for gitweb and/or daemon
access.

However, **setting a description** for the project also enables gitweb
permissions so you can do it that way if you want.  Of course in this case you
have to deal with each repo separately.  Add lines like this to gitolite.conf:

    foo = "some description"
    bar = "some other description"
    baz = "yet another description"

You can also **specify an owner** for gitweb to show, if you like; for example
I might use:

    gitolite "Sitaram Chamarty" = "fast, secure, fine-grained, access control for git"

These lines are standalone, so you can add them anywhere in the conf file.

Note that gitolite does **not** install or configure gitweb/git-daemon -- that
is a one-time setup you must do separately.  All gitolite does is:

  * for daemon, create the file `git-daemon-export-ok` in the repository
  * for gitweb, add the repo (plus owner name, if given) to the list of
    projects to be served by gitweb (see the config file variable
    `$PROJECTS_LIST`, which should have the same value you specified for
    `$projects_list` when setting up gitweb)
  * put the description, if given, in `$repo/description`

The "compile" script will keep these files consistent with the config settings
-- this includes removing such settings/files if you remove "read" permissions
for the special usernames or remove the description line.

Please **note** that giving permissions to these special users via `@all`
(that is, using either `repo @all` or `R = @all`), will not work unless you
set the rc-file variable `$GL_ALL_INCLUDES_SPECIAL` to `1`.  Also, **NOTE**
that giving them read access to `repo @all` means the `gitolite-admin` repo is
also accessible.  **It is upto you to decide if that is OK in your
environment**.

### #rsgc repo specific `git config` commands

(Thanks to teemu dot matilainen at iki dot fi)

>   ----

>   **Note**: this won't work unless the rc file has the right settings;
>   please see `$GL_GITCONFIG_KEYS` in the [rc file doc][rc] for details and
>   security information.

>   ----

Sometimes you want to specify `git config` settings for your repos.

For example, say you have a custom post-receive hook that sends an email when
a push happens, and this hook looks in the config for whom to send the email
to, etc.

You can set these git config values within a "repo" paragraph:

    repo gitolite
        config hooks.mailinglist = gitolite-commits@example.tld
        config hooks.emailprefix = "[gitolite] "
        config foo.bar = ""
        config foo.baz =

The syntax is simple:

    config sectionname.keyname = [optional value_string]

This does either a plain "git config section.key value" (for the first 3
examples above) or "git config --unset-all section.key" (for the last
example).  Other forms of the `git config` command (`--add`, the
`value_regex`, etc) are not supported.

>   ----

>   **WARNING**: simply deleting the config line from the `conf/gitolite.conf`
>   file will *not* delete the variable from `repo.git/config`.  The syntax in
>   the last example is the *only* way to make gitolite execute a
>   `--unset-all` operation on the given key.

>   ----

You can repeat the 'config' line as many times as you like, and the last
occurrence will be the one in effect.  This allows you to override settings
just for one project, as in this example:

    repo @all
        config gitolite.mirror.master = "frodo"
        config gitolite.mirror.slaves = "sam gollum"

    repo top-secret-project
        # only sam, because we don't trust gollum
        config gitolite.mirror.slaves = "sam"

The "delete config variable" syntax can also be used, if you wish:

    repo highlander     # there can be only one!
        config gitolite.mirror.master =
        config gitolite.mirror.slaves =

As you can see, the general idea is to place the most generic ones (`repo
@all`, or repo patterns like `repo foo.*`) first, and place more specific ones
later to override the generic settings.