wiki:TracRepositoryAdmin

Version 4 (modified by trac, 3 years ago) ( diff )

--

Repository Administration

Quick start

  • Enable the repository connector(s) for the version control system(s) that you will use.
  • Add repositories through the Repositories admin page, using trac-admin or by editing the [repositories] section of trac.ini.
  • Synchronize the repositories with the cache, if using cached repositories.
  • Configure your repository hooks to synchronize the repository. Alternatively you can synchronize on every request or disable the use of cached repositories, both of which have performance drawbacks and reduced functionality, but are easier to configure.

Enabling the components

Support for version control systems is provided by optional components distributed with Trac, which are disabled by default (since 1.0). Subversion and Git must be explicitly enabled if you wish to use them.

The version control systems can be enabled by adding the following to the [components] section of your trac.ini, or enabling the components through the Plugins admin page.

tracopt.versioncontrol.svn.* = enabled
tracopt.versioncontrol.git.* = enabled

Specifying repositories

Trac supports multiple repositories per environment, and the repositories may be for different version control systems. Each repository must be defined in a repository configuration provider. Repository providers included with Trac are the database store, the trac.ini configuration file and the GitWeb configuration file. Additional providers are available as plugins.

You can define your repositories through a mix of providers, but each repository should only be defined in a single provider. The repository names must be unique across all providers and duplicate names are discarded.

It is possible to define aliases of repositories that act as "pointers" to real repositories. This can be useful when renaming a repository, to avoid breaking links to the old name.

Default Repository

Trac's repositories are listed in the Repository Index when navigating to Browse Source. The default repository is displayed first, followed by the Repository Index. Trac Links without a repository name in the path specification (e.g. [1] rather than [1/repos1]) refer to the default repository. TracLinks for repositories other than the default must include the repository name in the path specification.

From the Repository Admin page, the default repository is specified by leaving the Name attribute empty. From the command line, the default repository is specified using the string (default) or "". In Trac Ini, the default repository is specified by leaving the {name} empty for each {name}.{attribute} option (e.g. .dir = /path/to/dir).

Repository Attributes

There are a number of attributes that can be specified for each repository, and additional attributes may be available through plugins. A repository name and one of the alias or dir attributes are mandatory. All others are optional.

The following attributes are supported:

Attribute Description
alias Defines an alias to a real repository. All Trac Links referencing the alias resolve to the aliased repository. Note that multiple indirection is not supported, so an alias must always point to a real repository. The alias and dir attributes are mutually exclusive.
cached For a version control system that support caching, specifies that repository caching should be used. Defaults to true for version control systems that support caching.
description The text specified in the description attribute is displayed below the top-level entry for the repository in the source browser. It supports Wiki Formatting.
dir The dir attribute specifies the location of the repository in the filesystem. The alias and dir attributes are mutually exclusive.
hidden When set to true, the repository is hidden from the repository index page in the source browser. Browsing the repository is still possible, and links referencing the repository remain valid.
sync_per_request When set to true the repository will be synchronized on every request (implicit synchronization). This is generally not recommended. See repository synchronization for a comparison of explicit and implicit synchronization. The attribute defaults to false.
type The type attribute specifies the version control system used by the repository. Trac provides support for Subversion and Git, and plugins add support for several other systems. If type is not specified, it defaults to the value of the [versioncontrol] default_repository_type option.
url The url attribute specifies the root URL to be used for checking out from the repository. When specified, a "Repository URL" link is added to the context navigation links in the source browser, that can be copied into the tool used for creating the working copy.

Scoped Repository

For some version control systems, it is possible to specify not only the path to the repository in the dir attribute, but also a scope within the repository. Trac will then only show information related to the files and changesets below that scope. The scope is specified by appending a path that is relative to the repository root. The Subversion backend for Trac supports this.

For example, assume a repository at filesystem path /var/svn/repos1 with several directories at the root of the repository: /proj1, /proj2, etc. The following configuration would scope the repository to /proj1:

proj1.dir = /var/svn/repos1/proj1
proj1.type = svn

For other repository types, check the corresponding plugin's documentation.

In the database

Repositories can also be specified in the database, using either the Repositories admin page under Version Control, or the trac-admin $ENV repository commands.

The admin panel shows the list of all repositories defined in the Trac environment. It allows adding repositories and aliases, editing repository attributes and removing repositories. Note that repositories defined in trac.ini are displayed but cannot be edited.

The following trac-admin commands can be used to perform repository operations from the command line.

repository add <repos> <dir> [type]
Add a repository <repos> located at <dir>, and optionally specify its type.
repository alias <name> <target>
Create an alias <name> for the repository <target>.
repository remove <repos>
Remove the repository <repos>.
repository set <repos> <key> <value>
Set the attribute <key> to <value> for the repository <repos>.

Note that the default repository has an empty name, so it will need to be quoted when running trac-admin from a shell. Alternatively, the name (default) can be used instead, for example when running trac-admin in interactive mode.

In trac.ini

Repositories and repository attributes can be specified in the [repositories] section of trac.ini. Every attribute consists of a key structured as {name}.{attribute} and the corresponding value separated with an equal sign (=). The name of the default repository is empty.

The main advantage of specifying repositories in trac.ini is that they can be inherited from a global configuration. Cached repositories defined in trac.ini at the time of environment initialization will be automatically synchronized if the repository's connector is enabled. One drawback is that due to limitations in the ConfigParser class used to parse trac.ini, the repository name is always lowercase.

The following example defines two Subversion repositories named project and lib, and an alias to project as the default repository. This is a typical use case where a Trac environment previously had a single repository (the project repository), and was converted to multiple repositories. The alias ensures that links predating the change continue to resolve to the project repository.

[repositories]
project.dir = /var/repos/project
project.description = This is the ''main'' project repository.
project.type = svn
project.url = http://example.com/svn/project
project.hidden = true

lib.dir = /var/repos/lib
lib.description = This is the secondary library code.
lib.type = svn
lib.url = http://example.com/svn/lib

.alias = project

Note that name.alias = target makes name an alias for the target repo, not the other way around.

In GitWeb

GitWeb is a CGI script that comes with Git for web-based visualization of repositories. Trac can read the gitweb-formatted project.lists file. The configuration is done through the [gitweb-repositories] section of trac.ini.

Repository caching

Caching improves the performance browsing the repository, viewing logs and viewing changesets. Cached repositories must be synchronized, using either explicit or implicit synchronization. When searching changesets, only cached repositories are searched.

Repositories that support caching are cached by default. The Subversion and Git backends support caching. The Mercurial plugin does not yet support caching (#8417). To disable caching, set the cached attribute to false.

After adding a cached repository, the cache must be populated with the trac-admin $ENV repository resync command.

repository resync <repos>
Re-synchronize Trac with a repository.

Repository synchronization

Either explicit or implicit synchronization can be used. Implicit synchronization is easier to configure, but may result in noticeably worse performance. The changeset added and modified events can't be triggered with implicit synchronization, so the commit ticket updater won't be available.

Explicit synchronization

This is the preferred method of repository synchronization. It requires adding a call to trac-admin in the post-commit hook of each repository. Additionally, if a repository allows changing revision metadata, a call to trac-admin must be added to the post-revprop-change hook as well.

changeset added <repos> <rev> […]
Notify Trac that one or more changesets have been added to a repository.
changeset modified <repos> <rev> […]
Notify Trac that metadata on one or more changesets in a repository has been modified.

The <repos> argument can be either a repository name (use "(default)" for the default repository) or the path to the repository.

Note that you may have to set the environment variable PYTHON_EGG_CACHE to the same value as was used for the web server configuration before calling trac-admin, if you changed it from its default location. See TracPlugins for more information.

Subversion

Using trac-svn-hook

In a Unix environment, the simplest way to configure explicit synchronization is by using the contrib/trac-svn-hook script. trac-svn-hook starts trac-admin asynchronously to avoid slowing the commit and log editing operations. The script comes with a number of safety checks and usage advice. Output is written to a log file with prefix svn-hooks- in the environment log directory, which can make configuration issues easier to debug.

There's no equivalent trac-svn-hook.bat for Windows yet, but the script can be run by Cygwin's bash. The documentation header of trac-svn-hook contains a Cygwin configuration example.

Follow the help in the documentation header of the script to configure trac-svn-hook. You'll need to minimally set the TRAC_ENV variable, and may also need to set TRAC_PATH and TRAC_LD_LIBRARY_PATH for a non-standard installation or a virtual environment.

Configuring the hook environment variables is even easier in Subversion 1.8 and later using the hook script environment configuration. Rather than directly editing trac-svn-hook to set the environment variables, or exporting them from the hook that invokes trac-svn-hook, they can be configured through the repository conf/hooks-env file.

Here is an example, using a Python virtual environment at /usr/local/venv:

[default]
TRAC_ENV=/var/trac/project-1
TRAC_PATH=/usr/local/venv/bin
Writing Your Own Hook Script

The following examples are complete post-commit and post-revprop-change scripts for Subversion. They should be edited for the specific environment, marked executable (where applicable) and placed in the hooks directory of each repository. On Unix (post-commit):

#!/bin/sh
export PYTHON_EGG_CACHE="/path/to/dir"
/usr/bin/trac-admin /path/to/env changeset added "$1" "$2"

Adapt the path to the actual location of trac-admin. On Windows (post-commit.cmd):

@C:\Python26\Scripts\trac-admin.exe C:\path\to\env changeset added "%1" "%2"

The post-revprop-change hook for Subversion is very similar. On Unix (post-revprop-change):

#!/bin/sh
export PYTHON_EGG_CACHE="/path/to/dir"
/usr/bin/trac-admin /path/to/env changeset modified "$1" "$2"

On Windows (post-revprop-change.cmd):

@C:\Python26\Scripts\trac-admin.exe C:\path\to\env changeset modified "%1" "%2"

The Unix variants above assume that the user running the Subversion commit has write access to the Trac environment, which is the case in the standard configuration where both the repository and Trac are served by the web server. If you access the repository through another means, for example svn+ssh://, you may have to run trac-admin with different privileges, for example by using sudo.

See the section about hooks in the Subversion book for more information. Other repository types will require different hook configuration.

Git

Git hooks can be used in the same way for explicit syncing of Git repositories.

If your repository is one that only gets pushed to, add the following to the hooks/post-receive file in the repo:

#!/bin/sh
tracenv=/path/to/env  # set to your Trac environment's path
repos=                # set to your repository's name
while read oldrev newrev refname; do
    if [ "$oldrev" = 0000000000000000000000000000000000000000 ]; then
        git rev-list --reverse "$newrev" --
    else
        git rev-list --reverse "$newrev" "^$oldrev" --
    fi | xargs trac-admin "$tracenv" changeset added "$repos"
done

The repos variable is the repository name (use "(default)" for the default repository).

Alternatively, if your git repository is one that gets committed to directly on the machine that hosts Trac, add the following to the hooks/post-commit file in your Git repository:

#!/bin/sh
tracenv=/path/to/env  # set to your Trac environment's path
repos=                # set to your repository's name
REV=$(git rev-parse HEAD)
trac-admin "$tracenv" changeset added "$repos" $REV

The post-commit hook will do nothing if you only update the repository by pushing to it.

Be sure to set the hook scripts as executable.

Mercurial

For Mercurial, add the following entries to the .hgrc file of each repository accessed by Trac (if TracMercurial is installed in a Trac plugins directory, download hooks.py and place it somewhere accessible):

[hooks]
; If mercurial-plugin is installed globally
commit = python:tracext.hg.hooks.add_changesets
changegroup = python:tracext.hg.hooks.add_changesets

; If mercurial-plugin is installed in a Trac plugins directory
commit = python:/path/to/hooks.py:add_changesets
changegroup = python:/path/to/hooks.py:add_changesets

[trac]
env = /path/to/env
trac-admin = /path/to/trac-admin

Per-request synchronization

If the post-commit hooks are not available, the environment can be set up for per-request synchronization. The sync_per_request attribute for each repository in the database and in trac.ini must be set to true.

Note that in this case, the changeset listener extension point is not called, and therefore plugins that depend on the changeset added and modified events won't work correctly. For example, automatic changeset references cannot be used with implicit synchronization.

Automatic changeset references in tickets

You can automatically add a reference to the changeset as a ticket comment whenever changes are committed to the repository. The description of the commit needs to contain one of the following patterns:

  • Refs #123 - to reference this changeset in #123 ticket
  • Fixes #123 - to reference this changeset and close #123 ticket with the default status fixed

This functionality requires installing a post-commit hook as described in explicit synchronization, and enabling the optional commit updater components through the Plugins admin panel or by adding the following line to the [components] section of your trac.ini:

tracopt.ticket.commit_updater.* = enabled

For more information, see the documentation of the CommitTicketUpdater component in the Plugins admin panel and the CommitTicketUpdater page.

Troubleshooting

My trac-post-commit-hook doesn't work anymore

You must now use the optional components from tracopt.ticket.commit_updater.*, which you can activate through the Plugins admin page, or by directly modifying the [components] section in the trac.ini. Be sure to use explicit synchronization.

See CommitTicketUpdater#Troubleshooting for more troubleshooting tips.

Git control files missing

If your repository is not browseable and you find a message in the log that looks like:

2017-08-08 10:49:17,339 Trac[PyGIT] ERROR: GIT control files missing in '/path/to/git-repository'
2017-08-08 10:49:17,339 Trac[git_fs] ERROR: GitError: GIT control files not found, maybe wrong directory?

First check that the path to your repository is correct. If the path is correct, you may have a permission problem whereby the web server cannot access the repository. You can use Git to verify the repository. On a Debian-like Linux OS, the following command should help:

$ sudo -u www-data git --git-dir=/path/to/git-repository fsck

On other platforms you may need to modify the command to use the user under which the webserver runs.

Note: See TracWiki for help on using the wiki.