Diffusion User Guide: Managing Repositories

Phabricator User Documentation (Application User Guides)

Guide to configuring and managing repositories in Diffusion.

Overview

After you create a new repository in Diffusion or select Manage Repository from the main screen if an existing repository, you'll be taken to the repository management interface for that repository.

On this interface, you'll find many options which allow you to configure the behavior of a repository. This document walks through the options.

Basics

The Basics section of the management interface allows you to configure the repository name, description, and identifiers. You can also activate or deactivate the repository here, and configure a few other miscellaneous settings.

Basics: Name

The repository name is a human-readable primary name for the repository. It does not need to be unique

Because the name is not unique and does not have any meaningful restrictions, it's fairly ambiguous and isn't very useful as an identifier. The other basic information (primarily callsigns and short names) gives you control over repository identifiers.

Basics: Callsigns

Each repository can optionally be identified by a "callsign", which is a short uppercase string like "P" (for Phabricator) or "ARC" (for Arcanist).

The primary goal of callsigns is to namespace commits to SVN repositories: if you use multiple SVN repositories, each repository has a revision 1, revision 2, etc., so referring to them by number alone is ambiguous.

However, even for Git and Mercurial they impart additional information to human readers and allow parsers to detect that something is a commit name with high probability (and allow distinguishing between multiple copies of a repository).

Configuring a callsign can make interacting with a commonly-used repository easier, but you may not want to bother assigning one to every repository if you have some similar, templated, or rarely-used repositories.

If you choose to assign a callsign to a repository, it must be unique within an install but do not need to be globally unique, so you are free to use the single-letter callsigns for brevity. For example, Facebook uses "E" for the Engineering repository, "O" for the Ops repository, "Y" for a Yum package repository, and so on, while Phabricator uses "P" and Arcanist uses "ARC". Keeping callsigns brief will make them easier to use, and the use of one-character callsigns is encouraged if they are reasonably evocative.

If you configure a callsign like XYZ, Phabricator will activate callsign URIs and activate the callsign identifier (like rXYZ) for the repository. These more human-readable identifiers can make things a little easier to interact with.

Basics: Short Name

Each repository can optionally have a unique short name. Short names must be unique and have some minor restrictions to make sure they are unambiguous and appropriate for use as directory names and in URIs.

Basics: Description

You may optionally provide a brief (or, at your discretion, excruciatingly long) human-readable description of the repository. This description will be shown on the main repository page.

You can also create a README file at the repository root (or in any subdirectory) to provide information about the repository. These formats are supported:

Basics: Encoding

Before content from the repository can be shown in the web UI or embedded in other contexts like email, it must be converted to UTF-8.

Most source code is written in UTF-8 or a subset of UTF-8 (like plain ASCII) already, so everything will work fine. The majority of repositories do not need to adjust this setting.

If your repository is primarily written in some other encoding, specify it here so Phabricator can convert from it properly when reading content to embed in a webpage or email.

Basics: Dangerous Changes

By default, repositories are protected against dangerous changes. Dangerous changes are operations which rewrite or destroy repository history (for example, by deleting or rewriting branches). Normally, these take the form of git push --force or similar.

It is normally a good idea to leave this protection enabled because most scalable workflows rarely rewrite repository history and it's easy to make mistakes which are expensive to correct if this protection is disabled.

If you do occasionally need to rewrite published history, you can treat this option like a safety: disable it, perform required rewrites, then enable it again.

If you fully disable this at the repository level, you can still use Herald to selectively protect certain branches or grant this power to a limited set of users.

This option is only available in Git and Mercurial, because it is impossible to make dangerous changes in Subversion.

This option has no effect if a repository is not hosted because Phabricator can not prevent dangerous changes in a remote repository it is merely observing.

Basics: Disable Publishing

You can disable publishing for a repository. For more details on what this means, see Diffusion User Guide: Permanent Refs.

This is primarily useful if you need to perform major maintenance on a repository (like rewriting a large part of the repository history) and you don't want the maintenance to generate a large volume of email and notifications. You can disable publishing, apply major changes, wait for the new changes to import, and then reactivate publishing.

Basics: Deactivate Repository

Repositories can be deactivated. Deactivating a repository has these effects:

• the repository will no longer be updated;
• users will no longer be able to clone/fetch/checkout the repository;
• users will no longer be able to push to the repository; and
• the repository will be hidden from view in default queries.

When repositories are created for the first time, they are deactivated. This gives you an opportunity to customize settings, like adjusting policies or configuring a URI to observe. You must activate a repository before it will start working normally.

Basics: Delete Repository

Repositories can not be deleted from the web UI, so this option only gives you information about how to delete a repository.

Repositories can only be deleted from the command line, with bin/remove:

$ ./bin/remove destroy <repository>

This command will permanently destroy the repository. For more information about destroying things, see Permanently Destroying Data.

Policies

The Policies section of the management interface allows you to review and manage repository access policies.

You can configure granular access policies for each repository to control who can view, clone, administrate, and push to the repository.

Policies: View

The view policy for a repository controls who can view the repository from the web UI and clone, fetch, or check it out from Phabricator.

Users who can view a repository can also access the "Manage" interface to review information about the repository and examine the edit history, but can not make any changes.

Policies: Edit

The edit policy for a repository controls who can change repository settings using the "Manage" interface. In essence, this is permission to administrate the repository.

You must be able to view a repository to edit it.

You do not need this permission to push changes to a repository.

Policies: Push

The push policy for a repository controls who can push changes to the repository.

This policy has no effect if Phabricator is not hosting the repository, because it can not control who is allowed to make changes to a remote repository it is merely observing.

You must also be able to view a repository to push to it.

You do not need to be able to edit a repository to push to it.

Further restrictions on who can push (and what they can push) can be configured for hosted repositories with Herald, which allows you to write more sophisticated rules that evaluate when Phabricator receives a push. To get started with Herald, see Herald User Guide.

Additionally, Git and Mercurial repositories have a setting which allows you to Prevent Dangerous Changes. This setting is enabled by default and will prevent any users from pushing changes which rewrite or destroy history.

URIs

The URIs panel allows you to add and manage URIs which Phabricator will fetch from, serve from, and push to.

These options are covered in detail in Diffusion User Guide: URIs.

Limits

The Limits panel allows you to configure limits and timeouts.

Filesize Limit: Allows you to set a maximum filesize for any file in the repository. If a commit creates a larger file (or modifies an existing file so it becomes too large) it will be rejected. This option only applies to hosted repositories.

This limit is primarily intended to make it more difficult to accidentally push very large files that shouldn't be version controlled (like logs, binaries, machine learning data, or media assets). Pushing huge datafiles by mistake can make the repository unwieldy by dramatically increasing how much data must be transferred over the network to clone it, and simply reverting the changes doesn't reduce the impact of this kind of mistake.

Clone/Fetch Timeout: Configure the internal timeout for creating copies of this repository during operations like intracluster synchronization and Drydock working copy construction. This timeout does not affect external users.

Touch Limit: Apply a limit to the maximum number of paths that any commit may touch. If a commit affects more paths than this limit, it will be rejected. This option only applies to hosted repositories. Users may work around this limit by breaking the commit into several smaller commits which each affect fewer paths.

This limit is intended to offer a guard rail against users making silly mistakes that create obviously mistaken changes, like copying an entire repository into itself and pushing the result. This kind of change can take some effort to clean up if it becomes part of repository history.

Note that if you move a file, both the old and new locations count as touched paths. You should generally configure this limit to be more than twice the number of files you anticipate any user ever legitimately wanting to move in a single commit. For example, a limit of 20000 will let users move up to 10,000 files in a single commit, but will reject users mistakenly trying to push a copy of another repository or a directory with a million logfiles or whatever other kind of creative nonsense they manage to dream up.

Branches

The Branches panel allows you to configure how Phabricator interacts with branches.

This panel is not available for Subversion repositories, because Subversion does not have formal branches.

You can configure a Default Branch. This controls which branch is shown by default in the UI. If no branch is provided, Phabricator will use master in Git and default in Mercurial.

Fetch Refs: In Git, if you are observing a remote repository, you can specify that you only want to fetch a subset of refs using "Fetch Refs".

Normally, all refs (refs/*) are fetched. This means all branches, all tags, and all other refs.

If you want to fetch only a few specific branches, you can list only those branches. For example, this will fetch only the branch "master":

refs/heads/master

You can fetch all branches and tags (but ignore other refs) like this:

refs/heads/*
refs/tags/*

This may be useful if the remote is on a service like GitHub, GitLab, or Gerrit and uses custom refs (like refs/pull/ or refs/changes/) to store metadata that you don't want to bring into Phabricator.

Permanent Refs: To learn more about permanent refs, see:

• Diffusion User Guide: Permanent Refs

By default, Phabricator considers all branches to be permanent refs. If you only want some branches to be treated as permanent refs, specify them here.

When specifying branches, you should enter one branch name per line. You can use regular expressions to match branches by wrapping an expression in regexp(...). For example:

Staging Area

The Staging Area panel configures staging areas, used to make proposed changes available to build and continuous integration systems.

For more details, see Harbormaster User Guide.

Automation

The Automation panel configures support for allowing Phabricator to make writes directly to the repository, so that it can perform operations like automatically landing revisions from the web UI.

For details on repository automation, see Drydock User Guide: Repository Automation.

Symbols

The Symbols panel allows you to customize how symbols (like class and function names) are linked when viewing code in the repository, and when viewing revisions which propose code changes to the repository.

To take advantage of this feature, you need to do additional work to build symbol indexes. For details on configuring and populating symbol indexes, see User Guide: Symbol Indexes.

Repository Identifiers and Names

Repositories have several short identifiers which you can use to refer to the repository. For example, if you use command-line administrative tools to interact with a repository, you'll provide one of these identifiers:

$ ./bin/repository update <identifier>

The identifiers available for a repository depend on which options are configured. Each repository may have several identifiers:

• An ID identifier, like R123. This is available for all repositories.
• A callsign identifier, like rXY. This is available for repositories with a callsign.
• A short name identifier, like xylophone. This is available for repositories with a short name.

All three identifiers can be used to refer to the repository in cases where the intent is unambiguous, but only the first two forms work in ambiguous contexts.

For example, if you type R123 or rXY into a comment, Phabricator will recognize them as references to the repository. If you type xylophone, it assumes you mean the word "xylophone".

Only the R123 identifier is immutable: the others can be changed later by adjusting the callsign or short name for the repository.

Commit Identifiers

Diffusion uses repository identifiers and information about the commit itself to generate globally unique identifiers for each commit, like rE12345.

Each commit may have several identifiers:

• A repository ID identifier, like R123:abcdef123....
• A repository callsign identifier, like rXYZabcdef123.... This only works if a repository has a callsign.
• Any unique prefix of the commit hash.

Git and Mercurial use commit hashes to identify commits, and Phabricator will recognize a commit if the hash prefix is unique and sufficiently long. Commit hashes qualified with a repository identifier must be at least 5 characters long; unqualified commit hashes must be at least 7 characters long.

In Subversion, commit identifiers are sequential integers and prefixes can not be used to identify them.

When rendering the name of a Git or Mercurial commit hash, Phabricator tends to shorten it to 12 characters. This "short length" is relatively long compared to Git itself (which often uses 7 characters). See this post on the LKML for a historical explanation of Git's occasional internal use of 7-character hashes:

https://lkml.org/lkml/2010/10/28/287

Because 7-character hashes are likely to collide for even moderately large repositories, Diffusion generally uses either a 12-character prefix (which makes collisions very unlikely) or the full 40-character hash (which makes collisions astronomically unlikely).

Next Steps

Continue by:

• returning to the Diffusion User Guide.