- SUPPORTED VERSION CONTROL SYSTEMS
- SEE ALSO
kgb-client - relay commits to KGB servers
- kgb-client --conf /path/to/config [other-option ...]
- kgb-client --uri http://some.server:port/service --password password --repo-id repository --timeout timeout-in-seconds
- kgb-client option... /svn/repo revision
- kgb-client option... old-rev new-rev ref-name
kgb-client is the client counterpart of kgb-bot(1). Intented usage is as a hook in your version control system, executed after the repository gets updated. It analyzes the commit(s) and then relays the information about the repository, branch, author, modified files and change log to the KGB server, whch will show it on IRC.
Configuration options (except --conf) may be specified both in the configuration file and on the command line. Usually you want to have all the options in a configuration file, because having passwords on the command line is insecure. The configuration file also gives more control, for example it supports multple servers and multiple ways of detection of branch and module names.
The configration file is in YAML format. Unless noted otherwise, all the options below can be used on the command line if prepended with two dashes. An example configuration file is shipped with the distribution.
- repository type
Specifies the type of the repository kgb-client shall be working with. Currently defaults to
- repo-id repository name
Short repository identifier. Will be used for identifying the repository to the KGB daemon, which will also use this for IRC notifications. Mandatory.
- uri URI
URI of the KGB server. Something like
- proxy URI
URI of the SOAP proxy. If not given, it is the value of the uri option, with
- password password
Password for authentication to the KGB server.
- timeout seconds
Timeout for server communication. Default is 15 seconds, as we want instant IRC and commit response.
Only available in the configuration file.
An array of servers, each described using uri, proxy, password and timeout options. When several servers are configured, kgb-client chooses one randomly. If a given server times out or there is another problem with communication, kgb-client tries another server.
The top-level uri, proxy, password and timeout options are treated as describing an extra server to the servers described in servers array.
The password and timeout options default too the top-level options of the same name.
Makes the whole process more verbose.
Sometimes development is done in multiple branches. Simetimes, a project consists of multiple sub-projects or modules. It is nice to have the module and branc highlighted in notifications. There are two options to help determining the module and branch names from a list of changes.
These options are mainly useful when using Subversion. Git commits carry implicit branch information and chances are that sub-projects use separate Git repositories.
A list of regular expressions that serve for detection of branch and module of commits. Each item from the list is tried in turn, until an item is found that matches all the paths that were modified by the commit. Regular expressions must have two captures: the first one giving the branch name, and the second one giving the module name.
All the paths that were modified by the commit must resolve to the same branch and module in order for the branch and module to be transmitted to the KGB server.
Hint: use () to match empty branch or module if the concept is not applicable. Like:
branch-and-module-re: - "^/(trunk)/([^/]+)/" - "^()/(website)/" # either a sub-project in /trunk/<subproject> # or a file in the website, which is matched like a module
- branch-and-module-re-swap 1
If you can only provide the module name in the first capture and the branch name in the second, use this option to signal the fact to kgb-client. The setting is in effect for all patterns.
branch-and-module-re-swap: 1 branch-and-module-re: - "^/([^/]+)/(trunk|tags)/" - "^/(website)/()" # either a sub-project in /<subproject> # or a file in the website, which is matched like a module
- module name
In the case of sub-projects that use separate Git repositories, you may want to use explicit module name. Having this on the command line would allow for all the sub-project to share the configuration file (same repo-id) while still having sub-project-specific notifications.
Installation requires calling kgb-client with two command line arguments:
- path to the subversion repository
This is the physical path to the Subversion repository. Something like /srv/svn/my-repo
This is the revision number of the commit, that has triggered the hook.
Both these arguments are supplied to the standard Subversion post-commit hooks.
kgb-client shall be installed as a post-recieve hook. Something along the following shall do:
#!/bin/sh /path/to/kgb-client --git-reflog - --conf /path/to.conf ...
--git-reflog - will make kgb-client read the reflog information from standard input as any standard Git post-receive hook.
There are other ways to give kgb-client information about Git reflog, mostly useful when debugging on in unusual situations. See App::KGB::Client::Git.