git-multimail
                           =============

git-multimail is a tool for sending notification emails on pushes to a
Git repository.  It includes a Python module called git_multimail.py,
which can either be used as a hook script directly or can be imported
as a Python module into another script.

git-multimail is derived from the Git project's old
contrib/hooks/post-receive-email, and is mostly compatible with that
script.  See README.migrate-from-post-receive-email for details about
the differences and for how to migrate from post-receive-email to
git-multimail.

git-multimail, like the rest of the Git project, is licensed under
GPLv2 (see the COPYING file for details).

Please note: although, as a convenience, git-multimail may be
distributed along with the main Git project, development of
git-multimail takes place in its own, separate project.  See section
"Getting involved" below for more information.


By default, for each push received by the repository, git-multimail:

1. Outputs one email summarizing each reference that was changed.
   These "reference change" (called "refchange" below) emails describe
   the nature of the change (e.g., was the reference created, deleted,
   fast-forwarded, etc.) and include a one-line summary of each commit
   that was added to the reference.

2. Outputs one email for each new commit that was introduced by the
   reference change.  These "commit" emails include a list of the
   files changed by the commit, followed by the diffs of files
   modified by the commit.  The commit emails are threaded to the
   corresponding reference change email via "In-Reply-To".  This style
   (similar to the "git format-patch" style used on the Git mailing
   list) makes it easy to scan through the emails, jump to patches
   that need further attention, and write comments about specific
   commits.  Commits are handled in reverse topological order (i.e.,
   parents shown before children).  For example,

   [git] branch master updated
   + [git] 01/08: doc: fix xref link from api docs to manual pages
   + [git] 02/08: api-credentials.txt: show the big picture first
   + [git] 03/08: api-credentials.txt: mention credential.helper explicitly
   + [git] 04/08: api-credentials.txt: add "see also" section
   + [git] 05/08: t3510 (cherry-pick-sequence): add missing '&&'
   + [git] 06/08: Merge branch 'rr/maint-t3510-cascade-fix'
   + [git] 07/08: Merge branch 'mm/api-credentials-doc'
   + [git] 08/08: Git 1.7.11-rc2

   Each commit appears in exactly one commit email, the first time
   that it is pushed to the repository.  If a commit is later merged
   into another branch, then a one-line summary of the commit is
   included in the reference change email (as usual), but no
   additional commit email is generated.

   By default, reference change emails have their "Reply-To" field set
   to the person who pushed the change, and commit emails have their
   "Reply-To" field set to the author of the commit.

3. Output one "announce" mail for each new annotated tag, including
   information about the tag and optionally a shortlog describing the
   changes since the previous tag.  Such emails might be useful if you
   use annotated tags to mark releases of your project.


Requirements
------------

* Python 2.x, version 2.4 or later.  No non-standard Python modules
  are required.  git-multimail does *not* currently work with Python
  3.x.

  The example scripts invoke Python using the following shebang line
  (following PEP 394 [1]):

      #! /usr/bin/env python2

  If your system's Python2 interpreter is not in your PATH or is not
  called "python2", you can change the lines accordingly.  Or you can
  invoke the Python interpreter explicitly, for example via a tiny
  shell script like

      #! /bin/sh
      /usr/local/bin/python /path/to/git_multimail.py "$@"

* The "git" command must be in your PATH.  git-multimail is known to
  work with Git versions back to 1.7.1.  (Earlier versions have not
  been tested; if you do so, please report your results.)

* To send emails using the default configuration, a standard sendmail
  program must be located at '/usr/sbin/sendmail' and configured
  correctly to send emails.  If this is not the case, see the
  multimailhook.mailer configuration variable below for how to
  configure git-multimail to send emails via an SMTP server.


Invocation
----------

git_multimail.py is designed to be used as a "post-receive" hook in a
Git repository (see githooks(5)).  Link or copy it to
$GIT_DIR/hooks/post-receive within the repository for which email
notifications are desired.  Usually it should be installed on the
central repository for a project, to which all commits are eventually
pushed.

For use on pre-v1.5.1 Git servers, git_multimail.py can also work as
an "update" hook, taking its arguments on the command line.  To use
this script in this manner, link or copy it to $GIT_DIR/hooks/update.
Please note that the script is not completely reliable in this mode
[2].

Alternatively, git_multimail.py can be imported as a Python module
into your own Python post-receive script.  This method is a bit more
work, but allows the behavior of the hook to be customized using
arbitrary Python code.  For example, you can use a custom environment
(perhaps inheriting from GenericEnvironment or GitoliteEnvironment) to

* change how the user who did the push is determined

* read users' email addresses from an LDAP server or from a database

* decide which users should be notified about which commits based on
  the contents of the commits (e.g., for users who want to be notified
  only about changes affecting particular files or subdirectories)

Or you can change how emails are sent by writing your own Mailer
class.  The "post-receive" script in this directory demonstrates how
to use git_multimail.py as a Python module.  (If you make interesting
changes of this type, please consider sharing them with the
community.)


Configuration
-------------

By default, git-multimail mostly takes its configuration from the
following "git config" settings:

multimailhook.environment

    This describes the general environment of the repository.
    Currently supported values:

    "generic" -- the username of the pusher is read from $USER and the
        repository name is derived from the repository's path.

    "gitolite" -- the username of the pusher is read from $GL_USER and
        the repository name from $GL_REPO.

    If neither of these environments is suitable for your setup, then
    you can implement a Python class that inherits from Environment
    and instantiate it via a script that looks like the example
    post-receive script.

    The environment value can be specified on the command line using
    the --environment option.  If it is not specified on the command
    line or by multimailhook.environment, then it defaults to
    "gitolite" if the environment contains variables $GL_USER and
    $GL_REPO; otherwise "generic".

multimailhook.repoName

    A short name of this Git repository, to be used in various places
    in the notification email text.  The default is to use $GL_REPO
    for gitolite repositories, or otherwise to derive this value from
    the repository path name.

multimailhook.mailinglist

    The list of email addresses to which notification emails should be
    sent, as RFC 2822 email addresses separated by commas.  This
    configuration option can be multivalued.  Leave it unset or set it
    to the empty string to not send emails by default.  The next few
    settings can be used to configure specific address lists for
    specific types of notification email.

multimailhook.refchangeList

    The list of email addresses to which summary emails about
    reference changes should be sent, as RFC 2822 email addresses
    separated by commas.  This configuration option can be
    multivalued.  The default is the value in
    multimailhook.mailinglist.  Set this value to the empty string to
    prevent reference change emails from being sent.

multimailhook.announceList

    The list of email addresses to which emails about new annotated
    tags should be sent, as RFC 2822 email addresses separated by
    commas.  This configuration option can be multivalued.  The
    default is the value in multimailhook.refchangelist or
    multimailhook.mailinglist.  Set this value to the empty string to
    prevent annotated tag announcement emails from being sent.

multimailhook.commitList

    The list of email addresses to which emails about individual new
    commits should be sent, as RFC 2822 email addresses separated by
    commas.  This configuration option can be multivalued.  The
    default is the value in multimailhook.mailinglist.  Set this value
    to the empty string to prevent notification emails about
    individual commits from being sent.

multimailhook.announceShortlog

    If this option is set to true, then emails about changes to
    annotated tags include a shortlog of changes since the previous
    tag.  This can be useful if the annotated tags represent releases;
    then the shortlog will be a kind of rough summary of what has
    happened since the last release.  But if your tagging policy is
    not so straightforward, then the shortlog might be confusing
    rather than useful.  Default is false.

multimailhook.refchangeShowLog

    If this option is set to true, then summary emails about reference
    changes will include a detailed log of the added commits in
    addition to the one line summary.  The log is generated by running
    "git log" with the options specified in multimailhook.logOpts.
    Default is false.

multimailhook.mailer

    This option changes the way emails are sent.  Accepted values are:

    - sendmail (the default): use the command /usr/sbin/sendmail or
      /usr/lib/sendmail (or sendmailCommand, if configured).  This
      mode can be further customized via the following options:

       multimailhook.sendmailCommand

           The command used by mailer "sendmail" to send emails.  Shell
           quoting is allowed in the value of this setting, but remember that
           Git requires double-quotes to be escaped; e.g.,

             git config multimailhook.sendmailcommand '/usr/sbin/sendmail -t -F \"Git Repo\"'

           Default is '/usr/sbin/sendmail -t' or '/usr/lib/sendmail
           -t' (depending on which file is present and executable).

       multimailhook.envelopeSender

           If set then pass this value to sendmail via the -f option to set
           the envelope sender address.

    - smtp: use Python's smtplib.  This is useful when the sendmail
      command is not available on the system.  This mode can be
      further customized via the following options:

       multimailhook.smtpServer

           The name of the SMTP server to connect to.  The value can
           also include a colon and a port number; e.g.,
           "mail.example.com:25".  Default is 'localhost' using port
           25.

       multimailhook.envelopeSender

           The sender address to be passed to the SMTP server.  If
           unset, then the value of multimailhook.from is used.

multimailhook.from

    If set then use this value in the From: field of generated emails.
    If unset, then use the repository's user configuration (user.name
    and user.email).  If user.email is also unset, then use
    multimailhook.envelopeSender.

multimailhook.administrator

    The name and/or email address of the administrator of the Git
    repository; used in FOOTER_TEMPLATE.  Default is
    multimailhook.envelopesender if it is set; otherwise a generic
    string is used.

multimailhook.emailPrefix

    All emails have this string prepended to their subjects, to aid
    email filtering (though filtering based on the X-Git-* email
    headers is probably more robust).  Default is the short name of
    the repository in square brackets; e.g., "[myrepo]".

multimailhook.emailMaxLines

    The maximum number of lines that should be included in the body of
    a generated email.  If not specified, there is no limit.  Lines
    beyond the limit are suppressed and counted, and a final line is
    added indicating the number of suppressed lines.

multimailhook.emailMaxLineLength

    The maximum length of a line in the email body.  Lines longer than
    this limit are truncated to this length with a trailing " [...]"
    added to indicate the missing text.  The default is 500, because
    (a) diffs with longer lines are probably from binary files, for
    which a diff is useless, and (b) even if a text file has such long
    lines, the diffs are probably unreadable anyway.  To disable line
    truncation, set this option to 0.

multimailhook.maxCommitEmails

    The maximum number of commit emails to send for a given change.
    When the number of patches is larger that this value, only the
    summary refchange email is sent.  This can avoid accidental
    mailbombing, for example on an initial push.  To disable commit
    emails limit, set this option to 0.  The default is 500.

multimailhook.emailStrictUTF8

    If this boolean option is set to "true", then the main part of the
    email body is forced to be valid UTF-8.  Any characters that are
    not valid UTF-8 are converted to the Unicode replacement
    character, U+FFFD.  The default is "true".

multimailhook.diffOpts

    Options passed to "git diff-tree" when generating the summary
    information for ReferenceChange emails.  Default is "--stat
    --summary --find-copies-harder".  Add -p to those options to
    include a unified diff of changes in addition to the usual summary
    output.  Shell quoting is allowed; see multimailhook.logOpts for
    details.

multimailhook.logOpts

    Options passed to "git log" to generate additional info for
    reference change emails (used only if refchangeShowLog is set).
    For example, adding --graph will show the graph of revisions, -p
    will show the complete diff, etc.  The default is empty.

    Shell quoting is allowed; for example, a log format that contains
    spaces can be specified using something like:

      git config multimailhook.logopts '--pretty=format:"%h %aN <%aE>%n%s%n%n%b%n"'

    If you want to set this by editing your configuration file
    directly, remember that Git requires double-quotes to be escaped
    (see git-config(1) for more information):

      [multimailhook]
              logopts = --pretty=format:\"%h %aN <%aE>%n%s%n%n%b%n\"

multimailhook.emailDomain

    Domain name appended to the username of the person doing the push
    to convert it into an email address (via "%s@%s" % (username,
    emaildomain)).  More complicated schemes can be implemented by
    overriding Environment and overriding its get_pusher_email()
    method.

multimailhook.replyTo
multimailhook.replyToCommit
multimailhook.replyToRefchange

    Addresses to use in the Reply-To: field for commit emails
    (replyToCommit) and refchange emails (replyToRefchange).
    multimailhook.replyTo is used as default when replyToCommit or
    replyToRefchange is not set.  The value for these variables can be
    either:

    - An email address, which will be used directly.

    - The value "pusher", in which case the pusher's address (if
      available) will be used.  This is the default for refchange
      emails.

    - The value "author" (meaningful only for replyToCommit), in which
      case the commit author's address will be used.  This is the
      default for commit emails.

    - The value "none", in which case the Reply-To: field will be
      omitted.


Email filtering aids
--------------------

All emails include extra headers to enable fine tuned filtering and
give information for debugging.  All emails include the headers
"X-Git-Repo", "X-Git-Refname", and "X-Git-Reftype".  ReferenceChange
emails also include headers "X-Git-Oldrev" and "X-Git-Newrev";
Revision emails also include header "X-Git-Rev".


Customizing email contents
--------------------------

git-multimail mostly generates emails by expanding templates.  The
templates can be customized.  To avoid the need to edit
git_multimail.py directly, the preferred way to change the templates
is to write a separate Python script that imports git_multimail.py as
a module, then replaces the templates in place.  See the provided
post-receive script for an example of how this is done.


Customizing git-multimail for your environment
----------------------------------------------

git-multimail is mostly customized via an "environment" that describes
the local environment in which Git is running.  Two types of
environment are built in:

* GenericEnvironment: a stand-alone Git repository.

* GitoliteEnvironment: a Git repository that is managed by gitolite
  [3].  For such repositories, the identity of the pusher is read from
  environment variable $GL_USER, and the name of the repository is
  read from $GL_REPO (if it is not overridden by
  multimailhook.reponame).

By default, git-multimail assumes GitoliteEnvironment if $GL_USER and
$GL_REPO are set, and otherwise assumes GenericEnvironment.
Alternatively, you can choose one of these two environments explicitly
by setting a "multimailhook.environment" config setting (which can
have the value "generic" or "gitolite") or by passing an --environment
option to the script.

If you need to customize the script in ways that are not supported by
the existing environments, you can define your own environment class
class using arbitrary Python code.  To do so, you need to import
git_multimail.py as a Python module, as demonstrated by the example
post-receive script.  Then implement your environment class; it should
usually inherit from one of the existing Environment classes and
possibly one or more of the EnvironmentMixin classes.  Then set the
"environment" variable to an instance of your own environment class
and pass it to run_as_post_receive_hook().

The standard environment classes, GenericEnvironment and
GitoliteEnvironment, are in fact themselves put together out of a
number of mixin classes, each of which handles one aspect of the
customization.  For the finest control over your configuration, you
can specify exactly which mixin classes your own environment class
should inherit from, and override individual methods (or even add your
own mixin classes) to implement entirely new behaviors.  If you
implement any mixins that might be useful to other people, please
consider sharing them with the community!


Getting involved
----------------

git-multimail is an open-source project, built by volunteers.  We
would welcome your help!

The current maintainer is Michael Haggerty <mhagger@alum.mit.edu>.

General discussion of git-multimail takes place on the main Git
mailing list,

    git@vger.kernel.org

Please CC emails regarding git-multimail to me so that I don't
overlook them.

The git-multimail project itself is currently hosted on GitHub:

    https://github.com/mhagger/git-multimail

We use the GitHub issue tracker to keep track of bugs and feature
requests, and GitHub pull requests to exchange patches (though, if you
prefer, you can send patches via the Git mailing list with cc to me).

Please note that although a copy of git-multimail will probably be
distributed in the "contrib" section of the main Git project,
development takes place in the separate git-multimail repository on
GitHub!  (Whenever enough changes to git-multimail have accumulated, a
new code-drop of git-multimail will be submitted for inclusion in the
Git project.)


Footnotes
---------

[1] http://www.python.org/dev/peps/pep-0394/

[2] Because of the way information is passed to update hooks, the
    script's method of determining whether a commit has already been
    seen does not work when it is used as an "update" script.  In
    particular, no notification email will be generated for a new
    commit that is added to multiple references in the same push.

[3] https://github.com/sitaramc/gitolite