magit.texi

\input texinfo.tex    @c -*-texinfo-*-
@c %**start of header
@setfilename magit.info
@settitle Magit User Manual
@documentencoding utf-8
@c %**end of header

@dircategory Emacs
@direntry
* Magit: (magit).        Using Git from Emacs with Magit.
@end direntry

@copying
Copyright @copyright{} 2008, 2009, 2011 Marius Vollmer et al. (See the
header of magit.el for the lengthy list of Magit contributors.)

@quotation
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License, Version 1.2 or
any later version published by the Free Software Foundation; with no
Invariant Sections, with no Front-Cover Texts, and with no Back-Cover
Texts.
@end quotation
@end copying

@node Top
@top Magit User Manual

Magit is an interface to the version control system Git, implemented
as an extension to Emacs.

@menu
* Introduction::
* Acknowledgments::
* Sections::
* Status::
* Untracked files::
* Staging and Committing::
* History::
* Reflogs::
* Diffing::
* Tagging::
* Resetting::
* Stashing::
* Branching::
* The branch list::
* Wazzup::
* Merging::
* Rebasing::
* Rewriting::
* Pushing and Pulling::
* Bisecting::
* Submodules::         
* Using Magit Extensions::
* Using Git Directly::
@end menu

@node Introduction
@chapter Introduction

With Magit, you can inspect and modify your Git repositories with
Emacs.  You can review and commit the changes you have made to the
tracked files, for example, and you can browse the history of past
changes.  There is support for cherry picking, reverting, merging,
rebasing, and other common Git operations.

Magit is not a complete interface to Git; it just aims to make the
most common Git operations convenient.  Thus, Magit will likely not
save you from learning Git itself.

This manual provides a tour of all Magit features.  It does not give a
introduction to version control in general, or to Git in particular.

The main entry point to Magit is @kbd{M-x magit-status}, which will
put you in Magit's status buffer.  You will be using it frequently, so
it is probably a good idea to bind @code{magit-status} to a key of
your choice.

In addition to the status buffer, Magit will also create buffers that
show lists of commits, buffers with diffs, and other kinds of buffers.
All these buffers are in @code{magit-mode} and have the same key
bindings.  Not all commands make sense in all contexts, but a given
key will always do the same thing in all Magit buffers.

Naturally, Magit runs the @code{git} command to do most of the work.
The @code{*magit-process*} buffer contains the transcript of the most
recent command.  You can switch to it with @kbd{$}.

@node Acknowledgments
@chapter Acknowledgments

From day one of the first Magit announcement, John Wiegley has
contributed numerous fixes, UI improvements, and new features.
Thanks!

Linh Dang and Christian Neukirchen also contributed from day one.
Thanks!

Phil Hagelberg joined a few days later.  Thanks!

Alex Ott contributed support for git svn.  Thanks!

Marcin Bachry contributed bug fixes and support for decorated logs.
Thanks!

Alexey Voinov fixed bugs.  Thanks!

Rémi Vanicat helped with Tramp support.  Thanks!

@node Sections
@chapter Sections

All Magit buffers are structured into nested 'sections'.  These
sections can be hidden and shown individually.  When a section is
hidden, only its first line is shown and all its children are
completely invisible.

The most fine-grained way to control the visibility of sections is the
@kbd{TAB} key.  It will to toggle the current section (the section
that contains point) between being hidden and being shown.

Typing @kbd{S-TAB} toggles the visibility of the children of the
current section.  When all of them are shown, they will all be hidden.
Otherwise, when some or all are hidden, they will all be shown.

The digit keys @kbd{1}, @kbd{2}, @kbd{3}, and @kbd{4} control the
visibility of sections based on levels.  Hitting @kbd{2}, for example,
will show sections on levels one and two, and will hide sections on
level 3.  However, only sections that are a parent or child of the
current section are affected.

For example, when the current section is on level 3 and you hit
@kbd{1}, the grand-parent of the current section (which is on level
one) will be shown, and the parent of the current section (level 2)
will be hidden.  The visibility of no other section will be changed.

This sounds a bit complicated, but you'll figure it out.

Using @kbd{M-1}, @kbd{M-2}, @kbd{M-3}, and @kbd{M-4} is similar to the
unmodified digits, but now all sections on the respective level are
affected, regardless of whether or not they are related to the current
section.

For example, @kbd{M-1} will only show the first lines of the top-level
sections and will hide everything else.  Typing @kbd{M-4} on the other
hand will show everything.

Because of the way the status buffer is set up, some changes to
section visibility are more common than others.  Files are on level 2
and diff hunks are on level 4.  Thus, you can type @kbd{2} to collapse
the diff of the current file, and @kbd{M-2} to collapse all files.
This returns the status buffer to its default setup and is a quick way
to unclutter it after drilling down into the modified files.

Because @kbd{2} and @kbd{M-2} are so common in the status buffer, they
are bound to additional, more mnemonic keys: @kbd{M-h} (hide) and
@kbd{M-H} (hide all).  Likewise @kbd{4} and @kbd{M-4} are also
available as @kbd{M-s} (show) and @kbd{M-S} (show all).

In other buffers than the status buffer, @kbd{M-h}, @kbd{M-H},
@kbd{M-s}, and @kbd{M-S} might work on different levels than on 2 and
4, but they keep their general meaning: @kbd{M-H} hides all detail,
and @kbd{M-S} shows everything.

@node Status
@chapter Status

Running @kbd{M-x magit-status} displays the main interface of Magit,
the status buffer.  You can have multiple status buffers active at the
same time, each associated with its own Git repository.

When invoking @kbd{M-x magit-status} from within a Git repository, it
will switch to the status buffer of that repository.  Otherwise, it
will prompt for a directory.  With a prefix argument, it will always
prompt.

You can set @code{magit-repo-dirs} to customize how
@code{magit-status} asks for the repository to work on.  When
@code{magit-repo-dirs} is nil, @code{magit-status} will simply ask for
a directory.

If you specify a directory that is not a Git repository, @kbd{M-x
magit-status} will offer to initialize it as one.

When @code{magit-repo-dirs} is not nil, it is treated as a list of
directory names, and @code{magit-status} will find all Git
repositories in those directories and offer them for completion.
(Magit will only look @code{magit-repo-dirs-depth} levels deep,
however.)

With two prefix arguments, @code{magit-status} will always prompt for
a raw directory.

Thus, you would normally set @code{magit-repo-dirs} to the places
where you keep most of your Git repositories and switch between them
with @kbd{C-u M-x magit-status}.  If you want to go to a repository
outside of your normal working areas, or if you want to create a new
repository, you would use @kbd{C-u C-u M-x magit-status}.

You need to explicitly refresh the status buffer when you have made
changes to the repository from outside of Emacs.  You can type @kbd{g}
in the status buffer itself, or just use @kbd{M-x magit-status}
instead of @kbd{C-x b} when switching to it.  You also need to refresh
the status buffer in this way after saving a file in Emacs.

The header at the top of the status buffer shows a short summary of
the repository state: where it is located, which branch is checked
out, etc.  Below the header are a number of sections that show details
about the working tree and the staging area.  You can hide and show
them as described in the previous section.

The first section shows @emph{Untracked files}, if there are any.  See
@ref{Untracked files} for more details.

Two section show your local changes.  They are explained fully in the
next chapter, @ref{Staging and Committing}.

If the current branch is associated with a remote tracking branch, the
status buffer shows the differences between the current branch and the
tracking branch.  See @ref{Pushing and Pulling} for more information.

During a history rewriting session, the status buffer shows the
@emph{Pending changes} and @emph{Pending commits} sections.  See
@ref{Rewriting} for more details.

@node Untracked files
@chapter Untracked files

Untracked files are shown in the @emph{Untracked files} section.

You can add a untracked file to the staging area with @kbd{s}.  If
point is on the @emph{Untracked files} section title when you it
@kbd{s}, all untracked files are staged.

Typing @kbd{C-u S} anywhere will also stage all untracked files,
together with all changes to the tracked files.

You can instruct Git to ignore them by typing @kbd{i}.  This will add
the filename to the @code{.gitignore} file.  Typing @kbd{C-u i} will
ask you for the name of the file to ignore.  This is useful to ignore
whole directories, for example.  The @kbd{I} command is similar to
@kbd{i} but will add the file to @code{.git/info/exclude} instead.

To delete a untracked file forever, use @kbd{k}.  If point is on the
@emph{Untracked files} section title when you it @kbd{k}, all
untracked files are deleted.

@node Staging and Committing
@chapter Staging and Committing

Comitting with Git is a two step process: first you add the changes
you want to commit to a 'staging area', and then you commit them to
the repository.  This allows you to only commit a subset of your local
changes.

Magit allows you to ignore the staging area if you wish.  As long as
your staging area is unused, Magit will show your uncomitted changes
in a section named @emph{Changes}.

When the staging area is in use, Magit uses two sections:
@emph{Unstaged changes} and @emph{Staged changes}.  The @emph{Staged
changes} section shows the changes that will be included in the next
commit, while the @emph{Unstaged changes} section shows the changes
that will be left out.

To move a unstaged hunk into the staging area, move point into the
hunk and type @kbd{s}.  Likewise, to unstage a hunk, move point into
it and type @kbd{u}.  If point is in a diff header when you type
@kbd{s} or @kbd{u}, all hunks belonging to that diff are moved at the
same time.

If the region is active when you type @kbd{s} or @kbd{u}, only the
changes in the region are staged or unstaged.  (This works line by
line: if the beginning of a line is in the region it is included in
the changes, otherwise it is not.)

Typing @kbd{C-u s} will ask you for a name of a file to be staged, for
example to stage files that are hidden.

To move all hunks of all diffs into the staging area in one go, type
@kbd{S}.  To unstage everything, type @kbd{U}.

Typing @kbd{C-u S} will stage all untracked files in addition to the
changes to tracked files.

You can discard uncommitted changes by moving point into a hunk and
typing @kbd{k}.  The changes to discard are selected as with @kbd{s}
and @kbd{u}.

Before committing, you should write a short description of the
changes.

Type @kbd{c} to pop up a buffer where you can write your change
description.  Once you are happy with the description, type @kbd{C-c
C-c} in that buffer to perform the commit.

If you want to write changes in a @file{ChangeLog} file, you can use
@kbd{C-x 4 a} on a diff hunk.

Typing @kbd{c} when the staging area is unused is a special situation.
Normally, the next commit would be empty, but you can configure Magit
to do something more useful by customizing the
@code{magit-commit-all-when-nothing-staged} variable.  One choice is
to instruct the subsequent @kbd{C-c C-c} to commit all changes.
Another choice is stage everything at the time of hitting @kbd{c}.

You can type @kbd{C-c C-a} in the buffer with the change description
to toggle a flag that determines whether the next commit will
@emph{amend} the current commit in HEAD.

Typing @kbd{C-c C-s} will toggle the @code{--signoff} option.  The
default is determined by the @code{magit-commit-signoff} customization
variable.

Typing @kbd{C-c C-e} will toggle the @code{--allow-empty} option.  This
allows you to make commits that serve as notes, without including any
changes.

If you change your mind and don't want to go ahead with your commit
while you are in the @code{*magit-log-edit*} buffer, you can just
switch to another buffer, continue editing there, staging and
unstaging things until you are happy, and then return to the
@code{*magit-log-edit*} buffer, maybe via @kbd{C-x b}, or by hitting
@kbd{c} again in a Magit buffer.

If you want to erase the @code{*magit-log-edit*} buffer and bury it,
you can hit @kbd{C-c C-k} in it.

Typing @kbd{C} will also pop up the change description buffer, but in
addition, it will try to insert a ChangeLog-style entry for the change
that point is in.

@node History
@chapter History

To show the repository history of your current head, type @kbd{l l}.  A
new buffer will be shown that displays the history in a terse form.
The first paragraph of each commit message is displayed, next to a
representation of the relationships between commits.

Giving a prefix argument to @kbd{l} will ask for the starting and end
point of the history.  This can be used to show the commits that are
in one branch, but not in another, for example.  The start point can
also be a range of revisions ``r1..r2''.  In that case ``r1'' is used
as the start and ``r2'' as the end point of the history.

Typing @kbd{l L} (or @kbd{l C-u L}) will show the log in a more verbose
form.

Magit will show only @code{magit-log-cutoff-length} entries. @kbd{e}
will show twice as many entries. @kbd{C-u e} will show all entries,
and given a numeric prefix argument, @kbd{e} will add this number of
entries. You can also show more entries by typing @kbd{RET} on the
text appended at the end of the buffer.

You can move point to a commit and then cause various things to happen
with it.  (The following commands work in any list of commit, such as
the one shown in the @emph{Unpushed commits} section.)

Typing @kbd{RET} will pop up more information about the current commit
and move point into the new buffer.  Typing @kbd{SPC} and @kbd{DEL}
will also show the information, but will scroll the new buffer up or
down (respectively) when typed again.

Typing @kbd{a} will apply the current commit to your current branch.
This is useful when you are browsing the history of some other branch
and you want to `cherry-pick' some changes from it.  A typical
situation is applying selected bug fixes from the development version
of a program to a release branch.  The cherry-picked changes will not
be committed automatically; you need to do that explicitly.

Typing @kbd{A} will cherry-pick the current commit and will also
commit the changes automatically when there have not been any
conflicts.

Typing @kbd{v} will revert the current commit.  Thus, it will apply
the changes made by that commit in reverse.  This is obviously useful
to cleanly undo changes that turned out to be wrong.  As with @kbd{a},
you need to commit the changes explicitly.

Typing @kbd{C-w} will copy the sha1 of the current commit into the
kill ring.

Typing @kbd{=} will show the differences from the current commit to
the @dfn{marked} commit.

You can mark the current commit by typing @kbd{.}.  When the current
commit is already marked, typing @kbd{.} will unmark it.  To unmark
the marked commit no matter where point is, use @kbd{C-u .}.

Some commands, such as @kbd{=}, will use the current commit and the
marked commit as implicit arguments.  Other commands will offer the
marked commit as a default when prompting for their arguments.

To grep the history press @kbd{s}.

@node Reflogs
@chapter Reflogs

You can use @kbd{l h} and @kbd{l H} to browse your @emph{reflog}, the
local history of changes made to your repository heads.  Typing
@kbd{H} will ask for a head, while @kbd{l h} will show the reflog of
@code{HEAD}.

The resulting buffer is just like the buffer produced by @kbd{l l} and
@kbd{l L} that shows the commit history.

@node Diffing
@chapter Diffing

Magit typically shows diffs in the ``unified'' format.

In any buffer that shows a diff, you can type @kbd{e} anywhere within
the diff to show the two versions of the file in Ediff.  If the diff
is of a file in the status buffer that needs to be merged, you will be
able to use Ediff as an interactive merge tool.  Otherwise, Ediff will
simply show the two versions of the file.

To show the changes from you working tree to another revision, type
@kbd{d}.  To show the changes between two arbitrary revisions, type
@kbd{D}.

You can use @kbd{a} within the diff output to apply the changes to
your working tree.  As usual when point is in a diff header for a
file, all changes for that file are applied, and when it is in a hunk,
only that hunk is.  When the region is active, the applied changes are
restricted to that region.

Typing @kbd{v} will apply the selected changes in reverse.

@node Tagging
@chapter Tagging

Typing @kbd{t t} will make a lighweight tag.  Typing @kbd{t T} will
make a annotated tag.  It will put you in the normal
@code{*magit-log-edit} buffer for writing commit messages, but typing
@kbd{C-c C-c} in it will make the tag instead.  This is controlled by
the @code{Tag} field that will be added to the @code{*magit-log-edit*}
buffer.  You can edit it, if you like.

@node Resetting
@chapter Resetting

Once you have added a commit to your local repository, you can not
change that commit anymore in any way.  But you can reset your current
head to an earlier commit and start over.

If you have published your history already, rewriting it in this way
can be confusing and should be avoided.  However, rewriting your local
history is fine and it is often cleaner to fix mistakes this way than
by reverting commits (with @kbd{v}, for example).

Typing @kbd{x} will ask for a revision and reset your current head to
it.  No changes will be made to your working tree and staging area.
Thus, the @emph{Staged changes} section in the status buffer will show
the changes that you have removed from your commit history.  You can
commit the changes again as if you had just made them, thus rewriting
history.

Typing @kbd{x} while point is in a line that describes a commit will
offer this commit as the default revision to reset to.  Thus, you can
move point to one of the commits in the @emph{Unpushed commits}
section and hit @kbd{x RET} to reset your current head to it.

Type @kbd{X} to reset your working tree and staging area to the most
recently committed state.  This will discard your local modifications,
so be careful.

You can give a prefix to @kbd{x} if you want to reset both the current
head and your working tree to a given commit.  This is the same as
first using an unprefixed @kbd{x} to reset only the head, and then
using @kbd{X}.

@node Stashing
@chapter Stashing

You can create a new stash with @kbd{z z}.  Your stashes will be listed
in the status buffer, and you can apply them with @kbd{a} and pop them
with @kbd{A}.  To drop a stash, use @kbd{k}.

With a prefix argument, both @kbd{a} and @kbd{A} will attempt to
reinstate the index as well as the working tree from the stash.

Typing @kbd{z Z} will create a stash just like @kbd{z z}, but will
leave the changes in your working tree and index.

You can visit and show stashes in the usual way: Typing @kbd{SPC} and
@kbd{DEL} will pop up a buffer with the description of the stash and
scroll it, typing @kbd{RET} will move point into that buffer. Using
@kbd{C-u RET} will move point into that buffer in other window.

@node Branching
@chapter Branching

The current branch is indicated in the header of the status buffer.
You can switch to a different branch by typing @kbd{b c}.  This will
immediately checkout the branch into your working copy, so you
shouldn't have any local modifications when switching branches.

If you try to switch to a remote branch, Magit will offer to create a
local tracking branch for it instead.  This way, you can easily start
working on new branches that have appeared in a remote repository.

Typing @kbd{b c} while point is at a commit description will offer
that commit as the default to switch to.  This will result in a
detached head.

To create a new branch and switch to it immediately, type @kbd{b B}.

Typing @kbd{b V} will list the local and remote branches in a new buffer
called @code{*magit-branches*} from which you can work with them. See
@ref{The branch list} for more details.


@node The branch list
@chapter The branch list

The branch list is a separate buffer called @code{*magit-branches*}
with its own local key map. The keybindings are modeled after magit's
main buffer. The buffer contains both local and remote branches. The
current local branch is marked by a ``*'' in front of the name.

Typing @kbd{k} will delete the branch in the current line, and
@kbd{C-u k} deletes it even if it hasn't been merged into the current
local branch. Deleting works for both local and remote branches.

You can merge the branch in the current line by typing @kbd{m m} for a
manual merge and @kbd{m m} for an automatic merge.

With @kbd{RET} or @kbd{b b} you can check out the branch in the current
line.

Typing @kbd{$} shows the @code{*magit-process*} buffer which contains
the transcript of the most recent command.

Typing @kbd{V} will refresh the branch list. @kbd{q} buries the branch
list buffer and deletes its window.


@node Wazzup
@chapter Wazzup

Typing @kbd{w} will show a summary of how your other branches relate
to the current branch.

For each branch, you will get a section that lists the commits in that
branch that are not in the current branch.  The sections are initially
collapsed; you need to explicitly open them with @kbd{TAB} (or
similar) to show the lists of commits.

When point is on a @emph{N unmerged commits in ...} title, the
corresponding branch will be offered as the default for a merge.

Hitting @kbd{i} on a branch title will ignore this branch in the
wazzup view.  You can use @kbd{C-u w} to show all branches, including
the ignored ones.  Hitting @kbd{i} on an already ignored branch in
that view will unignore it.

@node Merging
@chapter Merging

Magit offers two ways to merge branches: manually and automatic.  A
manual merge will apply all changes to your working tree and staging
area, but will not commit them, while a automatic merge will go ahead
and commit them immediately.

Type @kbd{m m} to initiate merge.

After initiating a merge, the header of the status buffer might remind
you that the next commit will be a merge commit (with more than one
parent).  If you want to abort a manual merge, just do a hard reset to
HEAD with @kbd{X}.

Merges can fail if the two branches you merge want to introduce
conflicting changes.  In that case, the automatic merge stops before the
commit, essentially falling back to a manual merge.  You need to resolve
the conflicts for example with @kbd{e} and stage the resolved files, for
example with @kbd{S}.

You can not stage individual hunks one by one as you resolve them, you
can only stage whole files once all conflicts in them have been
resolved.

@node Rebasing
@chapter Rebasing

Typing @kbd{R} in the status buffer will initiate a rebase or, if one
is already in progress, ask you how to continue.

When a rebase is stopped in the middle because of a conflict, the
header of the status buffer will indicate how far along you are in the
series of commits that are being replayed.  When that happens, you
should resolve the conflicts and stage everything and hit @kbd{R c} to
continue the rebase.  Alternatively, hitting @kbd{c} or @kbd{C} while
in the middle of a rebase will also ask you whether to continue the
rebase.

Of course, you can initiate a rebase in any number of ways, by
configuring @code{git pull} to rebase instead of merge, for example.
Such a rebase can be finished with @kbd{R} as well.

@node Rewriting
@chapter Rewriting

As hinted at earlier, you can rewrite your commit history.  For
example, you can reset the current head to an earlier commit with
@kbd{x}.  This leaves the working tree unchanged, and the status
buffer will show all the changes that have been made since that new
value of the current head.  You can commit these changes again,
possibly splitting them into multiple commits as you go along.

Amending your last commit is a common special case of rewriting
history like this.

Another common way to rewrite history is to reset the head to an
earlier commit, and then to cherry pick the previous commits in a
different order.  You could pick them from the reflog, for example.

Magit has several commands that can simplify the book keeping
associated with rewriting.  These commands all start with the @kbd{r}
prefix key.

Typing @kbd{r b} will start a rewrite operation.  You will be prompted
for a @emph{base} commit, and all commits between the current head and
this commit are put in a list of @emph{Pending commits} (including the
base commit).  The current head will then be reset to the parent of
the base commit.

You would then typically use @kbd{a} and @kbd{A} to cherry pick
commits from the list of pending commits in the desired order, until
all have been applied.  Magit shows which commits have been applied by
changing their marker from @code{*} to @code{.}.

Using @kbd{A} will immediately commit the commit (as usual).  If you
want to combine multiple previous commits into a single new one, use
@kbd{a} to apply them all to your working tree, and then commit them
together.

Magit has no explicit support for rewriting merge commits.  It will
happily include merge commits in the list of pending commits, but
there is no way of replaying them automatically.  You have to redo the
merge explicitly.

You can also use @kbd{v} to revert a commit when you have changed your
mind.  This will change the @code{.} mark back to @code{*}.

Once you are done with the rewrite, type @kbd{r t} to remove the book
keeping information from the status buffer.

If you rather wish to start over, type @kbd{r a}.  This will abort the
rewriting, resetting the current head back to the value it had before
the rewrite was started with @kbd{r s}.

Typing @kbd{r f} will @emph{finish} the rewrite: it will apply all
unused commits one after the other, as if you would us @kbd{A} with
all of them.

You can change the @kbd{*} and @kbd{.} marks of a pending commit
explicitly with @kbd{r *} and @kbd{r .}.

In addition to a list of pending commits, the status buffer will show
the @emph{Pending changes}.  This section shows the diff between the
original head and the current head.  You can use it to review the
changes that you still need to rewrite, and you can apply hunks from
it, like from any other diff.

@node Pushing and Pulling
@chapter Pushing and Pulling

Magit will run @code{git push} when you type @kbd{P p}.  If you give a
prefix argument to @kbd{P p}, you will be prompted for the repository to
push to.  When no default remote repository has been configured yet for
the current branch, you will be prompted as well.  Typing @kbd{P p} will
only push the current branch to the remote.  In other words, it will run
@code{git push <remote> <branch>}.  The branch will be created in the
remote if it doesn't exist already.  The local branch will be configured
so that it pulls from the new remote branch.

Typing @kbd{f f} will run @code{git fetch}.  It will prompt for the name
of the remote to update if there are no default one.  Typing @kbd{f o}
will always prompt for the remote.  Typing @kbd{F F} will run @code{git
pull}.  When you don't have a default branch configured to be pulled
into the current one, you will be asked for it.

If there is a default remote repository for the current branch, Magit
will show that repository in the status buffer header.

In this case, the status buffer will also have a @emph{Unpushed
commits} section that shows the commits on your current head that are
not in the branch named @code{<remote>/<branch>}.  This section works
just like the history buffer: you can see details about a commit with
@kbd{RET}, compare two of them with @kbd{.} and @kbd{=}, and you can
reset your current head to one of them with @kbd{x}, for example. If
you want to push the changes then type @kbd{P p}.

When the remote branch has changes that are not in the current branch,
Magit shows them in a section called @emph{Unpulled changes}.  Typing
@kbd{F p} will merge them into the current branch.

@node Submodules
@chapter Submodules

@table @kbd
@item M u
Update the submodules, with a prefix argument it will initializing.

@item M i
Initialize the submodules.

@item M b
Update and initialize the submodules in one go.

@item M s
Synchronizes submodules' remote URL configuration setting to the value
specified in .gitmodules.
@end table


@node Bisecting
@chapter Bisecting

magit supports bisecting by showing how many revisions and steps are
left to be tested in the status buffer. You can control the bisect
session from both the status and from log buffers with the @kbd{B} key
menu.

Typing @kbd{B s} will start a bisect session.  You will be prompted
for a revision that is known to be bad (defaults to @emph{HEAD}) and
for a revision that is known to be good (defaults to the revision at
point if there is one). git will select a revision for you to test,
and magit will update its status buffer accordingly.

You can tell git that the current revision is good with @kbd{B g},
that it is bad with @kbd{B b} or that git should skip it with @kbd{B
k}. You can also tell git to go into full automatic mode by giving it
the name of a script to run for each revision to test with @kbd{B u}.

The current status can be shown as a log with @kbd{B l}. It contains
the revisions that have already been tested and your decisions about
their state.

The revisions left to test can be visualized in gitk with @kbd{B v}.

When you're finished bisecting you have to reset the session with
@kbd{B r}.


@node Using Magit Extensions
@chapter Magit Extensions

@menu
* Interfacing with Subversion::
* Interfacing with Topgit::
* Developing Extensions::
@end menu


@node Interfacing with Subversion
@section Interfacing with Subversion

Typing @kbd{N r} runs @code{git svn rebase}, typing @kbd{N c} runs
@code{git svn dcommit} and typing @kbd{N f} runs @code{git svn fetch}.

@kbd{N s} will prompt you for a (numeric, Subversion) revision and
then search for a corresponding Git sha1 for the commit. This is
limited to the path of the remote Subversion repository. With a prefix
(@kbd{C-u N s} the user will also be prompted for a branch to search
in.

@node Interfacing with Topgit
@section Interfacing with Topgit

Topgit (http://repo.or.cz/r/topgit.git) is a patch queue manager that
aims at being close as possible to raw Git, which makes it easy to use
with Magit. In particular, it does not require to use a different set of
commands for ``commit'', ``update'',… operations.

@file{magit-topgit.el} provides basic integration with Magit, mostly by
providing a ``Topics'' section.

Topgit branches can be created the regular way, by using a ``t/'' prefix
by convention. So, creating a ``t/foo'' branch will actually populate
the ``Topics'' section with one more branch after committing
@file{.topdeps} and @file{.topmsg}.

Also, the way we pull (see @ref{Pushing and Pulling}) such a branch is
slightly different, since it requires updating the various dependencies
of that branch. This should be mostly transparent, except in case
of conflicts.

@node Developing Extensions
@section Developing Extensions

Magit provides a generic mechanism to allow cooperation with Git-related
systems, such as foreign VCS, patch systems,…

In particular it allows to:

@itemize @bullet
@item
Define sections to display specific informations about the current state
of the repository, and place them relatively to existing sections.

@code{magit-define-inserter} automagically defines two hooks called
@code{magit-before-insert-SECTION-hook} and
@code{magit-after-insert-SECTION-hook} that allow to generate and place
more sections.

In the following example, we use the builtin ``stashes'' section to
place our own ``foo'' one.

@example
(magit-define-inserter foo ()
  (magit-git-section 'foo
                     "Foo:" 'foo-wash-function
                     "foo" "arg1" "arg2"))
(add-hook 'magit-after-insert-stashes-hook 'magit-insert-foo)
@end example

@item
Define new types of objects in those sections.

The function @code{foo-wash-function} defined above post-processes each
line of the output of the ``git foo arg1 arg2'' command, and is able to
associate a type to certain lines.

A simple implementation could be:

@example
(defun foo-wash-function ()
  (let ((foo (buffer-substring (line-beginning-position) (line-end-position))))
    (goto-char (line-beginning-position))
    (magit-with-section foo 'foo
      (magit-set-section-info foo)
      (forward-line))))
@end example

In this case, every line of the command output is transformed into an
object of type @code{'foo}.

@item
Alter behavior of generic commands to dispatch them correctly to the
relevant system, optionally making use of the newly defined types.

@example
(magit-add-action (item info "discard")
  ((foo)
   (do-something-meaningful-for-discarding-a-foo)))
@end example

This will alter the behavior of @kbd{k}, when applied to those objects.

@item
Plug a different logic into basic commands, to reflect the presence of
the extension.

@code{magit-define-command} automagically defines
a @code{magit-CMD-command-hook} that can contain a list of functions to
call before the actual core code. Execution stops after the first hook
that returns a non-nil value. This leaves room for extension logic.

@example
(add-hook 'magit-create-branch-command-hook 'foo-create-branch)
@end example

The function @code{foo-create-branch} will be called each time an
attempt is made to create a branch, and can, for example, react to
a certain name convention.

@item
Define new commands and associated menu.

This part is not really specific to extensions, except that menus take
place in the ``Extensions'' submenu.

@end itemize

@node Using Git Directly
@chapter Using Git Directly

For situations when Magit doesn't do everything you need, you can run
raw Git commands using @kbd{:}.  This will prompt for a Git command, run
it, and refresh the status buffer.  The output can be viewed by typing
@kbd{$}.

@bye