slackbuilds/magit.texi
Marius Vollmer f89a045282 Only ever commit staged changes when amending a commit.
The usual behavior of committing all unstaged changes when there are
no staged changes makes it impossible to amend a commit just to fix
the commit message, which is a quite common thing to do.

* magit.el (magit-log-edit-commit): Do not pass "--all" to git commit
when amending.
2009-02-10 23:31:16 +02:00

550 lines
22 KiB
Text

\input texinfo.tex @c -*-texinfo-*-
@c %**start of header
@setfilename magit.info
@settitle Magit User Manual
@c %**end of header
@dircategory Emacs
@direntry
* Magit: (magit). Using Git from Emacs with Magit.
@end direntry
@copying
Copyright @copyright{} 2008, 2009 Marius Vollmer
@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::
* Wazzup::
* Merging::
* Rebasing::
* Rewriting::
* Pushing and Pulling::
* Interfacing with Subversion::
@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, Linh Dang,
and Christian Neukirchen have contributed fixes, UI improvements, and
new features. 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!
@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 show, 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 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 in the status buffer.
@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 rpefix argument, it will always
prompt.
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.
Typing a digit in the status buffer will move point to a section
header: @kbd{1} will move to @emph{Untracked files}, @kbd{2} will move
to @emph{Unstaged changes} or @emph{Changes}, @kbd{3} will move to
@emph{Staged changes}, and @kbd{4} will move to @emph{Unpushed
commits}. Note that these relations are fixed, @kbd{3} always moves
to @emph{Unstaged changes}, not the third visible section.
@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}.
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}.
@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.)
To move all hunks of all diffs into the staging area in one go, type
@kbd{S}. To unstage everything, type @kbd{U}.
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. When your stagin area is
unused at this point, all your uncommitted changes are committed.
When the staging area is in use, only the staged changes are
committed.
You can type @kbd{C-c C-a} in the buffer with the change description
toggles a flag that determines whether the next commit will be a
@emph{amending} one. When amending a commit, only the staged changes
are committed, even if there are none. (Normally, all unstaged
changes are committed when there are no staged changes.) This allows
you to easily modify the message of the last commit.
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}. 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.
Typing @kbd{L} (or @kbd{C-u L}) will show the log in a more verbose
form.
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.
@node Reflogs
@chapter Reflogs
You can use @kbd{h} and @kbd{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{h} will show the reflog of
@code{HEAD}.
The resulting buffer is just like the buffer produced by @kbd{l} and
@kbd{L} that shows the commit history.
@node Diffing
@chapter Diffing
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} will make a lighweight tag. Typing @kbd{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.
@node Stashing
@chapter Stashing
You can create a new stash with @kbd{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}.
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.
@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}. This will
immediately checkout the branch into your working copy, so you
shouldn't have any local modifications when switching branches.
Similar to @kbd{x}, typing @kbd{b} 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}.
@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.
@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} to initiate a manual merge, and type @kbd{M} for a
automatic merge.
A manual merge is useful when carefully merging a new feature that you
want to review and test before even committing it. A automatic merge
is appropriate when you are on a feature branch and want to catch up
with the master, say.
After initiating a manual merge, the header of the status buffer will
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 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.
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 he 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 s} 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}. If you give a
prefix argument to @kbd{P}, you will be prompted for the repository to
push to. When no default remote repositor has been configured yet for
the current branch, you will be prompted as well. Typing @kbd{P} will
only push the current branch to the remote. In other words, it will
run @code{git push <remote> <branch>}.
Typing @kbd{f} will run @code{git remote update} and @kbd{F} will run
@code{git pull}. You should have setup your Git configuration to do
the right thing for @code{git pull}.
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.
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} will merge them into the current branch.
@node Interfacing with Subversion
@chapter Interfacing with Subversion
Typing @kbd{N r} runs @code{git svn rebase} and typing @kbd{N c} runs
@code{git svn dcommit}.
@bye