Patches
*******

.. module:: patches
	:synopsis: Describe patch operations

Defining a patch
================

These links are recommended reading for understanding the history and nature
of patches and how they can be used with Mercurial.

* `The patch management problem <https://tortoisehg.bitbucket.io/hgbook/1.7/managing-change-with-mercurial-queues.html#sec:mq:patch-mgmt>`_
* `Understanding patches <https://tortoisehg.bitbucket.io/hgbook/1.7/managing-change-with-mercurial-queues.html#sec:mq:patch>`_
* `More about patches <https://tortoisehg.bitbucket.io/hgbook/1.7/managing-change-with-mercurial-queues.html#sec:mq:adv-patch>`_

Pitfalls
========

The standard patch format cannot describe binary files, renames, copies,
or permission changes.  If your patch needs to record any of those
things, you will need to enable **git** patches via::

	[diff]
	git=True

Mercurial 1.5 improves its behavior in this regard.  It will warn you
when git diffs are required, or sometimes upgrade to the git format
automatically.  See also the
`diff section <https://www.mercurial-scm.org/doc/hgrc.5.html#diff>`_ of
the hgrc documentation.

Mercurial's patch routines do not deal well with mixed EOLN between
source files and patches.  The **patch.eol** setting was introduced in
1.3 to improve this situation::

	[patch]
	eol = auto #strict, lf, or crlf

.. note::
	When eol is set to *auto*, the patching engine will preserve the line
	endings of the patched file regardless of the line endings in the
	patch itself.  You almost always want eol to be configured to *auto*.
	The only downside is that you cannot make a patch that changes the
	line endings of a source file.

See also the `patch section <https://www.mercurial-scm.org/doc/hgrc.5.html#patch>`_
of the hgrc documentation.

Applying a patch is not a foolproof operation.  If the source file has
diverged from the file that was used to create the patch, there may be
conflicts during the patch application.  These are written to a file
with an .rej extension.  TortoiseHg 2.0 includes a :command:`thg
rejects` command that can aid in the merging of the rejected chunks into
the source file.


Export Patches
==============

Changeset
---------

To export a changeset as a patch file, use the changeset context menu of
the Workbench to select :menuselection:`Export --> Export Patch`.  It
writes the changeset to a file in the repository root folder.

Changeset Ranges
----------------

Select the start and end of a range of changesets in the Workbench and
open the special revision range context menu.  From this menu you can
generate patches, generate a bundle, send emails, or visually diff the
accumulated changes.

This is a very powerful feature and there is no restriction on the base
and target changesets you can select.

Email
-----

.. figure:: figures/email.png
	:alt: Email dialog

	Email dialog of Workbench

To send a changeset as an email, use the changeset context menu of the
Workbench. :menuselection:`Export --> Email Patch`.  This opens the
e-mail dialog for this single changeset.

To send a changeset range, use the changeset range selection feature of
the Workbench and select :menuselection:`Email selected...` or
:menuselection:`Email DAG range...`.

Lastly, you can use the :guilabel:`Email` button on the sync tab of the
Workbench to email all outgoing changes to the selected remote
repository.

.. note::
	You must configure
	`SMTP <https://www.mercurial-scm.org/doc/hgrc.5.html#smtp>`_
	to send patches via email

Import Patches
==============

.. figure:: figures/import.png
	:alt: Import tool

	Import dialog of the WorkBench

The import dialog can be opened from the :guilabel:`Repository` menu of
the Workbench, or via :command:`thg import`.  The dialog supports file
and directory drag and drop.

You have the choice of importing directly into the repository, the
working folder, a shelf file, or your patch queue.

.. note::
	Importing a patch requires a clean working directory state.  You
	must commit, revert, or shelve changes before importing a patch.

.. note::
	If a patch does not import itself cleanly into the repository, the
	recommended recourse is to import the patch into your patch queue
	(qimport) and then qpush the patch.  This uses TortoiseHg patch
	rejection dialog and preserves the meta-data in the patch header.
	Do not forget to qrefresh after resolving the rejected chunks.

.. warning::
	If the patch you are importing does not have a commit
	message, Mercurial will try to launch your editor, just as if you
	had tried to import the patch from the command line.  Your ui.editor
	needs to be a GUI app to make this work correctly.


Patch Queues
============

.. figure:: figures/patchqueue.png
	:alt: Patch Queue in Workbench

	A patch queue in the repository graph

When MQ is enabled, several Workbench features are exposed. Context menu
options are exposed in the changeset menus, your patch queue is graphed
together with your repository's history, and a Patch Queue is activated.

Double clicking on an unapplied patch, an applied patch other than the
current qtip, or the qparent triggers a qgoto command; making the double
clicked revision the current patch.  Double clicking on any other
revision will trigger a visual diff of that revision.

.. note::
	The Workbench must be restarted after enabling or disabling the MQ
	extension in a repository.  This is true of most extensions.

.. note::
	It is recommended to learn the MQ extension before using the MQ
	features of the Workbench.

.. versionchanged:: 2.10
   The :guilabel:`Patch Queue` task tab has been superseded by the
   :guilabel:`Commit` task tab and the :guilabel:`Patch Queue` dock.

Patch Rejects
=============

As explained previously, patches are not guaranteed to apply cleanly to
their intended source files.  Prior to TortoiseHg 2.0, the only recourse
available when patch chunks were rejected was to open the source file
and the rejects file in an editor and manually fixup the rejected
chunks.

TortoiseHg 2.0 introduces a dialog that makes this a little bit easier.
If the shelve tool detects chunk rejections, it offers to open the
rejected chunks in the rejects editor.  The MQ tool also does this for
qpush commands.

.. figure:: figures/rejects.png
	:alt: Patch Rejects Editor

	Resolve rejected patch chunks

The rejects editor is very basic.  Your source file is opened in a
QScintilla2 window for edit.  Below the source file is the list of
chunks that failed to apply to this file.  When you click on a chunk in
the list the editor jumps to the line where the chunk context was
supposed to match.  It is up to you to figure out why the chunk did not
apply and to resolve it (perhaps even by ignoring the chunk).  The
resolved/unresolved states are for your own book keeping, so you know
when all of the chunks have been dealt with.  Once you have marked all
of the chunks resolved, the :guilabel:`Save` button will become
sensitive.

.. vim: noet ts=4