Allow rm to recursively remove directories

[fdroidserver.git] / docs / fdroid.texi

\input texinfo   @c -*-texinfo-*-
@c %**start of header
@setfilename fdroid.info
@settitle F-Droid Server Manual
@c %**end of header

@copying
This manual is for the F-Droid repository server tools.

Copyright @copyright{} 2010, 2011, 2012, 2013 Ciaran Gultnieks

Copyright @copyright{} 2011 Henrik Tunedal, Michael Haas, John Sullivan

Copyright @copyright{} 2013 David Black, Daniel Martí

@quotation
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License, Version 1.3
or any later version published by the Free Software Foundation;
with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts.
A copy of the license is included in the section entitled "GNU
Free Documentation License".
@end quotation
@end copying

@titlepage
@title F-Droid Server Manual
@author Ciaran Gultnieks and the F-Droid project
@page
@vskip 0pt plus 1filll
@insertcopying

@end titlepage

@contents

@ifnottex
@node Top
@top F-Droid Server

@insertcopying
@end ifnottex

@menu
* Overview::
* System Requirements::
* Setup::
* Simple Binary Repository::
* Building Applications::
* Importing Applications::
* Metadata::
* Update Processing::
* Build Server::
* Signing::
* GNU Free Documentation License::
* Index::
@end menu

@node Overview
@chapter Overview

The F-Droid server tools provide various scripts and tools that are used
to maintain the main F-Droid application repository. You can use these same
tools to create your own additional or alternative repository for publishing,
or to assist in creating, testing and submitting metadata to the main
repository.


@node System Requirements
@chapter System Requirements

@cindex installation

The system requirements for using the tools will vary depending on your
intended usage. At the very least, you'll need:

@itemize @bullet
@item
GNU/Linux
@item
Python 2.x
@item
The Android SDK Tools and Build-tools. 
Note that F-Droid does not assume that you have the Android SDK in your 
@code{PATH}: these directories will be specified in your repository 
configuration. Recent revisions of the SDK have @code{aapt} located in 
android-sdk/build-tools/ and it may be necessary to make a symlink to it in 
android-sdk/platform-tools/
@end itemize

If you intend to build applications from source you'll also need most, if not
all, of the following:

@itemize @bullet
@item
All SDK platforms requested by the apps you want to build
(The Android SDK is made available by Google under a proprietary license but
within that, the SDK platforms, support library and some other components are 
under the Apache license and source code is provided. 
Google APIs, used for building apps using Google Maps, are free to the extent 
that the library comes pre-installed on the device.
Google Play Services, Google Admob and others are proprietary and shouldn't be
included in the main F-Droid repository.)
@item
A version of the Android NDK
@item
Ant
@item
Ant Contrib Tasks (Debian package ant-contrib)
@item
Maven (Debian package maven)
@item
JavaCC (Debian package javacc)
@item
JDK (Debian package openjdk-6-jdk): openjdk-6 is recommended though openjdk-7
should work too
@item
VCS clients: svn, git, git-svn, hg, bzr
@item
Miscellaneous packages listed in 
buildserver/cookbooks/fdroidbuild-general/recipes/default.rb
of the F-Droid server repository
@item
A keystore for holding release keys. (Safe, secure and well backed up!)
@end itemize

If you intend to use the 'Build Server' system, for secure and clean builds
(highly recommended), you will also need:

@itemize @bullet
@item
VirtualBox (debian package virtualbox)
@item
Ruby (debian packages ruby and rubygems)
@item
Vagrant (gem install vagrant)
@item
Paramiko (debian package python-paramiko)
@item
Imaging (debian package python-imaging)
@item
Magic (debian package python-magic)
@end itemize


@node Setup
@chapter Setup

@cindex setup, installation

Because the tools and data will always change rapidly, you will almost
certainly want to work from a git clone of the tools at this stage. To
get started:

@example
git clone git://gitorious.org/f-droid/fdroidserver.git
@end example

You now have lots of stuff in the fdroidserver directory, but the most
important is the 'fdroid' command script which you run to perform all tasks.
This script is always run from a repository data directory, so the
most sensible thing to do next is to put your new fdroidserver directory
in your @code{PATH}.

@section Data

To do anything, you'll need at least one repository data directory. It's
from this directory that you run the @code{fdroid} command to perform all
repository management tasks. You can either create a brand new one, or
grab a copy of the data used by the main F-Droid repository:

@example
git clone git://gitorious.org/f-droid/fdroiddata.git
@end example

Regardless of the intended usage of the tools, you will always need to set
up some basic configuration details. This is done by creating a file called
@code{config.py} in the data directory. You should do this by copying the
example file (@code{config.sample.py}) from the fdroidserver project to your
data directory and then editing according to the instructions within.

Once configured in this way, all the functionality of the tools is accessed
by running the @code{fdroid} command. Run it on its own to get a list of the
available sub-commands.

You can follow any command with @code{--help} to get a list of additional
options available for that command.

@example
fdroid update --help
@end example


@node Simple Binary Repository
@chapter Simple Binary Repository

@cindex binary

If you want to maintain a simple repository hosting only binary APKs obtained
and compiled elsewhere, the process is quite simple:

@enumerate
@item
Set up the server tools, as described in Setup.
@item
Make a directory for your repository. This is the directory from which you
will do all the work with your repository. Create a config file there, called
@code{config.py}, by copying the @code{config.sample.py} from the server
project and editing it.
@item
Within that, make a directory called @code{repo} and put APK files in it.
@item
Run @code{fdroid update}.
@item
If it reports that any metadata files are missing, you can create them
in the @code{metadata} directory and run it again.
@item
To ease creation of metadata files, run @code{fdroid update} with the @code{-c}
option. It will create 'skeleton' metadata files that are missing, and you can
then just edit them and fill in the details.
@item
Then, if you've changed things, run @code{fdroid update} again.
@item
Running @code{fdroid update} adds an Icons directory into the repo directory,
and also creates the repository index (index.xml, and also index.jar if you've
configured the system to use a signed index).
@item
Publish the resulting contents of the @code{repo} directory to your web server.
@end enumerate

Following the above process will result in a @code{repo} directory, which you
simply need to push to any HTTP (or preferably HTTPS) server to make it
accessible.

While some information about the applications (and versions thereof) is
retrieved directly from the APK files, most comes from the corresponding file
in the @code{metadata} directory. The metadata file covering ALL versions of a
particular application is named @code{package.id.txt} where package.id is the
unique identifier for that package.

See the Metadata chapter for details of what goes in the metadata file. All
fields are relevant for binary APKs, EXCEPT for 'Build Version' entries, which
should be omitted.


@node Building Applications
@chapter Building Applications

Instead of (or as well as) including binary APKs from external sources in a
repository, you can build them directly from the source code.

Using this method, it is is possible to verify that the application builds
correctly, corresponds to the source code, and contains only free software.
Unforunately, in the Android world, it seems to be very common for an
application supplied as a binary APK to present itself as Free Software
when in fact some or all of the following are true:

@enumerate
@item
The source code (either for a particular version, or even all versions!) is
unavailable or incomplete.
@item
The source code is not capable of producing the actual binary supplied.
@item
The 'source code' contains binary files of unknown origin, or with proprietary
licenses.
@end enumerate

For this reason, source-built applications are the preferred method for the
main F-Droid repository, although occasionally for technical or historical
reasons, exceptions are made to this policy.

When building applications from source, it should be noted that you will be
signing them (all APK files must be signed to be installable on Android) with
your own key. When an application is already installed on a device, it is not
possible to upgrade it in place to a new version signed with a different key
without first uninstalling the original. This may present an inconvenience to
users, as the process of uninstalling loses any data associated with the
previous installation.

The process for managing a repository for built-from-source applications is
very similar to that described in the Simple Binary Repository chapter,
except now you need to:

@enumerate
@item
Include Build Version entries in the metadata files.
@item
Run @code{fdroid build} to build any applications that are not already built.
@item
Run @code{fdroid publish} to finalise packaging and sign any APKs that have
been built.
@end enumerate


@section More about "fdroid build"

When run without any parameters, @code{fdroid build} will build any and all
versions of applications that you don't already have in the @code{repo}
directory (or more accurately, the @code{unsigned} directory). There are various
other things you can do. As with all the tools, the @code{--help} option is
your friend, but a few annotated examples and discussion of the more common
usage modes follows:

To build a single version of a single application, you could run the
following:

@example
./fdroid build --package=org.fdroid.fdroid --vercode 16
@end example

This attempts to build version code 16 (which is version 0.25) of the F-Droid
client. Many of the tools recognise this @code{--package} parameter, allowing
their activity to be limited to just a single package.

If the build above was successful, two files will have been placed in the
@code{unsigned} directory:

@example
org.fdroid.fdroid_16.apk
org.fdroid.fdroid_16_src.tar.gz
@end example

The first is the (unsigned) APK. You could sign this with a debug key and push
it direct to your device or an emulator for testing. The second is a source
tarball containing exactly the source that was used to generate the binary.

If you were intending to publish these files, you could then run:

@example
./fdroid publish
@end example

The source tarball would move to the @code{repo} directory (which is the
directory you would push to your web server). A signed and zip-aligned version
of the APK would also appear there, and both files would be removed from the
@code{unsigned} directory.

If you're building purely for the purposes of testing, and not intending to
push the results to a repository, at least yet, the @code{--test} option can be
used to direct output to the @code{tmp} directory instead of @code{unsigned}.
A similar effect could by achieved by simply deleting the output files from
@code{unsigned} after the build, but with the risk of forgetting to do so!

Along similar lines (and only in conjunction with @code{--test}, you can use
@code{--force} to force a build of a Disabled application, where normally it 
would be completely ignored. Similarly a version that was found to contain 
ELFs or known non-free libraries can be forced to build. See also — 
scanignore= in the Build Version section.

If the build was unsuccessful, you can find out why by looking at the output 
in the logs/ directory. If that isn't illuminating, try building the app the 
regular way, step by step: android update project, ndk-build, ant debug.

Note that source code repositories often contain prebuilt libraries. If the 
app is being considered for the main F-Droid repository, it is important that 
all such prebuilts are built either via the metadata or by a reputable third 
party.


@section Direct Installation

You can also build and install directly to a connected device or emulator using
the @code{--install} switch. If you do this without using @code{--package} and
@code{--vercode} then all versions of all packages will be installed (with each
individual version overwriting the previous!). In most cases, this will not be
what you want to do, so execution will stop straight away. However, you can
override this if you're sure that's what you want, by using @code{--all}.
Note that currently, no sanity checks are performed with this mode, so that if 
the version is incorrect or that if the package name is different, you won't 
be informed.


@node Importing Applications
@chapter Importing Applications

To help with starting work on including a new application, @code{fdroid import}
will take a URL and optionally some other parameters, and attempt to construct
as much information as possible by analysing the source code. Basic usage is:

@example
./fdroid import --url=http://address.of.project
@end example

For this to work, the URL must point to a project format that the script
understands. Currently this is limited to one of the following:

@enumerate
@item
Gitorious - @code{https://gitorious.org/PROJECTNAME/REPONAME}
@item
Github - @code{https://github.com/USER/PROJECT}
@item
Google Code - @code{http://code.google.com/p/PROJECT/}
Supports git, svn and hg repos.

Some Google Code projects have multiple repositories, identified by a
dropdown list on the @code{source/checkout} page. To access one other than
the default, specify its name using the @code{--repo} switch.
@item
Bitbucket - @code{https://bitbucket.org/USER/PROJECT/}
@item
Git - @code{git://REPO}
@end enumerate

Depending on the project type, more or less information may be gathered. For
example, the license will be retrieved from a Google Code project, but not a
GitHub one. A bare repo url, such as the git:// one, is the least preferable
optional of all, since you will have to enter much more information manually.

If the import is successful, a metadata file will be created. You will need to
edit this further to check the information, and fill in the blanks.

If it fails, you'll be told why. If it got as far as retrieving the source
code, you can inspect it further by looking in @code{tmp/importer} where a full
checkout will exist.

A frequent cause of initial failure is that the project directory is actually
a subdirectory in the repository. In this case, run the importer again using
the @code{--subdir} option to tell it where. It will not attempt to determine
this automatically, since there may be several options.


@node Metadata
@chapter Metadata

@cindex metadata

Information used by update.py to compile the public index comes from two
sources:

@enumerate
@item
the APK files in the repo directory, and
@item
the metadata files in the metadata directory.
@end enumerate

The metadata files are simple, easy to edit text files, always named as the
application's package ID with '.txt' appended.

Note that although the metadata files are designed to be easily read and
writable by humans, they are also processed and written by various scripts.
They are capable of rewriting the entire file when necessary. Even so,
the structure and comments will be preserved correctly, although the order
of fields will be standardised. (In the event that the original file was
in a different order, comments are considered as being attached to the field
following them). In fact, you can standardise all the metadata in a single
command, without changing the functional content, by running:

@example
fdroid rewritemetadata
@end example

The following sections describe the fields recognised within the file.

@menu
* Category::
* License::
* Name::
* Auto Name::
* Web Site::
* Source Code::
* Issue Tracker::
* Donate::
* FlattrID::
* Bitcoin::
* Litecoin::
* Summary::
* Description::
* Repo Type::
* Repo::
* Build Version::
* AntiFeatures::
* Disabled::
* Requires Root::
* Archive Policy::
* Update Check Mode::
* Update Check Data::
* Auto Update Mode::
* Current Version::
* Current Version Code::
* No Source Since::
@end menu

@node Category
@section Category

A single category for the application to be placed in. There is no fixed list
of categories - both the client and the web site will automatically show any
categories that exist in any applications. However, if your metadata is
intended for the main F-Droid repository, you should use one of the existing
categories (look at the site/client), or discuss the proposal to add
a new one.

Additional categories can be specified, by using ';' as a separator. In this
case the first is the primary category, and the only one that will be seen by
clients that only understand one.

@node License
@section License

@cindex license

The overall license for the application, or in certain cases, for the 
source code only. 

Common values: 

@itemize @bullet

@item
@samp{GPLv2}
GNU GPL version 2

@item
@samp{GPLv2+}
GNU GPL version 2 or later

@item
@samp{GPLv3}
GNU GPL version 3

@item
@samp{GPLv3+}
GNU GPL version 3 or later

@item
@samp{GPL}
An unspecified GPL version. Use this only as a last resort or if there is 
some confusion over compatiblity of component licenses: particularly the use of 
Apache libraries with GPLv2 source code.

@item
@samp{AGPL}
Afferro GPL version 3.

@item
@samp{Apache2}
Apache 2

@item
@samp{MIT}
MIT X11 license

@item
@samp{BSD}
BSD license - the original '4-clause' version.

@item
@samp{NewBSD}
BSD license - the new, or modified, version.

@end itemize

@node Auto Name
@section Auto Name

@cindex Auto Name

The name of the application as can best be retrieved from the source code.
This is done so that the commitupdates script can put a familiar name in the
description of commits created when a new update of the application is
found. The Auto Name entry is generated automatically when @code{fdroid
checkupdates} is run.

@node Name
@section Name

@cindex Name

The name of the application. Normally, this field should not be present since 
the application's correct name is retrieved from the APK file. However, in a 
situation where an APK contains a bad or missing application name, it can be 
overridden using this. Note that this only overrides the name in the list of 
apps presented in the client; it doesn't changed the name or application label
in the source code.

@node Web Site
@section Web Site

@cindex Web Site

The URL for the application's web site.

@node Source Code
@section Source Code

@cindex Source Code

The URL to view or obtain the application's source code. This should be
something human-friendly. Machine-readable source-code is covered in the
'Repo' field.

@node Issue Tracker
@section Issue Tracker

@cindex Issue Tracker

The URL for the application's issue tracker. Optional, since not all
applications have one.

@node Donate
@section Donate

@cindex Donate

The URL to donate to the project. This should be the project's donate page
if it has one.

It is possible to use a direct PayPal link here, if that is all that is
available. However, bear in mind that the developer may not be aware of
that direct link, and if they later changed to a different PayPal account,
or the PayPal link format changed, things could go wrong. It is always
best to use a link that the developer explicitly makes public, rather than
something that is auto-generated 'button code'.

@node FlattrID
@section FlattrID

@cindex FlattrID

The project's Flattr (http://flattr.com) ID, if it has one. This should be
a numeric ID, such that (for example) https://flattr.com/thing/xxxx leads
directly to the page to donate to the project.

@node Bitcoin
@section Bitcoin

@cindex Bitcoin

A bitcoin address for donating to the project.

@node Litecoin
@section Litecoin

@cindex Litecoin

A litecoin address for donating to the project.

@node Summary
@section Summary

@cindex Summary

A brief summary of what the application is. Since the summary is only allowed 
one line on the list of the F-Droid client, keeping it to within 32 characters 
will ensure it fits even on the smallest screens.

@node Description
@section Description

@cindex Description

A full description of the application, relevant to the latest version. 
This can span multiple lines (which should be kept to a maximum of 80 
characters), and is terminated by a line containing a single '.'.

Basic MediaWiki-style formatting can be used. Leaving a blank line starts a
new paragraph. Surrounding text with @code{''} make it italic, and with
@code{'''} makes it bold.

You can link to another app in the repo by using @code{[[app.id]]}. The link
will be made appropriately whether in the Android client, the web repo
browser or the wiki. The link text will be the apps name.

Links to web addresses can be done using @code{[http://example.com Text]}.

Bulletted lists are done by simply starting each item with a @code{*} on
a new line, and numbered lists are the same but using @code{#}. There is
currently no support for nesting lists - you can have one level only.

It can be helpful to note information pertaining to updating from an 
earlier version; whether the app contains any prebuilts built by the 
upstream developers or whether non-free elements were removed; whether the 
app is in rapid development or whether the latest version lags behind the 
current version; whether the app supports multiple architectures or whether 
there is a maximum SDK specified (such info not being recorded in the index).

@node Repo Type
@section Repo Type

@cindex Repo Type

The type of repository - for automatic building from source. If this is not
specified, automatic building is disabled for this application. Possible
values are:

@itemize @bullet
@item
@samp{git}
@item
@samp{svn}
@item
@samp{git-svn}
@item
@samp{hg}
@item
@samp{bzr}
@item
@samp{srclib}
@end itemize
@node Repo
@section Repo

@cindex Repo

The repository location. Usually a git: or svn: URL, for example.

The git-svn option connects to an SVN repository, and you specify the URL in
exactly the same way, but git is used as a back-end. This is preferable for
performance reasons, and also because a local copy of the entire history is
available in case the upstream repository disappears. (It happens!). In
order to use Tags as update check mode for this VCS type, the URL must have
the tags= special argument set. Likewise, if you intend to use the
RepoManifest/branch scheme, you would want to specify branches= as well.
Finally, trunk= can also be added. All these special arguments will be passed
to "git svn" in order, and their values must be relative paths to the svn repo
root dir.
Here's an example of a complex git-svn Repo URL:
http://svn.code.sf.net/p/project/code/svn;trunk=trunk;tags=tags;branches=branches

For a Subversion repo that requires authentication, you can precede the repo
URL with username:password@ and those parameters will be passed as @option{--username}
and @option{--password} to the SVN checkout command. (This works only for
plain svn and not for git-svn - one of the very few cases where using svn is
advisable).

If the Repo Type is @code{srclib}, then you must specify the name of the
according srclib .txt file. For example if @code{scrlibs/FooBar.txt} exist
and you want to use this srclib, then you have to set Repo to
@code{FooBar}.

@node Build Version
@section Build Version

@cindex Build Version

Any number of these fields can be present, each specifying a version to
automatically build from source. The value is a comma-separated list.
For example:

@samp{Build Version:0.12,3,651696a49be2cd7db5ce6a2fa8185e31f9a20035}

The above specifies to build version 0.12, which has a version code of 3.
The third parameter specifies the tag, commit or revision number from
which to build it in the source repository.

In addition to the three, always required, parameters described above,
further parameters can be added (in name=value format) to apply further
configuration to the build. These are (roughly in order of application):

@table @code

@item disable=<message>
Disables this build, giving a reason why. (For backwards compatibility, this
can also be achieved by starting the commit ID with '!')

The purpose of this feature is to allow non-buildable releases (e.g. the source
is not published) to be flagged, so the scripts don't generate repeated
messages about them. (And also to record the information for review later).
If an apk has already been built, disabling causes it to be deleted once 
@code{fdroid update} is run; this is the procedure if ever a version has to 
be replaced.

@item subdir=<path>
Specifies to build from a subdirectory of the checked out source code.
Normally this directory is changed to before building,

@item submodules=yes
Use if the project (git only) has submodules - causes git submodule
init and update to be executed after the source is cloned.

@item init=xxxx
As for 'prebuild', but runs on the source code BEFORE any other processing
takes place.

You can use $$SDK$$, $$NDK$$ and $$MVN3$$ to substitute the paths to the
android SDK and NDK directories, and maven 3 executable respectively.

@item oldsdkloc=yes
The sdk location in the repo is in an old format, or the build.xml is
expecting such. The 'new' format is sdk.dir while the VERY OLD format
is sdk-location. Typically, if you get a message along the lines of:
"com.android.ant.SetupTask cannot be found" when trying to build, then
try enabling this option.

@item target=<target>
Specifies a particular SDK target for compilation, overriding the 
project.properties of the app and possibly sub-projects. Note that this does 
not change the target SDK in the AndroidManifest.xml — the level of features 
that can be included in the build. This is likely to cause the whole build.xml 
to be rewritten, which is fine if it's a 'standard' android file or doesn't 
already exist, but not a good idea if it's heavily customised. If you get an 
error about invalid target, first try @code{init=rm -rf bin/}; otherwise this 
parameter should do the trick.

Please note that gradle builds should be using compilesdk=.

@item compilesdk=<level>
Practically accomplishes the same that target= does when used in ant and maven
projects. compilesdk= is used rather than target= so as to not cause any more
confusion. It only takes effect on gradle builds in the build.gradle file,
thus using it in any other case is not wise.

@item update=xxx
By default, 'android update project' is used to generate or update the
build.xml file. Specifying update=no bypasses that.

Specifiying update=force forces rebuilding of the build.xml file at the
same time - this is frequently needed with r14 of the Android platform
tools. Be aware of any customisations in build.xml when using
update=force.

Otherwise, value can be a semicolon-separated list of directories in
which to run 'android update project' relative to the main
application directory (which may include '@code{subdir}' parameter).
Default value is '@code{.}', and passing non-default value may be
useful for multi-component projects. Note that @code{--subprojects}
switch is automatically passed to 'android update project', so using
explicit list may be needed only for peculiar source layouts.

@item encoding=xxxx
Adds a java.encoding property to local.properties with the given
value. Generally the value will be 'utf-8'. This is picked up by the
SDK's ant rules, and forces the Java compiler to interpret source
files with this encoding. If you receive warnings during the compile
about character encodings, you probably need this.

@item forceversion=yes
If specified, the package version in AndroidManifest.xml is replaced
with the version name for the build as specified in the metadata.

This is useful for cases when upstream repo failed to update it for
specific tag; to build an arbitrary revision; to make it apparent that 
the version differs significantly from upstream; or to make it apparent 
which architecture or platform the apk is designed to run on.

@item forcevercode=yes
If specified, the package version code in the AndroidManifest.xml is
replaced with the version code for the build. See also forceversion.

@item rm=<relpath1;relpath2;...>
Specifies the relative paths of files or directories to delete before
the build is done. The paths are relative to the base of the build
directory - i.e. the root of the directory structure checked out from
the source respository - not necessarily the directory that contains
AndroidManifest.xml.

Multiple files/directories can be specified by separating them with ';'.
Directories will be recursively deleted.

@item fixtrans=yes
Modifies any instances of string resources that use multiple
formatting arguments, but don't use positional notation. For example,
"Hello %s, %d" becomes "Hello %1$s, %2$d". Newer versions of the
Android platform tools enforce this sensible standard. If you get
error messages relating to that, you need to enable this.

@item fixapos=yes
Like fixtrans, but deals with an even older issue relating to
'unescaped apostrophes' in translation strings.

@item extlibs=a;b;c
Specifies a list of external libraries (jar files) from the
@code{build/extlib} library, which will be placed in the @code{libs} directory
of the project. Separate items with semicolons.

@item srclibs=a@@r;b@@r1;
Specifies a list of source libraries or Android projects. Separate items with 
semicolons, and each item is of the form name@@rev where name is the predefined 
source library name and rev is the revision or tag in source control to use. 

Each srclib has a metadata file under srclibs/ in the repository directory, 
and the source code is stored in build/srclib/. 
Repo Type: and Repo: are specified in the same way as for apps; Subdir: can be 
a comma separated list, for when directories are renamed by upstream; Update 
Project: updates the projects in the working directory and one level down; 
Prepare: can be used for any kind of preparation: in particular if you need to 
update the project with a particular target. You can then also use $$name$$ in 
the init/prebuild/build command to substitute the relative path to the library 
directory, but it could need tweaking if you've changed into another directory.

@item patch=x
Apply patch(es). 'x' names one (or more - comma-seperated)
files within a directory below the metadata, with the same
name as the metadata file but without the extension. Each of
these patches is applied to the code in turn.

@item prebuild=xxxx
Specifies a shell command (or commands - chain with &&) to run before
the build takes place. Backslash can be used as an escape character to
insert literal commas, or as the last character on a line to join that
line with the next. It has no special meaning in other contexts; in
particular, literal backslashes should not be escaped.

The command runs using bash.

Note that nothing should be build during this prebuild phase - scanning
of the code and building of the source tarball, for example, take place
after this. For custom actions that actually build things, use 'build'
instead.

You can use $$name$$ to substitute the path to a referenced srclib - see
the @code{srclib} directory for details of this.

You can use $$SDK$$, $$NDK$$ and $$MVN3$$ to substitute the paths to the
android SDK and NDK directories, and maven 3 executable respectively e.g. 
for when you need to run @code{android update project} explicitly.

@item initfun=yes
Enables a selection of mad hacks to make com.funambol.android build.
Probably not useful for any other application.

@item scanignore=path1;path2;...
Enables one or more files/paths to be exlcuded from the scan process.
This should only be used where there is a very good reason, and
probably accompanied by a comment explaining why it is necessary.

When scanning, files whose relative paths start with any of the paths
given here are ignored.

@item build=xxxx
As for 'prebuild', but runs during the actual build phase (but before the
main ant/maven build). Use this only for actions that do actual building.
Any prepartion of the source code should be done using 'init' or 'prebuild'.

Any building that takes place before build= will be ignored, as either ant,
mvn or gradle will be executed to clean the build environment right before
build= (or the final build) is run.

You can use $$SDK$$, $$NDK$$ and $$MVN3$$ to substitute the paths to the
android SDK and NDK directories, and maven 3 executable respectively.

@item buildjni=[yes|no|<dir list>]
Enables building of native code via the ndk-build script before doing
the main ant build. The value may be a list of directories relative
to the main application directory in which to run ndk-build, or 'yes'
which corresponds to '.' . Using explicit list may be useful to build
multi-component projects.

The build and scan processes will complain (refuse to build) if this
parameter is not defined, but there is a @code{jni} directory present.
If the native code is being built by other means, you can specify
@code{no} here to avoid that. However, if the native code is actually
not required, remove the directory instead (using @code{prebuild} for
example).

@item gradle=<flavour>[@@<dir>]
Build with gradle instead of ant, specifying what flavour to assemble.
If <flavour> is 'yes', 'main' or empty, no flavour will be used. Note
that this will not work on projects with flavours, since it will build
all flavours and there will be no 'main' build.
If @@<dir> is attached to <flavour>, then the gradle tasks will be run in that
directory. This might be necessary if gradle needs to be run in the parent
directory, in which case one would use 'gradle=<flavour>@..'.

@item maven=yes[@@<dir>]
Build with maven instead of ant. Like gradle, an extra @@<dir> tells f-droid
to run maven inside that relative subdirectory.

@item preassemble=<task1> <task2>
Space-separated list of gradle tasks to be run before the assemble task
in a gradle project build.

@item bindir=<path>
Normally the build output (apk) is expected to be in the bin
subdirectory below the ant build files. If the project is configured
to put it elsewhere, that can be specified here, relative to the base
of the checked out repo. Not yet implemented for gradle.

@item antcommand=xxx
Specify an alternate ant command (target) instead of the default
'release'. It can't be given any flags, such as the path to a build.xml.

@item novcheck=yes 
Don't check that the version name and code in the resulting apk are
correct by looking at the build output - assume the metadata is
correct. This takes away a useful level of sanity checking, and should
only be used if the values can't be extracted.

@end table

Another example, using extra parameters:

@samp{Build Version:1.09.03,10903,45,subdir=Timeriffic,oldsdkloc=yes}

@node AntiFeatures
@section AntiFeatures

@cindex AntiFeatures

This is optional - if present, it contains a comma-separated list of any of
the following values, describing an anti-feature the application has. 
Even though such apps won't be displayed unless a settings box is ticked, 
it is a good idea to mention the reasons for the anti-feature(s) in the 
description:

@itemize @bullet

@item
@samp{Ads} - the application contains advertising.

@item
@samp{Tracking} - the application tracks and reports your activity to
somewhere without your consent. It's commonly used for when developers 
obtain crash logs without the user's consent, or when an app is useless 
without some kind of authentication.

@item
@samp{NonFreeNet} - the application relies on computational services that 
are impossible to replace or that the replacement cannot be connected to 
without major changes to the app.

@item
@samp{NonFreeAdd} - the application promotes non-Free add-ons, such that the 
app is effectively an advert for other non-free software and such software is 
not clearly labelled as such.

@item
@samp{NonFreeDep} - the application depends on a non-Free application (e.g.
Google Maps) - i.e. it requires it to be installed on the device, but does not
include it.

@end itemize

@node Disabled
@section Disabled

@cindex Disabled

If this field is present, the application does not get put into the public
index. This allows metadata to be retained while an application is temporarily
disabled from being published. The value should be a description of why the
application is disabled. No apks or source code archives are deleted: to purge 
an apk see the Build Version section or delete manually for developer builds. 
The field is therefore used when an app has outlived it's usefulness, because 
the source tarball is retained.

@node Requires Root
@section Requires Root

@cindex Requires Root

Set this optional field to "Yes" if the application requires root
privileges to be usable. This lets the client filter it out if the
user so desires. Whether root is required or not, it is good to give 
a paragraph in the description to the conditions on which root may be 
asked for and the reason for it.

@node Update Check Mode
@section Update Check Mode

@cindex Update Check Mode

This determines the method using for determining when new releases are
available - in other words, the updating of the Current Version and Current
Version Code fields in the metadata by the @code{fdroid checkupdates} process.

Valid modes are:

@itemize
@item
@code{None} - No checking is done because there's no appropriate automated way
of doing so. Updates should be checked for manually. Use this, for example, 
when deploying betas or patched versions; when builds are done in a directory 
different to where the AndroidManifest.xml is; if the developers use the 
gradle build system and store version info in a separate file; if the 
developers make a new branch for each release and don't make tags; or if you've 
changed the package name or version code logic.
@item
@code{Static} - No checking is done - either development has ceased or new versions
are not desired. This method is also used when there is no other checking method 
available and the upstream developer keeps us posted on new versions.
@item
@code{RepoManifest} - At the most recent commit, the AndroidManifest.xml file 
is looked for in the directory where it was found in the the most recent build.
The appropriateness of this method depends on the development process used by 
the application's developers. You should not specify this method unless you're 
sure it's appropriate. For example, some developers bump the version when 
commencing development instead of when publishing.
It will return an error if the AndroidManifest.xml has moved to a different 
directory or if the package name has changed. 
The current version that it gives may not be accurate, since not all 
versions are fit to be published. Therefore, before building, it is often 
necessary to check if the current version has been published somewhere by the 
upstream developers, either by checking for apks that they distribute or for 
tags in the source code repository. 

It currently works for every repository type to different extents, except 
the srclib repo type. For git, git-svn and hg repo types, you may use 
"RepoManifest/yourbranch" as UCM so that "yourbranch" would be the branch used 
in place of the default one.  The default values are "master" for git, 
"default" for hg and none for git-svn (it stays in the same branch). 
On the other hand, branch support hasn't been implemented yet in bzr and svn, 
but RepoManifest may still be used without it.
@item
@code{RepoTrunk} - For svn and git-svn repositories, especially those who
don't have a bundled AndroidManifest.xml file, the Tags and RepoManifest
checks will not work, since there is no version information to obtain. But,
for those apps who automate their build process with the commit ref that HEAD
points to, RepoTrunk will set the Current Version and Current Version Code to
that number.
@item
@code{Tags} - The AndroidManifest.xml file in all tagged revisions in the
source repository is checked, looking for the highest version code. The
appropriateness of this method depends on the development process used by the
application's developers. You should not specify this method unless you're sure
it's appropriate. It shouldn't be used if the developers like to tag betas or 
are known to forget to tag releases. Like RepoManifest, it will not return the 
correct value if the directory containing the AndroidManifest.xml has moved.
Despite these caveats, it is the often the favourite update check mode. 

It currently only works for git, hg and git-svn repositories. In the case of
the latter, the repo URL must encode the path to the trunk and tags or else no
tags will be found.
@item
@code{HTTP} - HTTP requests are used to determine the current version code and
version name. This is controlled by the @code{Update Check Data} field, which
is of the form @code{urlcode|excode|urlver|exver}.

Firstly, if @code{urlcode} is non-empty, the document from that URL is
retrieved, and matched against the regular expression @code{excode}, with the
first group becoming the version code.

Secondly, if @code{urlver} is non-empty, the document from that URL is
retrieved, and matched against the regular expression @code{exver}, with the
first group becoming the version name. The @code{urlver} field can be set to
simply '.' which says to use the same document returned for the version code
again, rather than retrieving a different one.
@end itemize

@node Update Check Data
@section Update Check Data

@cindex Update Check Data

Used in conjunction with @code{Update Check Mode} for certain modes.

@node Archive Policy
@section Archive Policy

@cindex Archive Policy

This determines the policy for moving old versions of an app to the archive
repo, if one is configured. The configuration sets a default maximum number
of versions kept in the main repo, after which older ones are moved to the
archive. This app-specific policy setting can override that.

Currently the only supported format is "n versions", where n is the number
of versions to keep.

@node Auto Update Mode
@section Auto Update Mode

@cindex Auto Update Mode

This determines the method using for auto-generating new builds when new 
releases are available - in other words, adding a new Build Version line to the 
metadata.
This happens in conjunction with the 'Update Check Mode' functionality - i.e.
when an update is detected by that, it is also processed by this.

Valid modes are:

@itemize
@item
@code{None} - No auto-updating is done
@item
@code{Version} - Identifies the target commit (i.e. tag) for the new build based
on the given version specification, which is simply text in which %v and %c are
replaced with the required version name and version code respectively.

For example, if an app always has a tag "2.7.2" corresponding to version 2.7.2,
you would simply specify "Version %v". If an app always has a tag "ver_1234"
for a version with version code 1234, you would specify "Version ver_%c".

Additionally, a suffix can be added to the version name at this stage, to
differentiate F-Droid's build from the original. Continuing the first example
above, you would specify that as "Version +-fdroid %v" - "-fdroid" is the suffix.
@end itemize


@node Current Version
@section Current Version

@cindex Current Version

The name of the version that is current. There may be newer versions of the
application than this (e.g. betas), and there will almost certainly be older
ones. This should be the one that is recommended for general use. 
In the event that there is no source code for the current version, or that 
non-free libraries are being used, this would ideally be the latest 
version that is still free, though it may still be expedient to 
retain the automatic update check — see No Source Since.

This field is normally automatically updated - see Update Check Mode.

@node Current Version Code
@section Current Version Code

@cindex Current Version Code

The version code corresponding to the Current Version field. Both these fields
must be correct and matching although it's the current version code that's
used by Android to determine version order and by F-Droid client to determine
which version should be recommended.

This field is normally automatically updated - see Update Check Mode.

@node No Source Since
@section No Source Since

@cindex No Source Since

In case we are missing the source code for the Current Version reported by
Upstream, or that non-free elements have been introduced, this defines the 
first version that began to miss source code.
Apps that are missing source code for just one or a few versions, but provide
source code for newer ones are not to be considered here - this field is
intended to illustrate which apps do not currently distribute source code, and
since when have they been doing so.

@node Update Processing
@chapter Update Processing

@section Detecting

There are various mechanisms in place for automatically detecting that updates
are available for applications, with the @code{Update Check Mode} field in the
metadata determining which method is used for a particular application.

Running the @code{fdroid checkupdates} command will apply this method to each
application in the repository and update the @code{Current Version} and
@code{Current Version Code} fields in the metadata accordingly.

As usual, the @code{-p} option can be used with this, to restrict processing
to a particular application.

Note that this only updates the metadata such that we know what the current
published/recommended version is. It doesn't make that version available in
the repository - for that, see the next section.

@section Adding

Adding updates (i.e. new versions of applications already included in the
repository) happens in two ways. The simple case is applications where the
APK files are binaries, retrieved from a developer's published build. In this
case, all that's required is to place the new binary in the @code{Repo}
directory, and the next run of @code{fdroid update} will pick it up.

For applications built from source, it is necessary to add a new
@code{Build Version} line to the metadata file. At the very least, the version
name, version code and commit will be different. It is also possible that the
additional build flags will change between versions.

For processing multiple updates in the metadata at once, it can be useful to
run @code{fdroid update --interactive}. This will check all the applications
in the repository, and where updates are required you will be prompted to
[E]dit the metadata, [I]gnore the update, or [Q]uit altogether.

@node Build Server
@chapter Build Server

The Build Server system isolates the builds for each package within a clean,
isolated and secure throwaway virtual machine environment.

@section Overview

Building applications in this manner on a large scale, especially with the
involvement of automated and/or unattended processes, could be considered
a dangerous pastime from a security perspective. This is even more the case
when the products of the build are also distributed widely and in a
semi-automated ("you have updates available") fashion.

Assume that an upstream source repository is compromised. A small selection
of things that an attacker could do in such a situation:

@enumerate
@item
Use custom ant build steps to execute virtually anything as the user doing
the build.
@item
Access the keystore.
@item
Modify the built apk files or source tarballs for other applications in the
repository.
@item
Modify the metadata (which includes build scripts, which again, also includes
the ability to execute anything) for other applications in the repository.
@end enumerate

Through complete isolation, the repurcussions are at least limited to the
application in question. Not only is the build environment fresh for each
build, and thrown away afterwards, but it is also isolated from the signing
environment.

Aside from security issues, there are some applications which have strange
requirements such as custom versions of the NDK. It would be impractical (or
at least extremely messy) to start modifying and restoring the SDK on a
multi-purpose system, but within the confines of a throwaway single-use
virtual machine, anything is possible.

All this is in addition to the obvious advantage of having a standardised
and completely reproducible environment in which builds are made. Additionally,
it allows for specialised custom build environments for particular
applications.

@section Setting up a build server

In addition to the basic setup previously described, you will also need
a Vagrant-compatible Ubuntu Raring base box called 'raring32' (or raring64
for a 64-bit VM, if you want it to be much slower, and require more disk
space).

You can use a different version or distro for the base box, so long as you
don't expect any help making it work. One thing to be aware of is that
working copies of source trees are moved from the host to the guest, so
for example, having subversion v1.6 on the host and v1.7 on the guest
would fail.

Unless you're very trusting. you should create one of these for yourself
from verified standard Ubuntu installation media. However, you could skip
over the next few paragraphs (and sacrifice some security) by downloading 
@url{https://f-droid.org/raring32.box} or @url{https://f-droid.org/raring64.box}.

Documentation for creating a base box can be found at
@url{http://docs.vagrantup.com/v1/docs/base_boxes.html}.

In addition to carefully following the steps described there, you should
consider the following:

@enumerate
@item
It is advisable to disable udev network device persistence, otherwise any
movement of the VM between machines, or reconfiguration, will result in
broken networking.

For a Debian/Ubuntu default install, just
@code{touch /etc/udev/rules.d/75-persistent-net-generator.rules} to turn
off rule generation, and at the same time, get rid of any rules it's
already created in @code{/etc/udev/rules.d/70-persistent-net.rules}.
@item
Unless you want the VM to become totally inaccessible following a failed
boot, you need to set @code{GRUB_RECORDFAIL_TIMEOUT} to a value other than
-1 in @code{/etc/grub/default} and then run @code{update-grub}.
@end enumerate

You may also want to edit @code{buildserver/Vagrantfile} - in particular
there is a path for retrieving the base box if it doesn't exist, and an
apt proxy definition, both of which may need customising for your
environment.

With this base box available, you should then create @code{makebs.config.py},
using @code{makebs.config.sample.py} as a reference - look at the settings and
documentation there to decide if any need changing to suit your environment.
You can then go to the @code{fdroidserver} directory and run this:

@example
./makebuildserver.py
@end example

This will take a long time, and use a lot of bandwidth - most of it spent
installing the necessary parts of the Android SDK for all the various
platforms. Luckily you only need to do it occasionally. Once you have a
working build server image, if the recipes change (e.g. when packages need
to be added) you can just run that script again and the existing one will
be updated in place.

The main sdk/ndk downloads will automatically be cached to speed things
up the next time, but there's no easy way of doing this for the longer
sections which use the SDK's @code{android} tool to install platforms,
add-ons and tools. However, instead of allowing automatic caching, you
can supply a pre-populated cache directory which includes not only these
downloads, but also .tar.gz files for all the relevant additions. If the
provisioning scripts detect these, they will be used in preference to
running the android tools. For example, if you have
@code{buildserver/addons/cache/platforms/android-15.tar.gz} that will be
used when installing the android-15 platform, instead of re-downloading it
using @code{android update sdk --no-ui -t android-15}.

Once it's complete you'll have a new base box called 'buildserver' which is
what's used for the actual builds. You can then build packages as normal,
but with the addition of the @code{--server} flag to @code{fdroid build} to
instruct it to do all the hard work within the virtual machine.

The first time a build is done, a new virtual machine is created using the
'buildserver' box as a base. A snapshot of this clean machine state is saved
for use in future builds, to improve performance. You can force discarding
of this snapshot and rebuilding from scratch using the @code{--resetserver}
switch with @code{fdroid build}.

@node Signing
@chapter Signing

There are two kinds of signing involved in running a repository - the signing
of the APK files generated from source builds, and the signing of the repo
index itself. The latter is optional, but very strongly recommended.

@section Repo Index Signing

When setting up the repository, one of the first steps should be to generate
a signing key for the repository index. This will also create a keystore, which
is a file that can be used to hold this and all other keys used. Consider the
location, security and backup status of this file carefully, then create it as
follows:

@code{keytool -genkey -v -keystore my.keystore -alias repokey -keyalg RSA -keysize 2048 -validity 10000}

In the above, replace 'my.keystore' with the name of the keystore file to be
created, and 'repokey' with a name to identify the repo index key by.

You'll be asked for a password for the keystore, AND a password for the key.
They shouldn't be the same. In between, you'll be asked for some identifying
details which will go in the certificate.

The two passwords entered go into @code{config.py}, as @code{keystorepass} and
@code{keypass} respectively. The path to the keystore file, and the alias you
chose for the key also go into that file, as @code{keystore} and
@code{repo_keyalias} respectively.

@section Package Signing

With the repo index signing configured, all that remains to be done for package
signing to work is to set the @code{keydname} field in @code{config.py} to
contain the same identifying details you entered before.

A new key will be generated using these details, for each application that is
built. (If a specific key is required for a particular application, this system
can be overridden using the @code{keyaliases} config settings.


@node GNU Free Documentation License
@appendix GNU Free Documentation License

@include fdl.texi

@node Index
@unnumbered Index

@printindex cp

@bye