// file : doc/intro.cli // copyright : Copyright (c) 2014-2018 Code Synthesis Ltd // license : MIT; see accompanying LICENSE file "\name=build2-toolchain-intro" "\subject=toolchain" "\title=Toolchain Introduction" // TODO // // @@ refs to further docs // // STYLE // // @@ section boundary page breaks (
) // @@ when printed, code background is gone, but spaces still there // // PDF // // @@ tree output is garbled // @@ Could we use a nicer font, seeing that we embed them? // // NOTES // // - Maximum
 line is 70 characters.
//

"
\h0#preface|Preface|

This document is an overall introduction to the \c{build2} toolchain that
shows how the main components, namely the build system, the package dependency
manager, and the project dependency manager are used together to handle the
entire C++ project development lifecycle: creation, development, testing, and
delivery. For additional information, including documentation for individual
toolchain components, man pages, etc., refer to the \c{build2} project
\l{https://build2.org/doc.xhtml Documentation} page.


\h1#tldr|TL;DR|

\
$ git clone ssh://example.org/hello.git
$ tree hello
hello/
├── hello/
│   ├── hello.cxx
│   └── buildfile
├── manifest
└── repositories.manifest

$ cd hello
$ bdep init --config-create ../hello-gcc cc config.cxx=g++
initializing in project /tmp/hello/
created configuration /tmp/hello-gcc/ default,auto-synchronized
synchronizing:
  new hello/0.1.0

$ b
c++ hello/cxx{hello}@../hello-gcc/hello/hello/
ld ../hello-gcc/hello/hello/exe{hello}
ln ../hello-gcc/hello/hello/exe{hello} -> hello/

$ hello/hello World
Hello, World!

$ edit repositories.manifest   # add https://example.org/libhello.git
$ edit manifest                # add 'depends: libhello ^1.0.0'
$ edit hello/buildfile         # import libhello
$ edit hello/hello.cxx         # use libhello

$ b
fetching from https://example.org/libhello.git
synchronizing /tmp/hello-gcc/:
  new libhello/1.0.0 (required by hello)
  reconfigure hello/0.1.0
c++ ../hello-gcc/libhello-1.0.0/libhello/cxx{hello}
ld ../hello-gcc/libhello-1.0.0/libhello/libs{hello}
c++ hello/cxx{hello}@../hello-gcc/hello/hello/
ld ../hello-gcc/hello/hello/exe{hello}
ln ../hello-gcc/hello/hello/exe{hello} -> hello/

$ bdep fetch                   # refresh available versions
$ bdep status -i               # review available versions
hello configured 0.1.0
  libhello ^1.0.0 configured 1.0.0 available [1.1.0]

$ bdep sync libhello           # upgrade to latest
synchronizing:
  new libformat/1.0.0 (required by libhello)
  new libprint/1.0.0 (required by libhello)
  upgrade libhello/1.1.0
  reconfigure hello/0.1.0

$ bdep sync libhello/1.0.0     # downgrade
synchronizing:
  drop libprint/1.0.0 (unused)
  drop libformat/1.0.0 (unused)
  downgrade libhello/1.0.0
  reconfigure hello/0.1.0
\

\h1#guide|Getting Started Guide|

The aim of this guide is to get you started developing C/C++ projects with the
\c{build2} toolchain. All the examples in this section include the relevant
command output so if you just want to get a sense of what \c{build2} is about,
then you don't have to install the toolchain and run the commands in order to
follow along. Or, alternatively, you can take a short detour to the
\l{https://build2.org/install.xhtml Installation Instructions} and then try
the examples for yourself.

One of the primary goals of the \c{build2} toolchain is to provide a uniform
interface across all the platforms and compilers. While most of the examples
in this document assume a UNIX-like operation system, they will look pretty
similar if you are on Windows. You just have to use appropriate paths,
compilers, and options.

The question we will try to answer in this section can be summarized as:

\
$ git clone .../hello.git && now-what?
\

That is, we clone an existing C/C++ project or would like to create a new one
and then start hacking on it. We want to spend as little time and energy as
possible on the initial and ongoing infrastructure maintenance: setting up
build configurations, managing dependencies, continuous integration and
testing, release management, etc. Or, as one C++ user aptly put it, \"\i{All I
want to do is program.}\"

\h#guide-hello|Hello, World|

Let's see what programming with \c{build2} feels like by starting with a
customary \i{\"Hello, World!\"} program (here we assume our current working
directory is \c{/tmp}):

\
$ bdep new -t exe -l c++ hello
created new executable project hello in /tmp/hello/
\

The \l{bdep-new(1)} command creates a \i{canonical} \c{build2} project. In
our case it is an executable implemented in C++.

\N|To create a library, pass \c{-t\ lib}. By default \c{new} also initializes
a \c{git} repository and generates suitable \c{.gitignore} files (pass \c{-s\
none} if you don't want that). And for details on naming your projects, see
\l{bpkg#package-name Package Name}.|

\N|Note to Windows users: the \c{build2-baseutils} package includes core
\c{git} utilities that are sufficient for the \c{bdep} functionality.|

Let's take a look inside our new project:

\
$ tree hello
hello/
├── .git/
├── .bdep/
├── build/
├── hello/
│   ├── hello.cxx
│   ├── buildfile
│   └── testscript
├── buildfile
├── manifest
└── repositories.manifest
\

\N|While the canonical project structure is strongly recommended, especially
for new projects, \c{build2} is flexible enough to allow most commonly used
arrangements.|

Similar to version control tools, we normally run all \c{build2} tools from
the project's source directory or one of its subdirectories, so:

\
$ cd hello
\

While the project layout is discussed in more detail in later sections, let's
examine a couple of interesting files to get a sense of what's going on. We
start with the source file which should look familiar:

\
$ cat hello/hello.cxx

#include 

int main (int argc, char* argv[])
{
  using namespace std;

  if (argc < 2)
  {
    cerr << \"error: missing name\" << endl;
    return 1;
  }

  cout << \"Hello, \" << argv[1] << '!' << endl;
}
\

\N|If you prefer the \c{.?pp} extensions over \c{.?xx} for your C++ source
files, pass \c{-l\ c++,cpp} to the \c{new} command. See \l{bdep-new(1)} for
details on this and other customization options.|

Let's take a look at the accompanying \c{buildfile}:

\
$ cat hello/buildfile

libs =
#import libs += libhello%lib{hello}

exe{hello}: {hxx ixx txx cxx}{**} $libs testscript
\

As the name suggests, this file describes how to build things. While its
content might look a bit cryptic, let's try to infer a couple of points
without going into too much detail (the details are discussed in the following
sections). That \c{exe{hello\}} on the left of \c{:} is a \i{target}
(executable named \c{hello}) and what we have on the right are
\i{prerequisites} (C++ source files, libraries, etc). This \c{buildfile} uses
\l{b#name-patterns wildcard patterns} (that \c{**}) to automatically locate
all the C++ source files. This means we don't have to edit our \c{buildfile}
every time we add, remove, or rename a source file in our project. There also
appears to be some (commented out) infrastructure for importing and linking
libraries (that \c{libs} variable). We will see how to use it in a moment.

\N|In simple projects that follow the canonical structure we can often
completely ignore the presence of the build definition files thus approaching
the \i{build system-less} workflow found in languages like Rust and Go.|

Finally, the \c{buildfile} also lists \c{testscript} as a prerequisite of
\c{hello}. This file tests our target. Let's take a look inside:

\
$ cat hello/testscript

: basics
:
$* 'World' >'Hello, World!'

: missing-name
:
$* 2>>EOE != 0
error: missing name
EOE
\

Again, we are not going into detail here (see \l{testscript#intro Testscript
Introduction} for a proper introduction), but to give you an idea, here we
have two tests: the first (with id \c{basics}) verifies that our program
prints the expected greeting while the second makes sure it handles the
missing name error condition. Tests written in Testscript are concise,
portable, and executed in parallel.

Next up is \c{manifest}:

\
$ cat manifest
: 1
name: hello
version: 0.1.0-a.0.z
summary: hello executable
license: TODO
url: https://example.org/hello
email: you@example.org
#depends: libhello ^1.0.0
\

The \c{manifest} file is what makes a build system project a \i{package}. It
contains all the metadata that a user of a package might need to know: its
name, version, license, dependencies, etc., all in one place.

\N|Refer to \l{bpkg#manifest-format Manifest Format} for the general format of
\c{build2} manifest files and to \l{bpkg#manifest-package Package Manifest}
for details on the package manifest values.|

As you can see, \c{manifest} created by \l{bdep-new(1)} contains some dummy
values which you would want to adjust before publishing your package. But
let's resist the urge to adjust that strange looking \c{0.1.0-a.0.z} until we
discuss package versioning.

\N|Next to \c{manifest} you might have noticed the \c{repositories.manifest}
file \- we will discuss its function later, when we talk about dependencies
and where they come from.|

Project in hand, let's build it. Unlike other programming languages, C++
development usually involves juggling a handful of build configurations:
several compilers and/or targets (\c{build2} is big on cross-compiling),
debug/release, different sanitizers and/or static analysis tools, and so
on. As a result, \c{build2} is optimized for multi-configuration
usage. However, as we will see shortly, one build configuration can be
designated as the default with additional conveniences.

The \l{bdep-init(1)} command is used to initialize a project in a build
configuration. As a shortcut, it can also create a new build configuration in
the process, which is just what we need here. Let's start with GCC (remember
we are in the project's root directory):

\
$ bdep init -C ../hello-gcc @gcc cc config.cxx=g++
initializing in project /tmp/hello/
created configuration @gcc /tmp/hello-gcc/ default,auto-synchronized
synchronizing:
  new hello/0.1.0-a.0.19700101000000
\

The \cb{--create|-C} option instructs \c{init} to create a new configuration
in the specified directory (\c{../hello-gcc} in our case). To make referring
to configurations easier, we can give it a name, which is what we do with
\c{@gcc}. The next argument (\c{cc}, stands for \i{C-common}) is the build
system module we would like to configure. It implements compilation and
linking rules for the C and C++ languages. Finally, \c{config.cxx=g++} is (one
of) this module's configuration variables that specifies the C++ compiler we
would like to use (the corresponding C compiler will be determined
automatically). Let's for now also ignore that \c{synchronizing:...} bit along
with strange-looking \c{19700101000000} in the version \- it will become clear
what's going on here in a moment.

\N|Note to Windows users: a command line argument with leading \c{@} has
a special meaning in PowerShell. To work around this, you can use the
alternative \c{-@gcc} syntax or the \c{-n\ gcc} option.|

Now the same for Clang:

\
$ bdep init -C ../hello-clang @clang cc config.cxx=clang++
initializing in project /tmp/hello/
created configuration @clang /tmp/hello-clang/ auto-synchronized
synchronizing:
  new hello/0.1.0-a.0.19700101000000
\

If we check the parent directory, we should now see two build configurations
next to our project:

\
$ ls ..
hello/
hello-gcc/
hello-clang/
\

\N|If, as in the above examples, our configuration directories are next to the
project and their names are in the \c{\i{prj-name}\b{-}\i{cfg-name}} form,
then we can use the shortcut version of the \c{init} command:

\
$ bdep init -C @clang cc config.cxx=clang++
\

|

Things will also look pretty similar if you are on Windows instead of a
UNIX-like operating system. For example, to initialize our project on Windows
with Visual Studio, start the Visual Studio development command prompt and
then run:

\N|Currently we have to run \c{build2} tools from a suitable Visual Studio
development command prompt. This requirement will likely be removed in the
future.|

\
> bdep init -C ..\hello-debug @debug cc     ^
  config.cxx=cl                             ^
  \"config.cc.coptions=/MDd /Od /Zi\"         ^
  config.cc.loptions=/DEBUG

> bdep init -C ..\hello-release @release cc ^
  config.cxx=cl                             ^
  config.cc.coptions=/O2
\

\N|Besides the \c{coptions} (compile options) and \c{loptions} (link options),
other commonly used \c{cc} module configuration variables are \c{poptions}
(preprocess options) and \c{libs} (extra libraries to link). We can also use
their \c{config.c.*} (C compilation) and \c{config.cxx.*} (C++ compilation)
variants if we only want them applied during the respective language
compilation. For example:

\
$ bdep init ... cc       \
  config.cxx=clang++     \
  config.cc.coptions=-g  \
  config.cxx.coptions=-stdlib=libc++
\

|

One difference you might have noticed when creating the \c{gcc} and \c{clang}
configurations above is that the first one was designated as the default. The
default configuration is used by \c{bdep} commands if no configuration is
specified explicitly (see \l{bdep-projects-configs(1)} for details). It is
also the configuration that is used if we run the build system in the
project's source directory. So, normally, you would make your every day
development configuration the default. Let's try that:

\
$ bdep status
hello configured 0.1.0-a.0.19700101000000

$ b
c++ hello/cxx{hello}@../hello-gcc/hello/hello/
ld ../hello-gcc/hello/hello/exe{hello}
ln ../hello-gcc/hello/hello/exe{hello} -> hello/

$ b test
test hello/test{testscript} ../hello-gcc/hello/hello/exe{hello}

$ hello/hello World
Hello, World!
\

\N|To see the actual compilation command lines, run \c{b\ -v} and for even
more details, run \c{b\ -V}. See \l{b(1)} for more information on these
and other build system options.|

In contrast, the Clang configuration has to be requested explicitly:

\
$ bdep status @clang
hello configured 0.1.0-a.0.19700101000000

$ b ../hello-clang/hello/
c++ hello/cxx{hello}@../hello-clang/hello/hello/
ld ../hello-clang/hello/hello/exe{hello}

$ b test: ../hello-clang/hello/
test hello/test{testscript} ../hello-clang/hello/hello/exe{hello}

$ ../hello-clang/hello/hello/hello World
Hello, World!
\

As you can see, using the build system directly on configurations other than
the default requires explicitly specifying their paths. It would have been
more convenient if we could refer to them by names. The \l{bdep-update(1)} and
\l{bdep-test(1)} commands allow us to do exactly that:

\
$ bdep test @clang
c++ hello/cxx{hello}@../hello-clang/hello/hello/
ld ../hello-clang/hello/hello/exe{hello}
test hello/test{testscript} ../hello-clang/hello/hello/exe{hello}
\

And we can also perform the desired build system operation on several (or
\c{--all|-a}) configurations at once:

\
$ bdep test @gcc @clang
in configuration @gcc:
test hello/test{testscript} ../hello-gcc/hello/hello/exe{hello}

in configuration @clang:
test hello/test{testscript} ../hello-clang/hello/hello/exe{hello}
\

\N|As we will see later, the \l{bdep-test(1)} command also allows us to test
immediate (\c{--immediate|-i}) or all (\c{--recursive|-r}) dependencies of our
project. We call it \i{deep testing}.|

While we are here, let's also check how hard it would be to cross-compile:

\
$ bdep init -C ../hello-mingw @mingw cc config.cxx=x86_64-w64-mingw32-g++
initializing in project /tmp/hello/
created configuration @mingw /tmp/hello-mingw/ auto-synchronized
synchronizing:
  new hello/0.1.0-a.0.19700101000000

$ bdep update @mingw
c++ hello/cxx{hello}@../hello-mingw/hello/hello/
ld ../hello-mingw/hello/hello/exe{hello}
\

As you can see, cross-compiling in \c{build2} is nothing special. In our case,
on a properly setup GNU/Linux machine (that automatically uses \c{wine} as an
\c{.exe} interpreter) we can even run tests (in \c{build2} this is called
\i{cross-testing}):

\
$ bdep test @mingw
test hello/test{testscript} ../hello-mingw/hello/hello/exe{hello}

$ ../hello-mingw/hello/hello/hello.exe Windows
Hello, Windows!
\

Let's review what it takes to initialize a project's infrastructure and
perform the first build. For an existing project:

\
$ git clone .../hello.git
$ cd hello
$ bdep init -C ../hello-gcc @gcc cc config.cxx=g++
$ b
\

For a new project:

\
$ bdep new -t exe -l c++ hello
$ cd hello
$ bdep init -C ../hello-gcc @gcc cc config.cxx=g++
$ b
\

If you prefer, the \c{new} and \c{init} steps can be combined into a single
command:

\
$ bdep new -t exe -l c++ hello -C hello-gcc @gcc cc config.cxx=g++
\

And if you need to deinitialize a project in one or more build configurations,
there is the \l{bdep-deinit(1)} command for that:

\
$ bdep deinit @gcc @clang
deinitializing in project /tmp/hello/
in configuration @gcc:
synchronizing:
  drop hello

in configuration @clang:
synchronizing:
  drop hello
\

Now is also a good time to get an overview of the \c{build2} toolchain. After
all, we have already used two of its tools (\c{bdep} and \c{b}) without a
clear understanding of what they actually are.

Unlike most other programming languages that encapsulate the build system,
package dependency manager, and project dependency manager into a single tool
(such as Rust's \c{cargo} or Go's \c{go}), \c{build2} is a hierarchy of
several tools that you will be using directly and which together with your
version control system (VCS) will constitute the core of your project
management toolset.

\N|While \c{build2} can work without a VCS, this will result in reduced
functionality.|

At the bottom of the hierarchy is the build system, \l{b(1)}. Next comes the
package dependency manager, \l{bpkg(1)}. It is primarily used for \i{package
consumption} and depends on the build system. The top of the hierarchy is the
project dependency manager, \l{bdep(1)}. It is used for \i{project
development} and relies on \c{bpkg} for building project packages and their
dependencies.

\N|The main reason for this separation is modularity and the resulting
flexibility: there are situations where we only need the build system (for
example, when building a package for a system package manager where all the
dependencies should be satisfied from the system repository), or only the
build system and package manager (for example, when a build bot is building a
package for testing).

Note also that strictly speaking \c{build2} is not C/C++-specific; its build
model is general enough to handle any DAG-based operations and its
package/project dependency management can be used for any compiled language.|

\N|As we will see in a moment, \c{build2} also integrates with your VCS in
order to automate project versioning. Note that currently only \c{git(1)} is
supported.|

Now that we understand the tooling, let's also revisit the notion of \i{build
configuration} (those \c{hello-gcc} and \c{hello-clang} directories). A
\c{bdep} build configuration is actually a \c{bpkg} build configuration which,
in the build system terms, is an \i{amalgamation} \- a project that contains
\i{subprojects}. In our case, the subprojects in these amalgamations will be
the projects we have initialized with \c{init} and, as we will see later,
packages that they depend on. For example, here is what our \c{hello-gcc}
contains:

\
$ tree hello-gcc
hello-gcc/
├── .bpkg/
├── build/
│   └── config.build
└── hello/
    ├── build/
    │   └── config.build
    └── hello/
        ├── hello
        └── hello.o
\

\N|Underneath \l{bdep-init(1)} with the \c{--config-create|-C} option calls
\l{bpkg-cfg-create(1)} which, in turn, performs the build system \c{create}
meta-operation (see \l{b(1)} for details).

The important point here is that the \c{bdep} build configuration is not a
black box that you should never look inside of. On the contrary, it is a
normal and predictable concept of the package manager and the build system and
as long as you understand what you are doing, you should feel free to interact
with it directly.|

Let's now move on to the reason why there is \i{dep} in the \c{bdep} name:
dependency management.


\h#guide-repositories|Package Repositories|

Say we have realized that writing \i{\"Hello, World!\"} programs is a fairly
common task and that someone must have written a library to help with that. So
let's see if we can find something suitable to use in our project.

Where should we look? That's a good question. But before we can try to answer
it, we need to understand where \c{build2} can source dependencies. In
\c{build2} packages come from \i{package repositories}. Two commonly used
repository types are \i{version control} and \i{archive}-based (see
\l{bpkg-repository-types(1)} for details).

As the name suggests, a version control-based repository uses a VCS as its
distribution mechanism. \N{Currently, only \c{git} is supported.} Such a
repository normally contains multiple versions of a single package or,
perhaps, of a few related packages.

An archive-based repository contains multiple, potentially unrelated
packages/versions as archives along with some meta information (package list,
prerequisite/complement repositories, signatures, etc) that are all accessible
via HTTP(S).

Version control and archive-based repositories have different
trade-offs. Version control-based repositories are great for package
developers: With services like GitHub they are trivial to setup. In fact, your
project's (already existing) VCS repository will normally be the \c{build2}
package repository \- you might need to add a few files, but that's about it.

However, version control-based repositories are not without drawbacks: It will
be hard for your users to discover your packages (try searching for \"hello
library\" on GitHub \- most of the results are not even in C++ let alone
packaged for \c{build2}). There is also the issue of continuous availability:
users can delete their repositories, services may change their policies or go
out of business, and so on. Version control-based repositories also lack
repository authentication and package signing. Finally, obtaining the
available package list for such repositories can be slow.

A central, archive-based repository would address all these drawbacks: It
would be a single place to search for packages. Published packages will never
disappear and can be easily mirrored. Packages are signed and the repository
is authenticated (see \l{bpkg-repository-signing(1)} for details). And, last,
but not least, archive-based repositories are fast.

\l{https://cppget.org cppget.org} is the \c{build2} community's central
package repository. While centralized, it is also easy to mirror since its
contents are accessible via plain HTTPS (you can browse
\l{https://pkg.cppget.org pkg.cppget.org} to get an idea). As an added
benefit, packages on \l{https://cppget.org cppget.org} are continuously
\l{https://cppget.org/?builds built and tested} on all the major
platform/compiler combinations with the results available as part of the
package description.

\N|The main drawback of archive-based repositories is the setup cost. Getting
a basic repository going is relatively easy \- all you need is an HTTP(S)
server. Adding a repository web interface like that on \l{https://cppget.org
cppget.org} will require running \l{https://cppget.org/brep \c{brep}}. And
adding CI will require running a bunch of build bots
(\l{https://cppget.org/bbot \c{bbot}}). Note also that in \c{build2}
archive-based repositories can be federated with different sections of
the repository being hosted/managed potentially independently.|

\N|CI support for version control-based repositories is a work in progress.|

To summarize, version control-based repositories are great for package
developers while a central, archive-based repository is convenient for package
consumers. A reasonable strategy is then for package developers to publish
their releases to a central repository. Package consumers can then decide
which repository to use based on their needs. For example, one could use
\l{https://cppget.org cppget.org} as a (fast, reliable, and secure) source of
stable versions but also add, say, \c{git} repositories for select packages
(perhaps with the \c{#HEAD} fragment filter to improve download speed) for
testing development snapshots. In this model the two repository types
complement each other.

\N|Support for automated publishing of tagged releases to an archive-based
repository is a work in progress.|

Let's see how all this works in practice. Go over to \l{https://cppget.org
cppget.org} and type \"hello library\" in the search box. At the top of the
search result you should see the \l{https://cppget.org/libhello \c{libhello}}
package and if you follow the link you will see the package description page
along with a list of available versions. Pick a version that you like and you
will see the package version description page with quite a bit of information,
including the list of platform/compiler combinations that this version has
been successfully (or unsuccessfully) tested with. If you like what you see,
copy the \c{location} value \- this is the repository location where this
package version can be sourced from.

\N|The \l{https://cppget.org cppget.org} repository is split into several
sections: \c{stable}, \c{testing}, \c{beta}, \c{alpha} and \c{legacy}, with
each section having its own repository location (see the repository's
\l{https://cppget.org/?about about} page for details on each section's
policies). Note also that \c{testing} is complemented by \c{stable}, \c{beta}
by \c{testing}, and so on, so you only need to choose the lowest stability
level and you will automatically \"see\" packages from the more stable
sections.|

\N|The \l{https://cppget.org cppget.org} \c{stable} sections will always
contain the \c{libhello} library version \c{1.0.X} that was generated using
the following \l{bdep-new(1)} command line:

\
$ bdep new -t lib -l c++ libhello
\

It can be used as a predictable test dependency when setting up new projects.|

Let's say we've visited the \c{libhello} project's
\l{https://git.build2.org/cgit/hello/libhello/ home page} (for example by
following a link from the package details page) and noticed that it is being
developed in a \c{git} repository. How can we see what's available there? If
the releases are tagged, then we can infer the available released versions
from the tags. But that doesn't tell us anything about what's happening on the
\c{HEAD} or in the branches. For that we can use the package manager's
\l{bpkg-rep-info(1)} command:

\
$ bpkg rep-info https://git.build2.org/hello/libhello.git
libhello/1.0.0
libhello/1.1.0
\

As you can see, besides \c{1.0.0} that we have seen in \c{cppget.org/stable},
there is also \c{1.1.0} (which is perhaps being tested in
\c{cppget.org/testing}). We can also check what might be available on the
\c{HEAD} (see \l{bpkg-repository-types(1)} for details on the \c{git}
repository URL format):

\
$ bpkg rep-info https://git.build2.org/hello/libhello.git#HEAD
libhello/1.1.1-a.0.20180504111511.2e82f7378519
\

\N|We can also use the \c{rep-info} command on archive-based repositories,
however, if available, the web interface is usually more convenient and
provides more information.|

To summarize, we found two repositories for the \c{libhello} package: the
archive-based \l{https://cppget.org cppget.org} that contains the released
versions as well as its development \c{git} repository where we can get the
bleeding edge stuff. Let's now see how we can add \c{libhello} to our
project.


\h#guide-add-remove-deps|Adding and Removing Dependencies|

So we found \c{libhello} that we would like to use in our \c{hello}
project. First, we edit the \c{repositories.manifest} file found in the root
directory of our project and add one of the \c{libhello} repositories as a
prerequisite. Let's start with \l{https://cppget.org cppget.org}:

\
role: prerequisite
location: https://pkg.cppget.org/1/stable
\

\N|Refer to \l{bpkg#manifest-repository Repository Manifest} for details on
the repository manifest values.|

Next, we edit the \c{manifest} file (again, found in the root of our project)
and specify the dependency on \c{libhello} with optional version constraint.
For example:

\
depends: libhello ^1.0.0
\

Let's briefly discuss version constraints (for details see the
\l{bpkg#manifest-package-depends \c{depends}} value documentation). A version
constraint can be expressed with a comparison operator (\c{==}, \c{>},
\c{<}, \c{>=}, \c{<=}), a range shortcut operator (\c{~} and \c{^}), or a
range. Here are a few examples:

\
depends: libhello == 1.2.3
depends: libhello >= 1.2.3

depends: libhello ~1.2.3
depends: libhello ^1.2.3

depends: libhello [1.2.3 1.2.9)
\

You may already be familiar with the tilde (\c{~}) and caret (\c{^})
constraints from dependency managers for other languages. To recap, tilde
allows upgrades to any further patch versions while caret also allows upgrades
to further minor versions. They are equivalent to the following ranges:

\
~X.Y.Z  [X.Y.Z  X.Y+1.0)

^X.Y.Z  [X.Y.Z  X+1.0.0)  if X >  0
^0.Y.Z  [0.Y.Z  0.Y+1.0)  if X == 0
\

\N|Zero major version component is customarily used during early development
where the minor version effectively becomes major. As a result, the tilde
constraint has a special treatment of this case.|

Unless you have good reasons not to (for example, a dependency does not use
semantic versioning), we suggest that you use the \c{^} constraint which
provides a good balance between compatibility and upgradability with \c{~}
being a more conservative option.

Ok, we've specified where our package comes from (\c{repositories.manifest})
and which versions we find acceptable (\c{manifest}). The next step is to edit
\c{hello/buildfile} and import the \c{libhello} library into our build:

\
import libs += libhello%lib{hello}
\

Finally, we modify our source code to use the library:

\
#include 
...

int main (int argc, char* argv[])
{
  ...
  hello::say_hello (cout, argv[1]);
}
\

\N|You are probably wondering why we have to specify this repeating
information in so many places. Let's start with the source code: we can't
specify the version constraint or location there because it will have to be
repeated in every source file that uses the dependency.

Moving up, \c{buildfile} is also not a good place to specify this information
for the same reason (a library can be imported in multiple buildfiles) plus
the build system doesn't really know anything about version constraints or
repositories which is the purview of the dependency management tools.

Finally, we have to separate the version constraint and the location because
the same package can be present in multiple repositories with different
policies. For example, when a package from a version control-based repository
is published in an archive-based repository, its \c{repositories.manifest}
file is ignored and all its dependencies should be available from the
archive-based repository itself (or its fixed set of prerequisite
repositories). In other words, \c{manifest} belongs to a package while
\c{repositories.manifest} \- to a repository.

Also note that this is unlikely to become burdensome since adding new
dependencies is not something that happens often. There are also plans to
automate this with a \c{bdep-add(1)} command in the future.|

To summarize, these are the files we had to modify to add a dependency
to our project:

\
repositories.manifest   # add https://pkg.cppget.org/1/stable
manifest                # add 'depends: libhello ^1.0.0'
buildfile               # import libhello
hello.cxx               # use libhello
\

With a new dependency added, let's check the status of our project:

\
$ bdep status
fetching pkg:cppget.org/stable (prerequisite of dir:/tmp/hello)
warning: authenticity of the certificate for pkg:cppget.org/stable
         cannot be established
certificate is for cppget.org, \"Code Synthesis\" 
certificate SHA256 fingerprint:
86:BA:D4:DE:2C:87:1A:EE:38:<...>:5A:EA:F4:F7:8C:1D:63:30:C6
trust this certificate? [y/n] y

hello configured 0.1.0-a.0.19700101000000
      available  0.1.0-a.0.19700101000000#1
\

The \l{bdep-status(1)} command has detected that the dependency information
has changed and tells us that a new \i{iteration} of our project (that \c{#1})
is now available for \i{synchronization} with the build configuration.

We've also been prompted to authenticate the prerequisite repository. This
will have to happen once for every build configuration we initialize our
project in and can quickly become tedious. To overcome this, we can mention
the certificate fingerprint that we wish to automatically trust in the
\c{repositories.manifest} file (replace it with the actual fingerprint from
the repository's about page):

\
role: prerequisite
location: https://pkg.cppget.org/1/stable
trust: 86:BA:D4:DE:2C:87:1A:EE:38:<...>:5A:EA:F4:F7:8C:1D:63:30:C6
\

To synchronize a project with one or more build configurations we use the
\l{bdep-sync(1)} command:

\
$ bdep sync
synchronizing:
  new libhello/1.0.0 (required by hello)
  upgrade hello/0.1.0-a.0.19700101000000#1
\

Or we could just build the project without an explicit \c{sync} \- if
necessary, it will be automatically synchronized:

\
$ b
synchronizing:
  new libhello/1.0.0 (required by hello)
  upgrade hello/0.1.0-a.0.19700101000000#1
c++ ../hello-gcc/libhello-1.0.0/libhello/cxx{hello}
ld ../hello-gcc/libhello-1.0.0/libhello/libs{hello}
c++ hello/cxx{hello}@../hello-gcc/hello/hello/
ld ../hello-gcc/hello/hello/exe{hello}
ln ../hello-gcc/hello/hello/exe{hello} -> hello/
\

The synchronization as performed by the \c{sync} command is two-way:
dependency packages are first added, removed, upgraded, or downgraded in build
configurations according to the project's version constraints and user
input. Then the actual versions of the dependencies present in the build
configurations are recorded in the project's \c{lockfile} so that if desired,
the build can be reproduced exactly. \N{The \c{lockfile} functionality is not
yet implemented.} For a new dependency the latest available version that
satisfies the version constraint is used.

\N|Synchronization is also the last step in the \l{bdep-init(1)} command's
logic.|

Let's now examine the status in all (\c{--all|-a}) the build configurations
and include the immediate dependencies (\c{--immediate|-i}):

\
$ bdep status -ai
in configuration @gcc:
hello configured 0.1.0-a.0.19700101000000#1
  libhello ^1.0.0 configured 1.0.0

in configuration @clang:
hello configured 0.1.0-a.0.19700101000000
      available 0.1.0-a.0.19700101000000#1
\

Since we didn't specify a configuration explicitly, only the default (\c{gcc})
was synchronized. Normally, you would try a new dependency in one
configuration, make sure everything looks good, then synchronize the rest with
\c{--all|-a} (or, again, just build what you need directly). Here are a few
examples (see \l{bdep-projects-configs(1)} for details):

\
$ bdep sync -a
$ bdep sync @gcc @clang
$ bdep sync -c ../hello-mingw
\

After adding a new (or upgrading/downgrading existing) dependency, it's a good
idea to \i{deep-test} our project: run not only our own tests but also of its
immediate (\c{--immediate|-i}) or even all (\c{--recursive|-r}) dependencies.
For example:

\
$ bdep test -ai
in configuration @gcc:
test hello/test{testscript} ../hello-gcc/hello/hello/exe{hello}
test ../hello-gcc/libhello-1.0.0/tests/basics/exe{driver}

in configuration @clang:
test hello/test{testscript} ../hello-clang/hello/hello/exe{hello}
test ../hello-clang/libhello-1.0.0/tests/basics/exe{driver}
\

To get rid of a dependency, we simply remove it from the \c{manifest} file
and synchronize the project. For example, assuming \c{libhello} is no longer
mentioned as a dependency in our \c{manifests}:

\
$ bdep status
hello configured 0.1.0-a.0.19700101000000#1
      available  0.1.0-a.0.19700101000000#2

$ bdep sync
synchronizing:
  drop libhello/1.0.0 (unused)
  upgrade hello/0.1.0-a.0.19700101000000#2
\

\N|If instead of building a dependency from source you would prefer to use a
version that is installed by your system package manager, see
\l{#guide-system-deps Using System-Installed Dependencies}. And for
information on using dependencies that are not \c{build2} packages refer to
\l{#guide-unpackaged-deps Using Unpackaged Dependencies}.|


\h#guide-upgrade-downgrade-deps|Upgrading and Downgrading Dependencies|

Let's say we would like to try that \c{1.1.0} version we have seen in
the \c{libhello} \c{git} repository. First, we need to add the
repository to the \c{repositories.manifest} file:

\
role: prerequisite
location: https://git.build2.org/hello/libhello.git
\

\N|Note that we don't need the \c{trust} value since \c{git} repositories
are not authenticated.|

To refresh the list of available dependency versions we use the
\l{bdep-fetch(1)} command (or the \c{--fetch|-f} option to \c{status}):

\
$ bdep fetch
$ bdep status libhello
libhello configured 1.0.0 available [1.1.0]
\

To upgrade (or downgrade) dependencies we again use the \l{bdep-sync(1)}
command. We can upgrade one or more specific dependencies by listing them
as arguments to \c{sync}:

\
$ bdep sync libhello
synchronizing:
  new libformat/1.0.0 (required by libhello)
  new libprint/1.0.0 (required by libhello)
  upgrade libhello/1.1.0
  upgrade hello/0.1.0-a.0.19700101000000#3
\

Without an explicit version or the \c{--patch|-p} option, \c{sync} will
upgrade the specified dependencies to the latest available versions. For
example, if we don't like version \c{1.1.0}, we can downgrade it back to
\c{1.0.0} by specifying the version explicitly (we pass \c{--old-available|-o}
to \c{status} to see the old versions):

\
$ bdep status -o libhello
libhello configured 1.1.0 available (1.1.0) [1.0.0]

$ bdep sync libhello/1.0.0
synchronizing:
  drop libprint/1.0.0 (unused)
  drop libformat/1.0.0 (unused)
  downgrade libhello/1.0.0
  reconfigure hello/0.1.0-a.0.19700101000000#3
\

\N|The available versions are listed in the descending order with \c{[]}
indicating that the version is only available as a dependency and \c{()}
marking the current version.|

Instead of specific dependencies we can also upgrade (\c{--upgrade|-u}) or
patch (\c{--patch|-p}) immediate (\c{--immediate|-i}) or all
(\c{--recursive|-r}) dependencies of our project.

As a more realistic example, version \c{1.1.0} of \c{libhello} depends on two
other libraries: \c{libformat} and \c{libprint}. Here is our project's
dependency tree while we were still using that version:

\
$ bdep status -r
hello configured 0.1.0-a.0.19700101000000#3
  libhello ^1.0.0 configured 1.1.0
    libformat ^1.0.0 configured 1.0.0
    libprint ^1.0.0 configured 1.0.0
\

A typical conservative dependency management workflow would look like this:

\
$ bdep status -fi  # refresh and examine immediate dependencies
hello configured 0.1.0-a.0.19700101000000#3
  libhello configured 1.1.0 available [2.0.0] [1.2.0] [1.1.2] [1.1.1]

$ bdep sync -pi    # upgrade immediate to latest patch version
synchronizing:
  upgrade libhello/1.1.2
  reconfigure hello/0.1.0-a.0.19700101000000#3
continue? [Y/n] y
\

Notice that in case of such mass upgrades you are prompted for confirmation
before anything is actually changed (unless you pass \c{--yes|-y}).

In contrast, the following would be a fairly aggressive workflow where we
upgrade everything to the latest available version (version constraints
permitting; here we assume \c{^1.0.0} was used for all the dependencies):

\
$ bdep status -fr  # refresh and examine all dependencies
hello configured 0.1.0-a.0.19700101000000#3
  libhello configured 1.1.0 available [2.0.0] [1.2.0] [1.1.1]
    libprint configured 1.0.0 available [2.0.0] [1.1.0] [1.0.1]
    libformat configured 1.0.0 available [2.0.0] [1.1.0] [1.0.1]

$ bdep sync -ur    # upgrade all to latest available version
synchronizing:
  upgrade libprint/1.1.0
  upgrade libformat/1.1.0
  upgrade libhello/1.2.0
  reconfigure hello/0.1.0-a.0.19700101000000#3
continue? [Y/n] y
\

We can also have something in between: patch all (\c{sync\ -pr}), upgrade
immediate (\c{sync\ -ui}), or even upgrade immediate and patch the rest
(\c{sync\ -ui} followed by \c{sync\ -pr}).



\h#guide-versioning-releasing|Versioning and Release Management|

Let's now discuss versioning and release management and, yes, that
strange-looking \c{0.1.0-a.0.19700101000000} we keep seeing. While a build
system project doesn't need a version and a \c{bpkg} package can use custom
versioning schemes (see \l{bpkg#package-version Package Version}), a project
managed by \c{bdep} must use \i{standard versioning}. \N{A dependency, which
is a \c{bpkg} package, need not use standard versioning.}

Standard versioning (\i{stdver}) is a \l{https://semver.org semantic
versioning} (\i{semver}) scheme with a more precisely defined pre-release
component and without any build metadata.

\N|If you believe that \i{semver} is just \c{\i{major}.\i{minor}.\i{patch}},
then in your worldview \i{stdver} would be the same as \i{semver}. In reality,
\i{semver} also allows loosely defined pre-release and build metadata
components. For example, \c{1.2.3-beta.1+build.23456} is a valid \i{semver}.|

A standard version has the following form:

\c{\i{major}\b{.}\i{minor}\b{.}\i{patch}[\b{-}\i{prerel}]}

The \ci{major}, \ci{minor}, and \ci{patch} components have the same meaning as
in \i{semver}. The \ci{prerel} component is used to provide \i{continuous
versioning} of our project between releases. Specifically, during development
of a new version we may want to publish several pre-releases, for example,
alpha or beta. In between those we may also want to publish a number of
snapshots, for example, for CI. With continuous versioning all these releases,
pre-releases, and snapshots are assigned unique, properly ordered versions.

\N|Continuous versioning is a cornerstone of the \c{build2} project dependency
management. In case of snapshots, an appropriate version is assigned
automatically in cooperation with your VCS.|

The \ci{prerel} component for a pre-release has the following form:

\c{(\b{a}|\b{b})\b{.}\i{num}}

Here \cb{a} stands for alpha, \cb{b} stands for beta, and \ci{num} is the
alpha/beta number. For example:

\
1.1.0        # final              release  for 1.1.0
1.2.0-a.1    # first  alpha   pre-release  for 1.2.0
1.2.0-a.2    # second alpha   pre-release  for 1.2.0
1.2.0-b.1    # first  beta    pre-release  for 1.2.0
1.2.0        # final              release  for 1.2.0
\

The \ci{prerel} component for a snapshot has the following form:

\c{(\b{a}|\b{b})\b{.}\i{num}\b{.}\i{snapsn}[\b{.}\i{snapid}]}

Where \ci{snapsn} is the snapshot sequence number and \ci{snapid} is
the snapshot id. In case of \c{git}, \ci{snapsn} is the commit timestamp
in the \c{YYYYMMDDhhmmss} form and UTC timezone while \ci{snapid} is
a 12-character abbreviated commit id. For example:

\
1.2.3-a.1.20180319215815.26efe301f4a7
\

Notice also that a snapshot version is ordered \i{after} the corresponding
pre-release version. That is, \c{1.2.3-a.1\ <\ 1.2.3-a.1.1}. As a result, it
is customary to start the development of a new version with \c{X.Y.Z-a.0.z},
that is, a snapshot after the (non-existent) zero'th alpha release. \N{We will
explain the meaning of \cb{z} in this version momentarily.} The following
chronologically-ordered versions illustrate a typical release flow of a
project that uses \c{git} as its VCS:

\
0.1.0-a.0.19700101000000               # snapshot (no commits yet)
0.1.0-a.0.20180319215815.26efe301f4a7  # snapshot (first commit)
...                                    # more commits/snapshots
0.1.0-a.1                              # pre-release (first alpha)
0.1.0-a.1.20180319221826.a6f0f41205b8  # snapshot
...                                    # more commits/snapshots
0.1.0-a.2                              # pre-release (second alpha)
0.1.0-a.2.20180319231937.b701052316c9  # snapshot
...                                    # more commits/snapshots
0.1.0-b.1                              # pre-release (first beta)
0.1.0-b.1.20180319242038.c812163417da  # snapshot
...                                    # more commits/snapshots
0.1.0                                  # release
0.2.0-a.0.20180319252139.d923274528eb  # snapshot (first in 0.2.0)
...
\

For a more detailed discussion of standard versioning and its support in
\c{build2} refer to \l{b#module-version Version Module}.

Let's now see how this works in practice by publishing a couple of versions
for our \c{hello} project. By now it should be clear what that
\c{0.1.0-a.0.19700101000000} means \- it is the first snapshot version of our
project. Since there are no commits yet, it has the UNIX epoch as its commit
timestamp. As the first step, let's try to commit our project and see what
changes:

\
$ git add .
$ git commit -m \"Start hello project\"

$ bdep status
hello configured 0.1.0-a.0.19700101000000
      available  0.1.0-a.0.20180507062614.ee006880fc7e
\

Just like with changes to dependency information, \c{status} has detected that
a new (snapshot) version of our project is available for synchronization.

\N|Another way to view the project's version (which works even if we are
not using \c{bdep}) is with the build system's \c{info} operation:

\
$ b info
project: hello
version: 0.1.0-a.0.20180507062614.ee006880fc7e
summary: hello executable project
...
\

|

Let's synchronize with the default build configuration:

\
$ bdep sync
synchronizing:
  upgrade hello/0.1.0-a.0.20180507062614.ee006880fc7e

$ bdep status
hello configured 0.1.0-a.0.20180507062614.ee006880fc7e
\

\N|Notice that we didn't have to manually change the version anywhere. All we
had to do was commit our changes and a new snapshot version was automatically
derived by \c{build2} from the new \c{git} commit. Without this automation
continuous versioning would hardly be practical.|

If we now make another commit, we will see a similar picture:

\
$ bdep status
hello configured 0.1.0-a.0.20180507062614.ee006880fc7e
      available  0.1.0-a.0.20180507062615.8fb9de05b38f
\

\N|Note that you don't need to manually run \c{sync} after every commit. As
discussed earlier, you can simply run the build system to update your project
and things will get automatically synchronized if necessary.|

Ok, time for our first release. Let's start with \c{0.1.0-a.1}. Unlike
snapshots, for pre-releases as well as final releases we have to update the
version in the \c{manifest} file manually:

\
version: 0.1.0-a.1
\

\N|The \c{manifest} file is the singular place where we specify the package
version. The build system's \l{b#module-version \c{version} module} makes it
available in various forms in buildfiles and even source code.|

To ensure continuous versioning, this change to version must be the last commit
for this (pre-)release which itself must be immediately followed by a second
change to the version starting the development of the next (pre-)release. We
also recommend that you tag the release commit with a tag name in the
\c{\b{v}\i{X}.\i{Y}.\i{Z}} form.

\N|Having regular release tag names with the \cb{v} prefix allows one to
distinguish them from other tags, for example, with wildcard patterns.|

Here is the release workflow for our example:

\
$ git commit -a -m \"Release version 0.1.0-a.1\"
$ git tag -a v0.1.0-a.1 -m \"Tag version 0.1.0-a.1\"
$ git push --follow-tags

# Version 0.1.0-a.1 is now public.

$ edit manifest  # change 'version: 0.1.0-a.1.z'
$ git commit -a -m \"Change version to 0.1.0-a.1.z\"
$ git push

# Master is now open for business.
\

\N|In the future release management will be automated with a
\c{bdep-release(1)} command.|

Notice also that when specifying a snapshot version in \c{manifest} we use the
special \cb{z} snapshot value (for example, \c{0.1.0-a.1.z}) which is
recognized and automatically replaced by \c{build2} with, in case of \c{git},
a commit timestamp and id (refer to \l{b#module-version Version Module} for
details).

Publishing the final release is exactly the same. For completeness, here
are the commands:

\
$ edit manifest  # change 'version: 0.1.0'
$ git commit -a -m \"Release version 0.1.0\"
$ git tag -a v0.1.0 -m \"Tag version 0.1.0\"
$ git push --follow-tags

$ edit manifest  # change 'version: 0.2.0-a.0.z'
$ git commit -a -m \"Change version to 0.2.0-a.0.z\"
$ git push
\

\N|One sticky point of continuous versioning is choosing the next version.
For example, above should we continue with \c{0.1.1-a.0}, \c{0.2.0-a.0},
or \c{1.0.0-a.0}? The important rule to keep in mind is that we can jump
forward to any further version at any time and without breaking continuous
versioning. But we can never jump backwards.

For example, we can start with \c{0.2.0-a.0} but if we later realize that this
will actually be a new major release, we can easily change it to
\c{1.0.0-a.0}. As a result, the general recommendation is to start
conservatively by either incrementing the patch or the minor version
component. The recommended strategy is to increment the minor component and,
if required, release patch versions from a separate branch (created by
branching off from the release commit).

Note also that you don't have to make any pre-releases if you don't need them.
While during development you would still keep the version as \c{X.Y.Z-a.0}, at
release you simply change it directly to the final \c{X.Y.Z}.|

When publishing the final release you may also want to clean up now
obsolete pre-release tags. For example:

\
$ git tag -l 'v0.1.0-*' | xargs git push --delete origin
$ git tag -l 'v0.1.0-*' | xargs git tag --delete
\

\N|While at first removing such tags may seem like a bad idea, pre-releases
are by nature temporary and their use only makes sense until the final release
is published.

Also note that having a \c{git} repository with a large number of published
but unused version tags may result in a significant download overhead.|

Let's also briefly discuss in which situations we should increment each of the
version components. While \i{semver} gives basic guidelines, there are several
ways to apply them in the context of C/C++ where there is a distinction
between binary and source compatibility. We recommend that you reserve
\i{patch} releases for specific bug fixes and security issues that you can
guarantee with a high level of certainty to be binary-compatible. Otherwise,
if the changes are source-compatible, increment \i{minor}. And if they are
breaking (that is, the user code likely will need adjustments), increment
\i{major}. During early development, when breaking changes are frequent, it is
customary to use the \c{0.Y.Z} versions where \c{Y} effectively becomes the
\i{major} component. Again, refer to the \l{b#module-version Version Module}
for a more detailed discussion of this topic.


\h#guide-dev-multi|Developing Multiple Packages and Projects|

How does a library like \c{libhello} get developed? It's possible someone woke
up one day and realized that they were going to build a useful library that
everyone was going to use. But somehow this doesn't feel like how it really
works. In the real world things start organically: someone had a project like
\c{hello} and then needed the same functionality in another project. Or
someone else needed it and asked the author to factor it out into a
library. For this approach to work, however, moving such common functionality
into a library and then continue its parallel development must be a simple,
frictionless process. Let's see how this works in \c{build2}.

First, we need to decide whether to make \c{libhello} another package in our
\c{hello} project (that is, in the same \c{git} repository) or a separate
project (with a separate repository). Both arrangements are equally well
supported.

\N|A multi-package project works best if all the packages have the same
version and are released together. While the packages themselves can have
different versions (since each has its own \c{manifest}), in this scenario
following the release tagging recommendations discussed earlier will be
problematic.|

Let's start with a separate project since it is simpler. As the first step we
use \l{bdep-new(1)} to create a new library project next to our \c{hello}:

\
$ bdep new -t lib -l c++ libhello
created new library project libhello in /tmp/libhello/

$ ls
hello/
libhello/
hello-gcc/
hello-clang/
\

Let's also edit the generated \c{manifest} file and add the \c{project} value
(customarily after \c{version}) to indicate that our library belongs to the
same overall project as our executable:

\
$ cat libhello/manifest
: 1
name: libhello
version: 0.1.0-a.0.z
project: hello
summary: hello library
...
\

\N|The \c{project} value is used to group related packages together in order
to help with their organization and discovery. For example, if later we create
\c{libhello2} or \c{libhello-extra}, then it would make sense for them to also
belong to the \c{hello} project. See the \l{bpkg#manifest-package-project
\c{project}} value documentation for details.|

Our two projects will be sharing the same set of build configurations, so
next we initialize \c{libhello} in \c{hello-gcc} and \c{hello-clang}:

\
$ cd libhello

$ bdep init -A ../hello-gcc @gcc
initializing in project /tmp/libhello/
added configuration @gcc /tmp/hello-gcc/ default,auto-synchronized
synchronizing:
  new libhello/0.1.0-a.0.19700101000000

$ bdep init -A ../hello-clang @clang
initializing in project /tmp/libhello/
added configuration @clang /tmp/hello-clang/ auto-synchronized
synchronizing:
  new libhello/0.1.0-a.0.19700101000000
\

\N|If two or more projects share the same build configuration, then all of
them are always synchronized at once, regardless of the originating project.
It also makes sense to have the same default configuration and use identical
configuration names in all the projects.|

The last step is to move the desired functionality from \c{hello} to
\c{libhello} and at the same time add a dependency on \c{libhello}, just as we
did earlier (add a \c{depends} entry to \c{manifest}, then import the library
in \c{buildfile}, and so on). One interesting question is what to put as a
prerequisite repository in \c{repositories.manifest}. Our own setup will work
even if we don't put anything there \- the dependency will be automatically
resolved to our local version of \c{libhello} since we have initialized it in
all our build configurations. However, in case our \c{hello} repository is
used by someone else, it's a good idea to add the remote \c{git} repository
for \c{libhello} as a prerequisite.

\N|By now you have probably realized that our project directory is just
another type of package repository. See \l{bpkg-repository-types(1)} for
more information.|

And that's it, now we can build and test our new arrangement:

\
$ cd ../hello   # back to hello project root
$ bdep test -i
c++ ../libhello/libhello/cxx{hello}
c++ ../libhello/tests/basics/cxx{driver}
c++ hello/cxx{hello}
ld ../hello-gcc/libhello/libhello/libs{hello}
ld ../hello-gcc/libhello/tests/basics/exe{driver}
ld ../hello-gcc/hello/hello/exe{hello}
test ../hello-gcc/libhello/tests/basics/exe{driver}
test hello/test{testscript} ../hello-gcc/hello/hello/exe{hello}
\

This is also the approach we would use if we wanted to fix a bug in someone
else's library. That is, we would clone their project repository and
initialize it in the build configurations of our project which will
\"upgrade\" the dependency to use the local version. Then we make the fix,
submit it upstream, and continue using the local version until our fix is
merged/published, at which point we deinitialize the project and switch
back to using the upstream version.

Let's now examine the second option: making \c{libhello} a package inside
\c{hello}. Here is the original structure of our \c{hello} project:

\
hello/
├── .git/
├── build/
├── hello/
│   ├── hello.cxx
│   └── buildfile
├── buildfile
├── manifest
└── repositories.manifest
\

As the first step, we move the \c{hello} program into its own subdirectory:

\
hello/
├── .git/
├── hello/
│   ├── build/
│   ├── hello/
│   │   ├── hello.cxx
│   │   └── buildfile
│   ├── buildfile
│   └── manifest
└── repositories.manifest
\

Next we again use \l{bdep-new(1)} to create a new library but this time
as a package inside an already existing project:

\
$ cd hello
$ bdep new --package -t lib -l c++ libhello
created new library package libhello in /tmp/hello/libhello/
\

Let's see what our project looks like now:

\
hello/
├── .git/
├── hello/
│   ├── ...
│   └── manifest
├── libhello/
│   ├── ...
│   └── manifest
├── packages.manifest
└── repositories.manifest
\

\N|Notice that, as discussed earlier, \c{repositories.manifest} belongs to
the project (repository) while \c{manifest} \- to the package.|

Besides the \c{libhello} directory the \c{new} command also created the
\c{packages.manifest} file in the root directory of our project. Let's take a
look inside:

\
$ cat packages.manifest
: 1
location: libhello/
\

Up until now our \c{hello} was a simple, single-package project that didn't
need this file \- \c{manifest} in its root directory was sufficient (see
\l{bpkg-repository-types(1)} for details on the project repository
structure). But now it contains several packages and we need to specify where
they are located within the project. So let's go ahead and add the location
of the \c{hello} package:

\
$ cat packages.manifest
: 1
location: libhello/
:
location: hello/
\

\N|Packages in a project can reside next to each other or in subdirectories
but they cannot nest. When published to an archive-based repository, each
such package will be placed into its own archive.|

Next we initialize the new package in all our build configurations:

\
$ cd libhello
$ bdep init -a
initializing in project /tmp/hello/
in configuration @gcc:
synchronizing:
  upgrade hello/0.1.0-a.0.19700101000000#1
  new libhello/0.1.0-a.0.19700101000000

in configuration @clang:
synchronizing:
  upgrade hello/0.1.0-a.0.19700101000000#1
  new libhello/0.1.0-a.0.19700101000000
\

\N|Notice that the \c{hello} package has been \"upgraded\" to reflect its
new location.|

Finally, as before, we move the desired functionality from \c{hello} to
\c{libhello} and at the same time add a dependency on \c{libhello}. Note,
however, that in this case we don't need to add anything to
\c{repositories.manifest} since both packages are in the same project
(repository). And that's it, now we can build and test our new arrangement:

\
$ cd ..      # back to hello project root
$ bdep test
c++ libhello/libhello/cxx{hello}
c++ libhello/tests/basics/cxx{driver}
c++ hello/hello/cxx{hello}
ld ../hello-gcc/libhello/libhello/libs{hello}
ld ../hello-gcc/libhello/tests/basics/exe{driver}
ld ../hello-gcc/hello/hello/exe{hello}
test ../hello-gcc/libhello/tests/basics/exe{driver}
test hello/hello/test{testscript} ../hello-gcc/hello/hello/exe{hello}
\


\h#guide-consume-pkg|Package Consumption|

Ok, now that we have published a few releases of \c{hello}, how would the
users of our project get them? While they could clone the repository and use
\c{bdep} just like we did, this is more of a development rather than
consumption workflow. For consumption it is much easier to use the package
dependency manager, \l{bpkg(1)}, directly.

\N|Note that this approach also works for libraries in case you wish to use
them in a project with a build system other than \c{build2}. See
\l{#guide-unpackaged-deps Using Unpackaged Dependencies} for background on
cross-build system library consumption.|

First, we create a suitable build configuration with the
\l{bpkg-cfg-create(1)} command. We can use the same place for building all our
tools so let's call the directory \c{tools}. Seeing that we are only
interested in using (rather than developing) such tools, let's build them
optimized and also configure a suitable installation location:

\
$ bpkg create -d tools cc        \
  config.cxx=g++                 \
  config.cc.coptions=-O3         \
  config.install.root=/usr/local \
  config.install.sudo=sudo
created new configuration in /tmp/tools/

$ cd tools
\

The same step on Windows using Visual Studio would look like this (again,
remember to run this from the Visual Studio development command prompt):

\
$ bpkg create -d tools cc ^
  config.cxx=cl           ^
  config.cc.coptions=/O2  ^
  config.install.root= C:\install
\

To fetch and build packages (as well as all their dependencies) we use the
\l{bpkg-pkg-build(1)} command. We can use either an archive-based repository
like \l{https://cppget.org cppget.org} or build directly from \c{git}:

\
$ bpkg build hello@https://git.build2.org/hello/hello.git
fetching from https://git.build2.org/hello/hello.git
  new libformat/1.0.0 (required by libhello)
  new libprint/1.0.0 (required by libhello)
  new libhello/1.1.0 (required by hello)
  new hello/1.0.0
continue? [Y/n] y
configured libformat/1.0.0
configured libprint/1.0.0
configured libhello/1.1.0
configured hello/1.0.0
c++ libprint-1.0.0/libprint/cxx{print}
c++ hello-1.0.0/hello/cxx{hello}
c++ libhello-1.1.0/libhello/cxx{hello}
c++ libformat-1.0.0/libformat/cxx{format}
ld libprint-1.0.0/libprint/libs{print}
ld libformat-1.0.0/libformat/libs{format}
ld libhello-1.1.0/libhello/libs{hello}
ld hello-1.0.0/hello/exe{hello}
updated hello/1.0.0
\

\N|Passing a repository URL to the \c{build} command is a shortcut to the
following sequence of commands:

\
$ bpkg add https://git.build2.org/hello/hello.git  # add repository
$ bpkg fetch                             # fetch package list
$ bpkg build hello                       # build package by name
\

|

Once built, we can install the package to the location that we have specified
with \c{config.install.root} using the \l{bpkg-pkg-install(1)} command:

\
$ bpkg install hello
...
install libformat-1.0.0/libformat/libs{format}
install libprint-1.0.0/libprint/libs{print}
install libhello-1.1.0/libhello/libs{hello}
install hello-1.0.0/hello/exe{hello}

$ hello World
Hello, World!
\

\N|If on your system the installed executables don't run from \c{/usr/local}
because of the unresolved shared libraries (or if you are installing somewhere
else, such as \c{/opt}), then the easiest way to fix this is with \i{rpath}.
Simply add the following configuration variable when creating the build
configuration (or as an argument to the \c{install} command):

\
config.bin.rpath=/usr/local/lib
\

Note to Windows users: this is not an issue on this platform since executables
and shared (DLL) libraries are installed into the same subdirectory (\c{bin})
of the installation directory.|

The installation contents and layout under \c{config.install.root} would be
along these lines:

\
/usr/local/
├── bin/
│   └── hello
├── include/
│   ├── libformat/
│   │   ├── export.hxx
│   │   ├── format.hxx
│   │   └── version.hxx
│   ├── libhello/
│   │   ├── export.hxx
│   │   ├── hello.hxx
│   │   └── version.hxx
│   └── libprint/
│       ├── export.hxx
│       ├── print.hxx
│       └── version.hxx
├── lib/
│   ├── libformat-1.0.so
│   ├── libformat.so -> libformat-1.0.so
│   ├── libhello-1.1.so
│   ├── libhello.so -> libhello-1.1.so
│   ├── libprint-1.0.so
│   ├── libprint.so -> libprint-1.0.so
│   └── pkgconfig
│       ├── libformat.shared.pc
│       ├── libhello.shared.pc
│       └── libprint.shared.pc
└── share/
    └── doc/
        ├── libformat/
        │   └── manifest
        ├── libhello/
        │   └── manifest
        └── libprint/
            └── manifest
\

\N|The installation locations of various types of files (executables,
libraries, headers, documentation, etc.) can be customized using a number of
the \c{config.install.*} variables with the most commonly used ones and their
defaults (relative to \c{config.install.root}) listed below (see the
\c{install} build system module documentation for the complete list).

\
config.install.bin     = root/bin/
config.install.lib     = root/lib/
config.install.doc     = root/share/doc/
config.install.man     = root/share/man/
config.install.include = root/include/
\

|


If we need to uninstall a previously installed package, there is the
\l{bpkg-pkg-uninstall(1)} command:

\
$ bpkg uninstall hello
uninstall hello-1.0.0/hello/exe{hello}
uninstall libhello-1.1.0/libhello/libs{hello}
uninstall libprint-1.0.0/libprint/libs{print}
uninstall libformat-1.0.0/libformat/libs{format}
...
\

To upgrade or downgrade packages we again use the \c{build} command. Here
is a typical upgrade workflow:

\
$ bpkg fetch              # refresh available package list
$ bpkg status             # see if new versions are available

$ bpkg uninstall hello    # uninstall old version
$ bpkg build     hello    # upgrade to the latest version
$ bpkg install   hello    # install new version
\

Similar to \c{bdep}, to downgrade we have to specify the desired version
explicitly. There are also the \c{--upgrade|-u} and \c{--patch|-p} as well as
\c{--immediate|-i} and \c{--recursive|-r} options that allow us to upgrade or
patch packages that we have built and/or their immediate or all dependencies
(see \l{bpkg-pkg-build(1)} for details). For example, to make sure everything
is patched, run:

\
$ bpkg fetch
$ bpkg build -pr
\

If a package is no longer needed, we can remove it from the configuration with
\l{bpkg-pkg-drop(1)}:

\
$ bpkg drop hello
following dependencies were automatically built but
will no longer be used:
  libhello
  libformat
  libprint
drop unused packages? [Y/n] y
  drop hello
  drop libhello
  drop libformat
  drop libprint
continue? [Y/n] y
purged hello
purged libhello
purged libformat
purged libprint
\

\h#guide-system-deps|Using System-Installed Dependencies|

Our operating system might already have a package manager (which we will refer
to as \i{system package manager}) and for various reasons we may want to use
the system-installed version of a dependency rather than building one from
source.

\N|Using system-installed versions works best for mature rather than
rapidly-developed packages since for the latter you often need to track the
latest version (which may not yet be available from the system repository)
and/or test with multiple versions (which is not something that many system
package managers support).|

We can instruct \c{build2} to configure a dependency package as available from
the system rather than building it from source. Let's see how this works in an
example. Say, we want to use \l{https://cppget.org/libsqlite3 \c{libsqlite3}}
in our \c{hello} project.

The first step is to add it as a dependency, just like we did for \c{libhello}.
That is, add another \c{depends} entry to \c{manifest}, then import it in
\c{buildfile}, and so on.

\N|Note that the dependency still has to be packaged and available from one of
the project's prerequisite repositories. However, it can be a \i{stub} \- a
package that does not contain any source code and that can only be \"obtained\"
from the system (see \l{bpkg#package-version Package Version} for
details). See also \l{#guide-unpackaged-deps Using Unpackaged Dependencies}
for how to deal with dependencies that are not packaged.|

Now, if we just run \c{sync} or try to build our project, \c{build2} will
download and build the new dependency from source, just like it did for
\c{libhello}. Instead, we can issue an explicit \c{sync} command that
configures the \c{libsqlite3} package as coming from the system:

\
$ bdep sync ?sys:libsqlite3
synchronizing:
  configure sys:libsqlite3/*
  upgrade hello/0.1.0-a.0.19700101000000#3
\

Here \cb{?} is a package \i{flag} that instructs \c{build2} to treat it as a
dependency and \cb{sys} is a package \i{scheme} that tells \c{build2} it comes
from the system. See \l{bpkg-pkg-build(1)} for details.

\N|We can have some build configurations using a system-installed version of
a dependency while others building it from source, for example, for testing.|

\N|The system-installed dependency doesn't really have to come from the system
package manager. It can also be manually installed and, as discussed in
\l{#guide-unpackaged-deps Using Unpackaged Dependencies}, not necessarily into
the system-default location like \c{/usr/local}.|

\N|Currently, unless we specify the installed version explicitly, a
system-installed package is assumed to satisfy any dependency constraint. In
the future, \c{build2} will automatically query commonly used system package
managers for the installed version and maybe even request installation of the
absent packages. To support this functionality, the package manifest may need
to specify package name mappings for various system package managers (which is
the rationale behind stub packages).|


\h#guide-unpackaged-deps|Using Unpackaged Dependencies|

Generally, we will have a much better time if all our dependencies come as
\c{build2} packages. Unfortunately, this won't always be the case in the real
world and some libraries that you may need will use other build systems.

\N|There is also the opposite problem: you may want to consume a library that
uses \c{build2} in a project that uses a different build system. For that
refer to \l{#guide-consume-pkg Package Consumption}.|

The standard way to consume such unpackaged libraries is to install them (not
necessarily into a system-default location like \c{/usr/local}) so that we
have a single directory with their headers and a single directory with their
libraries. We can then configure our builds to use these directories when
searching for imported libraries.

\N|Needless to say, none of the \c{build2} dependency management mechanisms
such as version constraints or upgrade/downgrade will work on such unpackaged
libraries. You will have to manage all these yourself manually.|

Let's see how this all works in an example. Say, we want to use \c{libextra}
that uses a different build system in our \c{hello} project. The first step
is to manually build and install this library for each build configuration
that we have. For example, we can install all such unpackaged libraries into
\c{unpkg-gcc} and \c{unpkg-clang}, next to our \c{hello-gcc} and
\c{hello-clang} build configurations:

\
$ ls
hello/
hello-gcc/
unpkg-gcc/
hello-clang/
unpkg-clang/
\

\N|If you would like to try this out but don't have a suitable \c{libextra},
you can create and install one with these commands:

\
$ bdep new -t lib -l c++ libextra -C libextra-gcc cc config.cxx=g++
$ b install: libextra-gcc/ config.install.root=/tmp/unpkg-gcc
\

|

If we look inside one of these \c{unpkg-*} directories, we should see
something like this:

\
$ tree unpkg-gcc
unpkg-gcc
├── include
│   └── libextra
│       └── extra.hxx
└── lib
    ├── libextra.a
    ├── libextra.so
    └── pkgconfig
        └── libextra.pc
\

Notice that \c{libextra.pc} \- it's a \cb{pkg-config(1)} file that contains
any extra compile and link options that may be necessary to consume this
library. This is the \i{de facto} standard for build systems to communicate
library build information to each other and is today supported by most
commonly used implementations. Speaking of \c{build2}, it both recognizes
\c{.pc} files when consuming third-party libraries and automatically produces
them when installing its own.

\N|While this may all seem foreign to Windows users, there is nothing
platform-specific about this approach, including support for \c{pkg-config},
which, at least in case of \c{build2}, works equally well on Windows.|

Next, we create a build configuration and configure it to use one of these
\c{unpkg-*} directories (replace \c{...} with the absolute path):

\
$ bdep init -C ../hello-gcc @gcc cc config.cxx=g++ \
  config.cc.poptions=-I.../unpkg-gcc/include       \
  config.cc.loptions=-L.../unpkg-gcc/lib
\

\N|If using Visual Studio, replace \c{-I} with \c{/I} and \c{-L} with
\c{/LIBPATH:}.|

Alternatively, if you want to reconfigure one of the existing build
configurations, then simply edit the \c{build/config.build} file (that is,
\c{hello-gcc/build/config.build} in our case) and adjust the \c{poptions} and
\c{loptions} values. Or you can use the build system directly to reconfigure
the build configuration (see \l{b(1)} for details):

\
b configure: ../hello-gcc/                    \
  config.cc.poptions+=-I.../unpkg-gcc/include \
  config.cc.loptions+=-L.../unpkg-gcc/lib
\

\N|If all the unpackaged libraries included \c{.pc} files, then the \c{-L}
alone would have been sufficient. However, it doesn't hurt to also add
\c{-I}, for good measure.|

Once this is done, adjust your \c{buildfile} to import the library:

\
import libs += libextra%lib{extra}
\

And your source code to use it:

\
#include 
\

\N|Notice that we don't add the corresponding \c{depends} value to the
project's \c{manifest} since this library is not a package. However, it is a
good idea to instead add a \l{bpkg#manifest-package-requires \c{requires}}
entry as a documentation to users of our project.|


\h1#structure|Project Structure|

\h#structure-canonical|Canonical Project Structure|

The primary goal of establishing the \i{canonical project structure} is to
create an ecosystem of packages that can coexist, are easy to comprehend by
both humans and tools, scale to complex, real-world requirements, and, last
but not least, are pleasant to develop.

The canonical structure is primarily meant for packages \- a single library or
program (or, sometimes, a collection of related programs) with a specific and
well-defined function. While it may be less suitable for more elaborate,
multi-library/program \i{end-products} that are not meant to be packaged, most
of the recommendations discussed below would still apply. Oftentimes, you
would start with a canonical project and expand from there. Note also that
while the discussion below focuses on C++, most of it applies equally to C
projects.

Projects created by the \l{bdep-new(1)} command have the canonical structure.
The overall layout for executable (\c{-t\ exe}) and library (\c{-t\ lib})
projects are presented below.

\
/
├── build/
├── /
│   ├── .cxx
│   ├── testscript
│   └── buildfile
├── buildfile
└── manifest
\

\
lib/
├── build/
├── lib/
│   ├── .hxx
│   ├── .cxx
│   ├── export.hxx
│   ├── version.hxx.in
│   └── buildfile
├── tests/
├── buildfile
└── manifest
\

The canonical structure for both project types is discussed in detail next.
Below is a short summary of the key points:

\ul|

\li|\n\n\i{Library project names start with the \c{lib} prefix, for example,
\c{libhello}.}|

\li|\n\i{Header and source files (or module interface and implementation
files) are next to each other (no \c{include/} and \c{src/} split).}|

\li|\n\i{Header inclusions use \c{<>} and contain the project directory
prefix, for example, \c{}.}|

\li|\n\i{Header and source file extensions are either \c{.hpp/.cpp} or
\c{.hxx/.cxx} (\c{.mpp} or \c{.mxx} for module interfaces).}|

\li|\n\i{No special characters other than \c{_} and \c{-} in file names with
\c{.} only used for extensions.}\n\n|

|

Let's start with naming projects: by convention, library names start with the
\c{lib} prefix and using this prefix for non-library projects should be
avoided. The \c{bdep-new} command warns about both violations.

The project's root directory should contain the root \c{buildfile} and package
\c{manifest}.

The project's source code is placed into a subdirectory of the root directory
named the same as the project, for example, \c{hello/hello/} or
\c{libhello/libhello/}. It is called the project's \i{source subdirectory}.

There are several reasons for this layout: It implements the canonical
inclusion scheme (discussed below) where each header is prefixed with its
project name. It also has a predictable name where users (and tools) can
expect to find our project's source code. Finally, this layout prevents
clutter in the project's root directory which usually contains various other
files (like \c{README}, \c{LICENSE}) and directories (like \c{tests/}).

\N|Another popular approach is to place (public) headers into the \c{include/}
subdirectory and source files (as well as private headers) into \c{src/}. The
advantage of this layout is the predictable location that contains only the
project's public headers (that is, its API). This can make the project easier
to navigate and understand while harder to misuse (for example, by including a
private header).

However, this split layout is not without drawbacks:

\ul|

\li|Navigating between corresponding headers and sources is cumbersome. This
affects editing, grep'ing as well as code browsing (for example, on GitHub).|

\li|Implementing the canonical inclusion scheme would require an extra level
of subdirectories (for example, \c{include/libhello/} and \c{src/libhello/}),
which only amplifies the previous issue.|

\li|Supporting generated source code can be challenging: Source code
generators rarely provide support for writing headers and sources into
different directories. Even if we can move things around post-generation,
build systems may not support this arrangement (for example, \c{build2} does
not currently support target groups with members in different directories).||

Also, the stated advantage of this layout (separation of public headers) is
not as clear cut as it may seem. The common assumption of the split layout is
that only headers from \c{include/} are installed and, conversely, to use the
headers in-place, all one has to do is add \c{-I} pointing to \c{include/}.
On the other hand, it is common for public headers to include private, for
example, to call an implementation-detail function in inline or template code
(note that the same applies to private modules imported in public module
interfaces). Which means such private, (or, probably now more accurately
called implementation-detail) headers have to be placed in the \c{include/}
directory as well, perhaps into a subdirectory (such \c{details/}) or with a
file name suffix (sich as \c{-impl}) to signal to the user that they are still
\"private\". Needless to say, keeping track of which private headers can still
stay in \c{src/} and which have to be moved to \c{include/} (or vice versa) is
an arduous, error-prone task. As a result, practically, the split layout often
degrades into the \"all headers in \c{include/}\" arrangement which negates
its main advantage.

It's also not clear how the split layout will fit modularized projects. With
modules, both the interface and implementation (including non-inline/template
function definitions) can reside in the same file with a substantial number of
developers finding this arrangement appealing. If a project consists of only
such single-file modules, then \c{include/} and \c{src/} are effectively
become the same thing. In a sense, we already have this situation with
header-only libraries except that in case of modules calling the directory
\c{include/} would be an anachronism.

To summarize, the split directory arrangement offers little benefit over the
single directory layout, has a number of real drawbacks, and does not fit
modularized projects well. In practice, private headers are placed into
\c{include/}, often either in a subdirectory or with a special file name
suffix, a mechanis that is readily available to the single directory layout.|

All headers within a project should be included using the \c{<>} style and
contain the project name as a directory prefix. And all headers means \i{all
headers} \- public, private, or implementation detail, in executables and in
libraries.

As an example, let's say we've added \c{utility.hxx} to our \c{hello}
executable project. This is how it should be included in \c{hello.cxx}:

\
// hello/hello.cxx

// #include \"utility.hxx\"           // Wrong.
// #include            // Wrong.
// #include \"../hello/utility.hxx\"  // Wrong.

#include 
\

Similarly, if we want to include \c{hello.hxx} from \c{libhello}, then it
should look like this:

\
#include 
\

\N|The problem with the \c{\"\"} style inclusion is if the header is not found
relative to the including file, most compilers will continue looking for it in
the include search paths (\c{-I}). As a result, if the header is not present
in the right place (for example, because it was mistakenly not listed as to be
installed), chances are that a completely unrelated header with the same name
will be found and included. Needless to say, debugging situations like these
is unpleasant.

Similarly, prefixing all inclusions with the project name makes sure that
headers with common names (for example, \c{utility.hxx}) can coexist (for
example, when installed into a system-wide directory, such as
\c{/usr/include}). The prefix also plays an important role in supporting
auto-generated headers.

Finally, note that this header inclusion scheme is consistent with the module
importation, for example:

\
import hello.utility;
\

So let's imagine the \c{\"\"} style inclusion does not exist and we will all
have a much better time.|

If you have to disregard every rule and recommendation in this section but
one, for example, because you are working an existing library, then insist on
this: \i{public header inclusions must use the library name as a directory
prefix}.

The project's source subdirectory can have subdirectories of its own, for
example, to organize the code into components. Naturally, header inclusions
will need to contain such subdirectories, for example
\c{}. When the project's headers are installed (for
example, into \c{/usr/include}), this subdirectory hierarchy is
automatically recreated.

If you would like to separate public API headers/modules from implementation
details, then the convention is to place then into the \c{details/}
subdirectory. For example:

\
libhello/
└── libhello/
    ├── details/
    │    └── utility.hxx
    └── ...
\

It is recommended that you still install the implementation details headers
and modules for the reasons discussed above. If, however, you would like to
disable their installation, you can add the following line to your source
subdirectory's \c{buildfile}:

\
details/hxx{*}: install = false
\

When naming source files, only use ASCII alphabetic characters, digits, as
well as \c{_} (underscore) and \c{-} (minus). Use \c{.} (dot/period) only for
extensions, that is, trailing parts of the name that \i{classify} your files.
Examples of good names:

\
SmallVector.hxx
small-vector.hxx
small_vector.hxx
small-vector.test.cxx
\

Example of bad names:

\
small+vector.hxx
small.vector.hxx
\

The C source file extensions are always \c{.h}/\c{.c}. The two alternative C++
source file extension schemes are \c{.?pp} and \c{.?xx}:

\
file        .?pp  .?xx

header      .hpp  .hxx
module      .mpp  .mxx
inline      .ipp  .ixx
template    .tpp  .txx
source      .cpp  .cxx
\

The use of inline and template files is a matter of taste. If used, they are
included at the end of the header/module files and contain definitions of
inline and non-inline template functions, respectively. The \c{.?xx}/\c{.?pp}
files with the same name (or, sometimes, name prefix) are assumed to be
related and are collectively called a \i{module}. \N{This term is meant to
correspond directly to a C++ module.}

By default the \l{bdep-new(1)} command use the \c{.?xx} scheme. To use
\c{.?pp} instead, pass \c{-t\ c++,cpp}.

\N|There are several reasons not to \"reuse\" the \c{.h} C header extension
for C++ files:

\ul|

\li|There can be a need for both C and C++ headers for the same module.|

\li|It allows tools to accurately determine the language from the file name.|

\li|It is easier to search for C++ source code using wildcard patterns
(\c{*.?pp}).||

The last two reasons are also why headers without extensions are probably not
worth the trouble.|

Let's now move inside our source file: All macros defined by a project, such
as include guards, version macro, etc., must all start with the project name
(including the \c{lib} prefix for libraries), for example
\c{LIBHELLO_VERSION}. Similarly, the library's namespace and module names all
start with the library name but without the \c{lib} prefix. For example:

\
// libhello/hello.mxx

export module hello.core

namespace hello
{
  ...
}
\

Executable project may use a namespace (in which case it is natural to name it
after the project) and its modules shouldn't be qualified with the project
name (in order not to clash with similarly named modules from the
corresponding library, if any).

\N|Hopefully by now the insistence on the \c{lib} prefix should be easy to
understand: oftentimes executables and libraries come in pairs, for example
\c{hello} and \c{libhello}, with the reusable functionality being factored out
from the executable and into the library. It is natural to want to use the
same name \i{stem} (\c{hello} in our case) for both.

The above naming scheme (with the \c{lib} prefix present in some names but not
the others) is carefully crafted to allow such library/executable pairs to
coexist and be used together without too much friction. For example, both the
library and executable can have a header called \c{utility.hxx} with the
executable being able to include both and even get the \"merged\"
functionality without any extra effort (since they use the same namespace):

\
// hello/hello.cxx

#include 
#include 

namespace hello
{
  // Contains names from both utilities.
}
\

|

The source file that implements a module's unit tests should be placed next to
that module's other files and called with the module's name plus the \c{.test}
second-level extension. If a module uses Testscript for unit testing, then the
corresponding file should be called with the module's name plus the \c{.test}
extension. For example:

\
libhello/
└── libhello/
    ├── hello.hxx
    ├── hello.cxx
    ├── hello.test.cxx
    └── hello.test
\

\N|All source files (that is, headers, modules, etc.) with the \c{.test}
second-level extension are assumed to belong to unit tests and are
automatically excluded from the library/executable sources.|

The canonical library project created by \c{bdep-new} includes the \c{tests/}
subdirectory which contains the library's integration (as opposed to unit)
tests. Or, in other words, these are the tests that excercise the library via
its public interface, just like the real users of the library would. The
\c{tests/} subdirectory is an unnamed subproject (in the build system terms)
which allows us to build and run tests against an installed version of the
library. \N{The \c{build2} CI implementation will automatically perform this
test is a library contains the \c{tests/} subproject. See \c{bbot}
\l{bbot#arch-worker Worker Logic} for details.}

By default executable projects do not have the \c{tests/} subprojects instead
placing integration tests next to the source code (the \c{testscript} file;
see \l{testscript The build2 Testscript Language} for details). However, if
desired, executable projects can have the \c{tests/} subproject, the same as
libraries.

Other recommended top-level subdirectory names are \c{examples/} (for
libraries it is normally a subproject like \c{tests/}), \c{doc/}, and \c{etc/}
(sample configurations, scripts, third-party contributions, etc).

Note that there are no \c{bin/} or \c{obj/} subdirectories: output (object
files, libraries, executables, etc.) go into a parallel directory structure
(in case of an out-of-source build) or next to the sources (in case of an
in-source build).

Projects managed with \l{bdep(1)} are always built out-of-source. However, by
default, the source directory is configured as \i{forwarded} to one of the
out-of-source builds. This has two effects: we can run the build system driver
\l{b(1)} directly in the source directory and certain \"interesting\" output
(such as executables, documentation, test results, etc) will be automatically
\i{backlinked} to the source directory (see \l{b(1)} for details on forwarded
configurations). The following listing illustrates this setup for our
\c{hello} project (executables are marked with \c{*}):

\
                 hello-gcc/
hello/    ~~~    └── hello/
├── build/    ~~~    ├── build/
└── hello/    ~~~    └── hello/
    ├── hello.cxx        ├── hello.o
    └── hello     -->    └── *hello
\

The result is an \i{as-if} in-source build with all the benfits (such as
having both source and relevant output in the same directory) but without any
of the drawback (such as the inability to have multiple builds or source
directory cluttered with object files).

\N|The often cited motivation for placing executables into \c{bin/} is that in
many build system by also copying shared libraries there it is the only way to
make things runnable in a reasonably cross-platform manner. The major drawback
of this arrangement is the need for unique executable names which is
especially constraining when writing tests where it is convenient to call the
executable just \c{driver} or \c{test}.

In \c{build2} there is not such restrictions and all executables can run
\i{in-place}. This is achieved with \c{rpath} which is emulated with DLL
assemblies on Windows.|
"