path: root/doc/manual.cli

// file      : doc/manual.cli
// copyright : Copyright (c) 2014-2017 Code Synthesis Ltd
// license   : MIT; see accompanying LICENSE file

"\name=build2-build-bot-manual"
"\subject=build bot"
"\title=Build Bot"

// NOTES
//
// - Maximum <pre> line is 70 characters.
//

"
\h0#preface|Preface|

This document describes \c{bbot}, the \c{build2} build bot.

\h1#intro|Introduction|

\h1#arch|Architecture|

The \c{bbot} architecture includes several layers for security and
manageability. At the top we have a \c{bbot} running in the \i{controller}
mode. The controller monitors various \i{build sources} for \i{build
tasks}. For example, a controller may poll a \c{brep} instances for any new
packages to built as well as monitor a \c{git} repository for any new commits
to test. There can be several layers of controllers with \c{brep} being just a
special kind. A machine running a \c{bbot} instance in the controller mode is
called a \i{controller host}.

Below the controllers we have a \c{bbot} running in the \i{agent} mode
normally on Build OS. The agent polls its controllers for \i{build tasks} to
perform. A machine running a \c{bbot} instance in the agent mode is called a
\i{build host}.

The actual building is performed in the virtual machines and/or containers
that are executed on the build host. Inside virtual machines/containers,
\c{bbot} is running in the \i{worker mode} and receives build tasks from its
agent. Virtual machines and containers running a \c{bbot} instance in the
worker mode are collectively called \i{build machines}.

Let's now examine the workflow in the other direction, that is, from a worker
to a controller. Once a build machine is booted (by the agent), the worker
inside connects to the TFTP server running on the build host and downloads the
\i{build task manifest}. It then proceeds to perform the build task and
uploads the \i{build result manifest} (which includes build logs) to the TFTP
server.

Once an agent receives a build task for a specific build machine, it goes
through the following steps. First, it creates a directory on its TFTP server
with the \i{machine name} as its name and places the build task manifest
inside. Next, it makes a throw-away snapshot of the build machine and boots
it. After booting the build machine, the agent monitors the machine directory
on its TFTP server for the build result manifest (uploaded by the worker once
the build has completed). Once the result manifest is obtained, the agent
shuts down the build machine and discards its snapshot.

To obtains a build task the agent polls via HTTP/HTTPS one or more
controllers. Before each poll request the agent enumerates the available build
machines and sends this information as part of the request. The controller
responds with a build task manifest that identifies a specific build machine
to use.

If the controller has higher-level controllers (for example, \c{brep}), then
it aggregates the available build machines from its agents and polls these
controllers (just as an agent would), forwarding build tasks to suitable
agents. In this case we say that the \i{controller act as an agent}. The
controller may also be configured to monitor build sources, such as SCM
repositories, directly in which case it generates build tasks itself.

In this architecture the build results are propagated up the chain: from a
worker, to its agent, to its controller, and so on. A controller that is the
final destination of a build result uses email to notify interested parties of
the outcome. For example, \c{brep} would send a notification to the package
owner if the build failed. Similarly, a \c{bbot} controller that monitors a
\c{git} repository would send an email to a committer if their commit caused a
build failure. The email would include a link (normally HTTP/HTTPS) to the
build logs hosted by the controller.

\h#arch-machine-config|Configurations|

The \c{bbot} architecture distinguishes between a \i{machine configuration}
and a \i{build configuration}. The machine configuration captures the
operating system, installed compiler toolchain, and so on. The same build
machine may be used to \"generate\" multiple \i{build configurations}. For
example, the same machine can normally be used to produce 32/64-bit and
debug/release builds.

The machine configuration is \i{approximately} encoded in its \i{machine
name}. The machine name is a list of components separated with \c{-}. Each
component can contain alpha-numeric characters, underscores, dots, and pluses
with the whole id being a portably-valid path component.

The encoding is approximate in a sense that it captures only what's important
to distinguish in a particular \c{bbot} deployment.

The first component normally identifies the operating system and has the
following recommended form:

\
[<arch>_][<class>_]<os>[_<version>]
\

For example:

\
windows
windows_10
windows_10.1607
i686_windows_xp
bsd_freebsd_10
linux_centos_6.2
linux_ubuntu_16.04
macos_10.12
\

The second component normally identifies the installed compiler toolchain and
has the following recommended form:

\
<id>[<version>][<runtime>]
\

For example:

\
gcc
gcc_6
gcc_6.3
clang_3.9_libc++
clang_3.9_libstdc++
msvc_14
msvc_14u3
icc
\

Some examples of complete machine names:

\
windows_10-msvc_14u3
macos_10.12-clang
linux_ubuntu_16.04-gcc_6.3
\

Similarly, the build configuration is encoded in a \i{configuration name}
using the same format. As described in \l{#arch-controller Controller Logic},
build configurations are generated from machine configurations. As a result,
it usually makes sense to have the first component identify the operating
systems and the second component \- the toolchain with the rest identifying
a particular build configuration. For example:

\
windows-vc_14-32-debug
linux-gcc_6-cross-arm-eabi
\

\h#arch-machine-header-manifest|Machine Header Manifest|

\
SYNOPSIS

id:      <machine-id>
name:    <machine-name>
summary: <string>
\

The build machine header manifest contains basic information about a build
machine on the build host. A list of machine header manifests is sent by
\c{bbot} agents to controllers.

\dl|

\li|\n\c{id: <machine-id>}\n

  The \i{machine-id} uniquely identifies a machine version/revision/build.
  For virtual machines this can be the disk image checksum. For a container
  this can be UUID that is re-generated every time a container filesystem
  is altered.|

\li|\n\c{name: <machine-name>}\n

  The machine name as described above.|

\li|\n\c{summary: <string>}\n

  A one-line description of the machine. For example:

  \
  id: windows_10-msvc_14-1.3
  name: windows_10-msvc_14
  summary: Windows 10 build 1607 with VC 14 update 3
  \

||

\h#arch-machine-manifest|Machine Manifest|

\
SYNOPSIS

id:      <machine-id>
name:    <machine-name>
summary: <string>

type: <machine-type>
mac:  <macaddr>
\

The build machine manifest contains the complete description of a build
machine on the build host (see the Build OS documentation for their origin and
location). The machine manifest starts with the machine manifest header. All
the header values must appear before any non-header values.

\dl|

\li|\n\c{type: <machine-type>}\n

  The machine type. Valid values are \c{kvm} (QEMU/KVM virtual machine) and
  \c{nspawn} (\c{systemd-nspawn} container).|

\li|\n\c{mac: <macaddr>}\n

  Optional fixed MAC address for the machine in the hexadecimal,
  comma-separated format. For example:

  \
  mac: de:ad:be:ef:de:ad
  \

  If it is not specified, then a random address is generated on the first
  machine bootstrap which is then reused for each build/re-bootstrap. Note
  that it you specify a fixed address, then the machine can only be used by a
  single \c{bbot} agent.

||


\h#arch-task-manifest|Task Manifest|

\
SYNOPSIS

name:       <package-name>
version:    <package-version>
#location:  <package-url>
repository: <repository-url>
trust:      <repository-fp>

machine: <machine-name>
target:  <target-triplet>
config:  <config-vars>
\

The task manifest describes a build task. It consists of two groups of values.
The first group defines the package to build. The second group defines the
build configuration to use for building the package.

\dl|

\li|\n\c{name: <package-name>}\n

  Package name to test.|

\li|\n\c{version: <package-version>}\n

  Package version to test.|

\li|\n\c{repository: <repository-url>}\n

  The \c{bpkg} repository that contains the package and its dependencies.|

\li|\n\c{trust: <repository-fp>}\n

  The SHA256 repository certificate fingerprint to trust (see the \c{bpkg}
  \c{--trust} option for details). This value may be specified multiple times
  to establish the authenticity of multiple certificates. If the special
  \c{yes} value is specified, then all repositories will be trusted without
  authentication (see the \c{bpkg} \c{--trust-yes} option).

  Note that while the controller may return a task with \c{trust} values,
  whether they will be used is up to the agent's configuration. For example,
  some agents may only trust their internally-specified fingerprints to
  prevent the \"man in the middle\" type of attacks.|

\li|\n\c{machine: <machine-name>}\n

  The name of the build machine to use.|

\li|\n\c{target: <target-triplet>}\n

  The target triplet to build for. If not specified, then the default target
  for this machine is used (which is usually the machine itself).

  Compared to the autotools terminology, the \c{machine} value corresponds
  to \c{--build} (the machine we are building on) and \c{target} \- to
  \c{--host} (the machine we are building for). While we use essentially
  the same \i{target triplet} format as autotools for \c{target}, it is
  not flexible enough for \c{machine}.|

\li|\n\c{config: <config-vars>}\n

  Additional build system configuration variables.

  A single level of quotes (either single or double) is removed in each
  variable before being passed to \c{bpkg}. For example, the following value:

  \
  config: config.cc.coptions=\"-O3 -stdlib='libc++'\"
  \

  Will be passed to \c{bpkg} as the following (single) argument:

  \
  config.cc.coptions=-O3 -stdlib='libc++'
  \

  Variables can be separated with spaces or newlines.

  ||


\h#arch-result-manifest|Result Manifest|

\
SYNOPSIS

name:     <package-name>
version:  <package-version>

status:           <status>
configure-status: <status>
update-status:    <status>
test-status:      <status>

configure-log: <text>
update-log:    <text>
test-log:      <text>
\

The result manifest describes a build result.

\dl|

\li|\n\c{name: <package-name>}\n

  Package name from the task manifest.|

\li|\n\c{version: <package-version>}\n

  Package version from the task manifest.|

\li|\n\c{status: <status>}\n

  An overall (cumulative) build result status. Valid values are:

  \
  success    # All operations completed successfully.
  warning    # One or more operations completed with warnings.
  error      # One or more operations completed with errors.
  abort      # One or more operations were aborted.
  abnormal   # One or more operations terminated abnormally.
  \

  The \c{abort} status indicates that the operation has been aborted by
  \c{bbot}, for example, because it was consuming too many resources and/or
  was taking too long. Note that a task can be aborted both by the \c{bbot}
  worker as well as the agent. In the later case the whole machine is shut
  down and no operation-specific status or logs will be included (@@ Maybe
  we should just include 'log:' with commands that start VM, for
  completeness?).

  The \c{abnormal} status indicates that the operation has terminated
  abnormally, for example, due to the package manager or build system crash.

  Note that the overall \c{status} value should appear before any
  per-operation \c{*-status} values.|

\li|\n\c{*-status: <status>}\n

  A per-operation result status. Note that the \c{*-status} values should
  appear in the same order as the corresponding operations were performed
  and for each \c{*-status} there should be a corresponding \c{*-log}.|

\li|\n\c{*-log: <text>}\n

  A per-operation result log. Note that the \c{*-log} values should appear
  last and in the same order as the corresponding \c{*-status} values.||


\h#arch-task-req-manifest|Task Request Manifest|

\
SYNOPSIS

agent:       <agent-name>
fingerprint: <agent-fingerprint>
\

An agent (or controller acting as an agent) sends a task request to its
controller via HTTP/HTTPS POST method (@@ URL/API endpoint). The task request
starts with the task request manifest followed by a list of machine manifests.

\dl|

\li|\n\c{agent: <agent-name>}\n

  The name of the agent host (\c{hostname}). This should be unique in a
  particular \c{bbot} deployment.|

\li|\n\c{fingerprint: <agent-fingerprint>}\n

  The SHA256 fingerprint of the agent's public key. If not present, and
  the controller is configured to require it, then it responds with the
  \c{unauthorized} (401) HTTP status code.||


\h#arch-task-res-manifest|Task Response Manifest|

\
SYNOPSIS

session:    <session-id>
challenge:  <text>
result-url: <url>
\

A controller sends the task response manifest in response to the task request
initiated by an agent. The response is delivered as a result of the POST
method. The task response starts with the task response manifest optionally
followed by a task manifest.

\dl|

\li|\n\c{session: <session-id>}\n

  An identifier assigned to this session by the controller. An empty value
  indicates that the controller has no tasks at this time in which case all
  the following values as well as the task manifest are absent.|

\li|\n\c{challenge: <text>}\n

  Random text (nonce) used to challenge the agent's private key. If present,
  then the agent must sign this text with its private key and include the
  signature in the result request.

  The signature should be derived by calculating the SHA256 checksum of the
  text, encrypting it with the agent's private key, and then base64-encoding
  the result.|

\li|\n\c{result-url: <url>}\n

  The URL to post the result (upload) request to.||


\h#arch-result-req-manifest|Result Request Manifest|

\
SYNOPSIS

session:   <session-id>
challenge: <text>
\

On completion of a task an agent (or controller acting as an agent) sends a
result (upload) request to the controller via HTTP/HTTPS POST method using the
URL returned in the task response. The result request starts with the
result request manifest followed by a result manifest. Note that there is no
result response and only a successful but empty POST result is returned.

\dl|

\li|\n\c{session: <session-id>}\n

  The session id as returned by the controller in the task response.|

\li|\n\c{challenge: <text>}\n

  The answer to the private key challenge as posed by the controller in the
  task response. Must present only if the challenge value was present in the
  task response.||


\h#arch-worker|Worker Logic|

The \c{bbot} worker builds each package in a \i{build environment} that is
established for a particular build target. The environment has three
components: the execution environment (environment variables, etc), build
system modules, and configuration variables.

Setting up of the environment is performed by an executable (script, batch
file, etc). Specifically, upon receiving a build task, the worker obtains its
target and looks for the environment setup executable with this name in a
specific directory. If not found or if the target is unspecified, then the
worker looks for the executable called \c{default}. Not being able to locate
the environment executable is an error.

Once the environment setup executable is determined, the worker re-executes
itself as that executable passing to it as command line arguments the target
name (or empty value if not specified), the path to the \c{bbot} worker to be
executed once the environment is setup, and any additional options that need
to be propagated to the re-executed worker. The environment setup executable
is executed in the build directory as its current working directory. The build
directory contains the build task \c{manifest} file.

The environment setup executable sets up the necessary execution environment
for example by adjusting \c{PATH} or running a suitable \c{vcvars} batch file.
It then re-executes itself as the \c{bbot} worker passing to it as command
line arguments (in addition to worker options) the list of build system
modules (\c{<env-modules>}) and the list of configuration variables
(\c{<env-config-vars>}). The environment setup executable must execute the
\c{bbot} worker in the build directory as the current working directory.

The re-executed \c{bbot} worker then proceeds to test the package from the
repository by executing the following commands (\c{<>}-values are from the
task manifest and environment):

\
bpkg -v create <env-module> <config-vars> <env-config-vars>
bpkg -v add <repository-url>
bpkg -v fetch --trust <repository-fp>
bpkg -v build --yes --configure-only <package-name>/<package-version>
bpkg -v update <package-name>
bpkg -v test <package-name>
\

As an example, the following POSIX shell script can be used to setup the
environment for building C and C++ packages with GCC 6 on most Linux
distributions.

\
#!/bin/sh

# Environment setup script for C/C++ compilation with GCC 6.
#
# $1  - target
# $2  - bbot executable
# $3+ - bbot options

set -e # Exit on errors.

t=\"$1\"
shift

if test -n \"$t\"; then
  echo \"unknown target: $t\" 1>&2
  exit 1
fi

exec \"$@\" cc config.c=gcc-6 config.cxx=g++-6
\

\h#arch-controller|Controller Logic|

A \c{bbot} controller that issues own build tasks maps available build
machines (as reported by agents) to \i{build configurations} according to the
\c{buildtab} configuration file. Blank lines and lines that start with \c{#}
are ignored. All other lines in this file have the following format:

\
<machine-name-pattern> <config-name> [<target>] [<config-vars>]
\

Where \c{<machine-name-pattern>} is filesystem wildcard pattern that is
matched against available machine names, \c{<config-name>} is the
configuration name, optional \c{<target>} is the build target, and optional
\c{<config-vars>} is a list of additional build system configuration variables
with the same quoting semantics as in the \c{config} value in the build task
manifest. The matched machine name, the target, and configuration variables
are included into the build task manifest.

Note that each machine name is matched against every pattern and all the
patterns that match produce a configuration. If a machine does not match any
pattern, then it is ignored (meaning that this controller is not interested in
testing its packages with this machine). If multiple machines match the same
pattern, then only a single configuration using any of the machines is
produced (meaning that this controller considers these machines equivalent).

As an example, let's say we have a machine named \c{windows_10-vc_14u3}.  If
we wanted to test both 32 and 64-bit builds as well as debug and release, then
we could have generated the following configurations:

\
windows*-vc_14* windows-vc_14-32-debug   i686-microsoft-win32-msvc14.0   config.cc.coptions=/Z7 config.cc.loptions=/DEBUG
windows*-vc_14* windows-vc_14-32-release i686-microsoft-win32-msvc14.0   config.cc.coptions=\"/O2 /Oi\"
windows*-vc_14* windows-vc_14-64-debug   x86_64-microsoft-win32-msvc14.0 config.cc.coptions=/Z7 config.cc.loptions=/DEBUG
windows*-vc_14* windows-vc_14-64-release x86_64-microsoft-win32-msvc14.0 config.cc.coptions=\"/O2 /Oi\"
\

As another example, let's say we have \c{linux_fedora_25-gcc_6} and
\c{linux_ubuntu_16.04-gcc_6}. If all we cared about it testing GCC 6 on Linux,
then our configurations could look like this (note the missing target):

\
linux*-gcc-6 linux-gcc_6-debug   config.cc.coptions=-g
linux*-gcc-6 linux-gcc_6-release config.cc.coptions=-O3
\
"