= Tebako: An advanced image packager for interpretive languages

Platform tests on GitHub:
image:https://github.com/tamatebako/tebako/actions/workflows/ubuntu.yml/badge.svg["Ubuntu amd64", link="https://github.com/tamatebako/tebako/actions/workflows/ubuntu.yml"]
image:https://github.com/tamatebako/tebako/actions/workflows/alpine.yml/badge.svg["Alpine", link="https://github.com/tamatebako/tebako/actions/workflows/alpine.yml"]
image:https://github.com/tamatebako/tebako/actions/workflows/macos.yml/badge.svg["macOS amd64", link="https://github.com/tamatebako/tebako/actions/workflows/macos.yml"]
image:https://github.com/tamatebako/tebako/actions/workflows/windows-msys.yml/badge.svg["Windows msys", link="https://github.com/tamatebako/tebako/actions/workflows/windows-msys.yml"]

Platform tests on Cirrus:
image:https://api.cirrus-ci.com/github/tamatebako/tebako.svg?branch=main&task=ubuntu-aarch64["Ubuntu aarch64", link="https://cirrus-ci.com/github/tamatebako/tebako"]

Tools tests on GitHub:
image:https://github.com/tamatebako/tebako-ci-containers/actions/workflows/build-containers.yml/badge.svg["Tebako cobtainers", link="https://github.com/tamatebako/tebako-ci-containers/actions/workflows/build-containers.yml"]

Quality:
image:https://github.com/tamatebako/tebako/actions/workflows/lint-and-rspec.yml/badge.svg["lint and rspec", link="https://github.com/tamatebako/tebako/actions/workflows/lint-and-rspec.yml"]
image:https://codecov.io/gh/tamatebako/tebako/graph/badge.svg?token=XD3emQ5qsY["Tebako cli rspec coverage", link="https://codecov.io/gh/tamatebako/tebako"]

== Purpose

Tebako is an advanced executable packager designed for applications written in
interpretive languages.

It simplifies distribution and deployment by packaging your entire project with
a bundled runtime into a single, performant, executable binary.

== Architecture

A Tebako package is effectively a self-executing container-in-a-file.

The package contains the following components:

* An on-file filesystem (OFFS) containing all the project files and
dependencies in DwarFS format ("application")

* A runtime environment that includes the the necessary libraries and interpreters,
with patched filesystem calls that redirect access of project files to the
on-file filesystem ("runtime")

Tebako is capable to create a single file that contains both runtime and
application or place runtime and application to separate files. In the latter
case the runtime can be used with different applications or versions of the same
application.

Please refer to `mode` option below that controls Tebako output.

== Supported runtimes, platforms and architectures

Tebako artifacts can be built and executed on the following platforms and
architectures.

.Supported platforms and architectures
[cols="3", options="header"]
|===
| Platform and version | Architectures | Build system

3+| **Linux**
| Ubuntu 20.04 | amd64, aarch64 | gcc/g+\+: 10; clang/clang++: 12
| Alpine 3.17 | amd64 | gcc/g+\+: default; clang/clang++: default

3+| **macOS**
| macOS 13 (Ventura)  | amd64, arm64 | tested agains xcode: [14.3.1]
| macOS 14 (Sonoma)   | amd64, arm64 | tested agains xcode: [15.0.1, 15.4]
| macOS 15 (Sequoia)  | amd64, arm64 | tested agains xcode: [16.1]

3+| **Windows**
| Windows 10 | amd64 | MinGW ucrt64
| Windows 11 | amd64 | MinGW ucrt64
| Windows Server 2019 | amd64 | MinGW ucrt64
| Windows Server 2022 | amd64 | MinGW ucrt64

|===

[NOTE]
====
Windows build caveats:

* Tebako may face errors related to CMake path length limitations
(https://gitlab.kitware.com/cmake/cmake/-/issues/25936).
This error may affect not Tebako itself but the gems that need to be package and
use CMake to build native extensions.
There is no workaround for this issue as it looks like is a limitation of the
manifest used to build CMake executable.

* MSys strip utility creates broken executable when Tebako image is processed.
Linking with `-s` flag produces unusable executables as well.
Until this issue (https://github.com/tamatebako/tebako/issues/172) is resolved
we plan to produce a Windows executable with debug information unstripped. You
can opt to run `strip -S` manually, it most cases it works.

MacOS build caveats:

* We saw `clang` compiler segmentation fault when packaging of very large
projects with XCode 14.3.1. This issue is not reproducible with XCode 15.0.1 or
higher.
====


.Supported Ruby versions
[cols="2", options="header"]
|===
| Ruby version | Supported platforms

| 2.7.8 | Linux, macOS
| 3.0.7 | Linux, macOS
| 3.1.6 | Linux, macOS, Windows
| 3.2.{4,5,6} | Linux, macOS, Windows
| 3.3.{3,4,5,6} | Linux, macOS, Windows
| 3.4.1 | Linux, macOS, Windows

|===

NOTE: Our goal is to support all maintained Ruby releases, including minor versions.


== Package portability

=== General

Tebako packages are designed to be "forward portable" across different operating
systems and architectures to allow for easy distribution and deployment.

Forward portability means that a package created on a specific platform can be
executed on a newer version of the same platform.

=== macOS

macOS packages are forward portable across different macOS versions.

[example]
A Tebako executable package built on macOS 13 (Ventura) can be executed on
macOS 14 (Sonoma), but not vice versa.

`x86_64` macOS packages can be run on Apple M (ARM) systems.


=== Linux distributions using `musl`

Packages built for the
https://musl.libc.org[`musl` implementation of the C standard library]
(such as https://alpinelinux.org[Alpine Linux]) are forward portable.

[example]
A Tebako executable package built on Alpine 3.17 can be executed on Alpine 3.19.

Usage of the
https://github.com/tamatebako/tebako-ci-containers[Tebako Docker containers] for
packaging is encouraged since it eliminates the effort needed for toolchain
setup and configuration.


=== Linux distributions using `glibc`

Packages built for the
https://sourceware.org/glibc[`glibc` implementation of the C standard library]
are forward portable if the `--patchelf` experimental option is enabled.

The `--patchelf` option allows these packages to be portable to Linux GNU
distributions with GLIBC version 2.31 and above.

[example]
A Tebako executable package built on Ubuntu 20.04 with `--patchelf` option can
be executed on Rocky Linux 9.

Usage of the
https://github.com/tamatebako/tebako-ci-containers[Tebako Docker containers] for
packaging is encouraged since it eliminates the effort needed for toolchain
setup and configuration.


.Minimum versions of GLIBC Linux distributions that support Tebako packages with forward portability
[cols="3", options="header"]
|===
| Distribution | Minimal supported version | GLIBC version

| Ubuntu | 20.04 (Focal Fossa) | GLIBC 2.31
| Debian | 11 (Bullseye) | GLIBC 2.31
| Rocky Linux | 9 | GLIBC 2.34
| Fedora | 33 | GLIBC 2.32
| CentOS | 9 | GLIBC 2.34
| Red Hat Enterprise Linux (RHEL) | 9 | GLIBC 2.34
| Oracle Linux | 9 | GLIBC 2.34

|===


== Future plans

* Downloading new DwarFS images to be stored in the local home directory
* Allowing loading multiple DwarFS images in a stacked way
* Supporting a COW mechanism that the newly written files are stored
  in a separate image that can be loaded on top of the read-only file systems.

== FAQ

=== Why use Tebako?

Tebako is particularly useful for developers who need to:

* Distribute applications without requiring users to have specific runtimes installed.
* Simplify the deployment process by packaging all dependencies into one binary.
* Ensure consistency across different environments by using a single executable.
* Flexibility to support different runtime versions on the user's machine.


=== How do I know I need Tebako?

You might need Tebako if you:

* Want to package your application into a single, self-contained binary.
* Want to avoid the complexities of managing runtime environments on target machines.
* Distribute software to environments where installing runtimes and their dependencies is challenging.
* Require a streamlined way to deliver applications to end-users.
* Need to ensure that your application runs consistently across different environments and architectures.


=== What is DwarFS?

https://github.com/mhx/dwarfs[DwarFS] is a fast, high compression read-only
user-land file system designed for efficient storage and access of large
collections of files.

It is used by Tebako to package applications into a compact and efficient format.

Tebako uses https://github.com/tamatebako/libdwarfs[libdwarfs], the library
form of https://github.com/mhx/dwarfs[DwarFS], developed for the Tebako project.

=== When is Tebako better than comparable solutions?

Tebako offers several advantages over comparable solutions for supported
interpretive languages.

They are listed in order of the degree of virtualization below.

Tebako stands out by providing a lightweight runtime bundling approach that
simplifies distribution and deployment while offering flexibility and
efficiency.

It eliminates the need for users to have specific runtimes installed and ensures
consistency across different environments.

With Tebako, you can package your entire project with a bundled runtime into a
single, performant, executable binary.

[cols="a,3a,3a"]
|===
| Solution | Pros | Cons

| Virtual machines (VMs)
|
- Provides full isolation and compatibility across environments
|
- Requires a separate VM installation for each application
- Heavy resource consumption for virtualization

| Docker
|
- Provides portable containers
- Isolates entire applications and their dependencies
- Supports easy deployment and scalability
|
- Requires Docker installation and management
- Requires administrative rights on machine
- Containerization overhead

| *Tebako*
|
- Packages all files and dependencies into a single binary
- Supports multiple operating systems and architectures
- Provides efficient packaging and execution with DwarFS
- Offers security features like signing on macOS
- Simplifies distribution and deployment
- Native running speed
|
- Initial packaging time longer than Ruby gems
- Minor runtime overhead

| Ruby Gems
|
- Easy installation of Ruby libraries
- Provides user-side version control and dependency management
|
- Requires Ruby installation and gem management
- Runtime execution dependent on the user's installed Ruby version and gems

|===


== Usage

=== Command-line interface

Tebako works by packaging your project into a single executable binary that
includes all the necessary dependencies.

The way to work with Tebako is through its command-line interface (CLI).
It provides the following commands:

`setup`::
Prepares the Tebako packaging environment.

`press`::
Packages a project into a single executable binary.

`clean`::
Removes Tebako artifacts.

`clean_ruby`::
Removes Tebako Ruby artifacts.

`hash`::
Calculates the Tebako script hash for use as a cache key in CI/CD environments.

`version`::
Displays the Tebako version.

`help`::
Displays the help message.


== Usage

=== General

Tebako can be used in two ways:

* Through the Tebako container
* Local installation

Please refer to the <<installation>> section on how to install Tebako.


[[installation]]
== Installation

=== General

Installation of Tebako is only needed in order to package an application.

There is no need to install anything for users who run the packaged application.


=== Using Docker

==== General

If you have Docker installed and available, the easiest way to run Tebako is
through the official Docker containers.

Docker containers with preinstalled Tebako packaging environments for Ubuntu and
Alpine Linux are available at
https://github.com/tamatebako/tebako-ci-containers[tebako-ci-containers].


==== Pull the container

Pull the Tebako container image.

[source,sh]
----
docker pull ghcr.io/tamatebako/tebako-<container_tag>:latest
----

`<container_tag>`:: is the desired image tag (e.g., `ubuntu-20.04` or `alpine-3.17`).


==== Running Tebako commands in the container

Simply prefix the Tebako command with `docker run` and the container image.

[source,sh]
----
docker run -v <application_folder>:/mnt/w \
  -t ghcr.io/tamatebako/tebako-<container_tag>:latest \
  tebako {command} {parameters}
----

==== Packaging from outside the container

To package your application from outside the container, just run a single Docker
command.

This command mounts the application folder into the container and runs the
`tebako press` command, specifying the application root, entry point, output
location, and Ruby version.

[source,sh]
----
docker run -v <application_folder>:/mnt/w \
  -t ghcr.io/tamatebako/tebako-<container_tag>:latest \
  tebako press <tebako-press-parameters>
----

`<application_folder>`:: is the path to your application folder.

`<container_tag>`:: is the desired image tag (e.g., `ubuntu-20.04` or `alpine-3.17`).


[example]
====
Assume that you have a Ruby application in the `fontist` folder of the current
directory.

You can package it to `./fontist-package` using the following command:

[source,sh]
----
docker run -v $PWD:/mnt/w \
  -t ghcr.io/tamatebako/tebako-ubuntu-20.04:latest \
  tebako press --root=/mnt/w/fontist --entry-point=fontist --output=/mnt/w/fontist-package --Ruby=3.2.4
----
====

==== Packaging from inside the container

It is also possible to package an application from inside the Tebako container.

Start and enter the container interactively.

[source,sh]
----
docker run -it --rm -v <application_folder>:/mnt/w \
  ghcr.io/tamatebako/tebako-<container_tag>:latest bash
----

`<application_folder>`:: is the path to your application folder.

`<container_tag>`:: is the desired image tag (e.g., `ubuntu-20.04` or `alpine-3.17`).


Once inside, run the `tebako press` command:

[source,sh]
----
tebako press <tebako press parameters>
----

[example]
====
Assume that you have a Ruby application in the `fontist` folder of the current
directory.

You can package it to `./fontist-package` using the following command:

[source,sh]
----
$ docker run -it --rm -v $PWD:/mnt/w ghcr.io/tamatebako/tebako-<container_tag>:latest bash

# Inside the container:
$ tebako press --root=/mnt/w/fontist --entry-point=fontist --output=/mnt/w/fontist-package --Ruby=3.2.4
----
====


=== Local installation

==== General

There are cases where Docker may not be suitable for your needs, such as:

. Admin privileges: Running Docker requires administrative privileges, which
means Docker may not be available to users on their machines.

. Performance penalty: Docker introduces a performance penalty due to the
overhead of running containers. This can be a concern when packaging complex
applications that require heavy memory usage.

In such cases, you can choose to install Tebako locally.

Tebako is distributed as a Ruby gem. A Ruby environment is necessary.


[source,sh]
----
$ gem install tebako
----


==== Prerequisites

These prerequisites are needed only for users who want to install Tebako on
their machine and build all Tebako components locally.

If you use Docker, there is no need to set up these prerequisites.

===== Ubuntu 20.04

====== General

There are several prerequisites that need to be installed on Ubuntu 20.04 for
Tebako to work correctly.


====== GNU C/C++ 10+ or Clang C/C++ 12+

[source,sh]
----
apt install -y gcc-10 g++-10
update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-10 10
update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-10 10
----

or

[source,sh]
----
apt install -y clang-12
update-alternatives --install /usr/bin/clang clang /usr/bin/clang-12 150
update-alternatives --install /usr/bin/clang++ clang++ /usr/bin/clang++-12 150
----

====== CMake version 3.20+

Tebako requires CMake at a version of at least 3.20+.

If such CMake version is not available as a default package, set it up as
follows.

.Installing CMake 3.20+
[source,sh]
----
apt-get remove --purge --auto-remove cmake
apt-get update
apt-get install -y software-properties-common lsb-release curl
apt-get clean all
curl https://apt.kitware.com/kitware-archive.sh | bash
apt-get install cmake
----

====== Other development tools and libraries

[source,sh]
----
apt-get -y install sudo git curl build-essential pkg-config bison flex autoconf  \
   binutils-dev libevent-dev acl-dev libfmt-dev libjemalloc-dev libiberty-dev    \
   libdouble-conversion-dev liblz4-dev liblzma-dev libssl-dev libunwind-dev      \
   libboost-filesystem-dev libboost-program-options-dev libboost-system-dev      \
   libboost-iostreams-dev  libboost-date-time-dev libboost-context-dev           \
   libboost-regex-dev libboost-thread-dev libbrotli-dev libdwarf-dev libelf-dev  \
   libgoogle-glog-dev libffi-dev libgdbm-dev libyaml-dev libncurses-dev          \
   libreadline-dev libncurses-dev libreadline-dev ruby-dev ruby-bundler          \
   libutfcpp-dev
----

===== Alpine 3.17

====== General

There are several prerequisites that need to be installed on Alpine 3.17 for
Tebako to work correctly.

====== APK packages

Run the following command to install all prerequisites.

[source,sh]
----
apk --no-cache --upgrade add build-base cmake git bash autoconf boost-static   \
   boost-dev flex-dev bison make binutils-dev libevent-dev acl-dev sed python3 \
   pkgconfig lz4-dev openssl-dev zlib-dev xz ninja zip unzip curl libdwarf-dev \
   libunwind-dev gflags-dev elfutils-dev libevent-static openssl-libs-static   \
   lz4-static xz-dev zlib-static libunwind-static acl-static tar libffi-dev    \
   gdbm-dev yaml-dev yaml-static ncurses-dev ncurses-static readline-dev       \
   readline-static p7zip ruby-dev gcompat gettext-dev gperf brotli-dev         \
   brotli-static jemalloc-dev fmt-dev xz-static
----

===== macOS

====== General

There are several prerequisites that need to be installed on macOS for Tebako to work correctly.

The following instructions work for:

* macOS 13 (Ventura) through macOS 15 (Sequoia)


====== Homebrew packages

We use Homebrew to install the necessary packages on macOS.

[source,sh]
----
brew update
brew install gnu-sed bash pkg-config bison flex binutils libffi gdbm zlib \
  ncurses double-conversion boost jemalloc fmt glog libevent libsodium lz4 xz \
  libyaml openssl@3
brew bundle
----

Additionaly tebako repository includes `Brewfile` that can be used with
`brew bundle` command.

[source,sh]
----
brew bundle
----

====== Bison 3+

Tebako requires Bison 3+.

On macOS 14, the default Bison version is 2.3, and the Homebrew formula is
keg-only, which means that the full path to the Bison binary must be used to
utilize the correct version.

Run the following command prior to using Tebako, or add it into your shell
profile.

[source,sh]
----
export PATH="$(brew --prefix bison)/bin:$PATH"
----

====== jemalloc library build

On macOS, the `libdwarfs` build script creates an additional `jemalloc`
installation. This is done to satisfy the magic applied by folly during linking
but uses a static library.

If the `jemalloc` library is built within an emulated environment (QEMU,
Rosetta, etc.), there are known issues
(link:https://github.com/jemalloc/jemalloc/issues/1997[jemalloc issue #1997])
where `jemalloc` incorrectly detects the number of significant virtual address
bits and therefore defines them wrongly (`lg-vaddr` parameter).

Technically, these issues can be fixed by explicitly setting the
`--with-lg-vaddr` parameter for the `jemalloc` build. However, we decided not to
automate this since we do not feel that we can provide reasonable test coverage.

Instead, our build script accepts the `LG_VADDR` environment variable and passes
it to the jemalloc build as `--with-lg-vaddr=${LG_VADDR}`.

The `LG_VADDR` parameter specifies the number of significant virtual address
bits, which can vary based on the CPU architecture and emulation status.

This is a simple example script to set `LG_VADDR`.

NOTE: This is provided for illustration only.

[example]
====
[source,sh]
----
#!/bin/bash

# Check the CPU architecture
ARCH=$(uname -m)

# Check if running under Rosetta 2 emulation
if [[ "$ARCH" == "x86_64" && $(sysctl -n sysctl.proc_translated) == "1" ]]; then
  echo "Running on Apple Silicon under Rosetta 2 emulation"
  export LG_VADDR=39
elif [[ "$ARCH" == "arm64" ]]; then
  echo "Running on Apple Silicon"
  export LG_VADDR=39
else
  echo "Running on Intel Silicon"
  export LG_VADDR=48
fi

echo "Setting lg-vaddr to $LG_VADDR"
----
====


===== Windows

====== General

The following instructions work for:

* Windows 10, 11
* Windows Server 2019, 2022

====== Ruby

To run Tebako you need to have Ruby installed.
It is simplest to use the Ruby development environment provided by
https://rubyinstaller.org[RubyInstaller].

For example, Ruby+Devkit 3.1.4-1.

====== MinGW ucrt64

Enable MinGW ucrt64 and install the necessary packages.

The `ridk` command originates from the RubyInstaller installation.

[source,sh]
----
$ ridk enable ucrt64
$ pacman -S git tar bison flex toolchain make cmake
          boost diffutils libevent double-conversion
          fmt glog dlfcn gtest autotools ncurses libyaml
----



== Packaging

[[root-folder-selection]]
=== Tebako root folder (aka prefix) selection

The Tebako prefix determines the base directory for the Tebako setup.

It is an essential part of configuring how Tebako operates within your system.

The selection of the Tebako prefix follows a specific order of precedence to
ensure flexibility and ease of use:

. *User-specified prefix*:
The most direct way to set the root folder is by specifying it through a
command-line argument.

. *Current Working Directory (PWD)*:
If the prefix option is explicitly set to `PWD`, Tebako uses the current working
directory as Tebako root folder.

. *Environment variable (`TEBAKO_PREFIX`)*:
In the absence of a user-specified option, Tebako looks for an environment
variable named `TEBAKO_PREFIX`. If found, its value is used as the root folder.

. *Default value*:
If no prefix is specified and the `TEBAKO_DIR` environment variable is not set,
Tebako defaults to using a directory named `.tebako` in the user's home
directory.


Path Expansion: Regardless of the method used to set the Tebako prefix, Tebako
expands the provided path to an absolute path. This expansion includes resolving
relative paths based on the current working directory and expanding user
directory shortcuts like `~`.


=== Commands

Tebako provides several commands to manage the packaging and deployment process.

==== Press

This command "presses" a Ruby project using the Tebako components built in the Tebako
root folder (`<tebako-root-folder>`).


[NOTE]
====
The first invocation of the `press` command can take up to an hour as it sets up
the packaging environment and collects the required dependencies. Subsequent
invocations are much faster.
====

Upon the next invocation, Tebako will use previously created packaging
environment. The press process itself takes minutes.

You can manage setup of packaging environment manually; please refer to
description of setup and clean commands below.

[source,sh]
----
tebako press \
  -e|--entry-point=<entry-point> \
  -r|--root=<project-root-folder> \
  [-p|--prefix=<tebako-root-folder>] \
  [-R|--Ruby=<ruby-version>] \
  [-o|--output=<packaged-file-name>] \
  [-l|--log-level=<error|warn|debug|trace>] \
  [-c|--cwd=<package current working directory>]
  [-D|--devmode] \
  [-P|--patchelf] \
  [-m|--mode=<bundle|both|application|runtime>] \
  [-u|--ref=<runtime-reference>] \
  [-t|--tebafile=<path-to-tebafile>]
----

Where:

`<tebako-root-folder>`::
the Tebako root folder (see details: <<root-folder-selection>>)

`Ruby`::
this parameter defines Ruby version that will be packaged (optional, defaults to
`3.2.6`)

`project-root`::
a folder at the host source file system where project files are located.
This parameter is not required if the operation mode is `runtime`.

`entry-point`::
an executable file (binary executable or script) that shall be started when
packaged file is called. This parameter is not required if the operation mode is `runtime`.

`output`::
(optional)
the output file name.
+
Defaults to `<current folder>/<entry point base name>`.
+
On Windows output file will have `exe` extension.
If the application is to be a separate file (`both` or `application` mode), it
will have the same name with the `.tebako` extension.

`log-level`::
logging level for the Tebako built-in memory filesystem driver
(optional, defaults to `error`)

`cwd`::
(optional)
a folder within Tebako memfs where the packaged application will start. This folder should be specified relative to the memfs root.
+
If not provided, the application will start within the current folder of the
host (i.e., at `$PWD`).
+
This argument is required because it is not possible to change the directory to
a memfs folder until the package is started, as opposed to any host folder that
can be set as the current directory before Tebako package invocation.  Tebako
saves the original working directory in a global Ruby variable
`$tebako_original_pwd`.

`devmode`:: flag that activates development mode, in which Tebako's cache and
packaging consistency checks are relaxed.

`patchelf`::
Allows forward-compatibility of Tebako packages with Linux GNU distributions.
+
Specifically, this is a flag that removes a reference to the `GLIBC_PRIVATE`
version of `libpthread` from a Tebako package.  This allows Linux GNU packages
to run against versions of `libpthread` that differ from the version used for
packaging.
+
NOTE: This option only works on GNU Linux only.
+
[example]
For example, a package created at Ubuntu 20 system can be used on Ubuntu 22.
+
NOTE: The feature is exeprimental, we may consider another approach in the future.

`mode`::
output mode for tebako package (optional, defaults to `bundle`).

`bundle`::: place runtime and application to a single file
`both`::: create both run-rime and application
`application`::: create appplication only
`runtime`::: create runtime only

`ref`::
(optional)
Defaults to `tebako-runtime`.
When a Tebako application package is created on Windows, it is linked against a
Tebako runtime file name. The `ref` parameter allows to specify the name of the
runtime file.
+
NOTE: The `ref` option specificies the name of the runtime -- the runtime file
can be recreated or changed but not renamed.

`tebafile`::
the tebako configuration file (optional, defaults to `$PWD/.tebako.yml`).
Please refer to the separate section below for description of the tebafile.
+
NOTE: Development mode is *not intended for production use* and should only be
used during development.
+
NOTE: `entry-point` and `project-root-folder` are required parameters and may be
provided either via command-line or in the tebafile.

[example]
====
[source,sh]
----
tebako press \
  --root='~/projects/myproject' \
  --entry=start.rb \
  --output=/temp/myproject.tebako
----
====

==== Setup

This command sets up the Tebako packaging environment.

Collects required packages, builds the and creates packaging environment. This
is a lengthy task that can take significant time, up to 1 hour.

Tebako supports several configurations at a single system given that their root
directories differ and multiple Ruby versions within single configuration

This command is optional, tebako creates packaging environment automatically
upon the first invocation of press command.

However, if you plan to use tebako in CI/CD environment with caching it is
highly recommended to build cache based on `tebako setup` output. Building cache
based on `tebako press` may create inconsistent environment upon restore.

[source,sh]
----
$ tebako setup \
  [-p|--prefix=<tebako-root-folder>] \
  [-R|--Ruby=<ruby-version>] \
  [-D|--devmode] \
  [-t|--tebafile=<path-to-tebafile>]
----

Where:

`<tebako-root-folder>`:: the Tebako root folder (see details: <<root-folder-selection>>)

`Ruby`:: parameter defines package Ruby version (optional). This version is used in conjuction with requirements
from the Gemfile as explained below in 'Package Ruby version selection rules' section.

`tebafile`::
the tebako configuration file (optional, defaults to `$PWD/.tebako.yml`).
Please refer to the separate section below for tebafile description.

`devmode`:: flag activates development mode, in which Tebako's cache and packaging consistency checks are relaxed.
Please note that this mode is not intended for production use and should only be used during development.

==== Clean

This command cleans up all Tebako artifacts in the specified prefix directory.

NOTE: These artifacts are created by the `setup` and `press` commands.
Normally you do not need to do it since tebako packager optimizes artifacts lifecycle on its own.

[source,sh]
----
$ tebako clean \
  [-p|--prefix=<tebako-root-folder>] \
  [-t|--tebafile=<path-to-tebafile>]
----

Where:

`<tebako-root-folder>`:: the Tebako root folder (see details: <<root-folder-selection>>)

`tebafile`::
the tebako configuration file (optional, defaults to `$PWD/.tebako.yml`).
Please refer to the separate section below for tebafile description.

[example]
====
[source,sh]
----
tebako clean --prefix='~/.tebako'
----
====


==== Clean Ruby

This command cleans up only the Ruby artifacts from the specified prefix
directory.

NOTE: These artifacts are created by the `setup` and `press` commands.
Normally you do not need to do it, since Tebako packager optimizes artifacts
lifecycle on its own.

NOTE: Compiled DwarFS libraries are not cleaned.

[source,sh]
----
$ tebako clean_ruby
  [-p|--prefix=<tebako-root-folder>] \
  [-R|--Ruby=<ruby-version>] \
  [-t|--tebafile=<path-to-tebafile>]

----

Where:

`<tebako-root-folder>`::
the Tebako setup folder (optional, defaults to current folder)

`Ruby`::
defines Ruby version that will cleaned (optional, cleans all versions by default)

`tebafile`::
the tebako configuration file (optional, defaults to `$PWD/.tebako.yml`).
Please refer to the separate section below for tebafile description.

[example]
====
[source,sh]
----
tebako clean_ruby --prefix='~/.tebako'
----
====

==== Build script hash

This command outputs a hash value for the Tebako build script, which can be used
as a cache key in CI/CD pipelines.

[source,sh]
----
$ tebako hash
----

=== Tebako configuration file

It is possible to provide all or some options for the `tebako
{setup | press | clean | clean_ruby}` commands via a Tebako configuration file
('tebafile').

Tebafile is a YAML file with a single key `options`. The options are the same as
long names for the command line.

For example, for the prefix option:

[source]
----
-p|--prefix=<tebako-root-folder>
----
the key in the YAML file would be 'prefix'.

.Example tebafile that sets values for prefix and Ruby options
[example]
====
This is an example tebafile that sets values for prefix and Ruby options:

[source,yaml]
----
options:
  prefix: /tmp/tebako
  Ruby: 3.2.4
----
====

=== Options prefernce order

Tebako supports several methods to set options. The table below show preference order and limitations for specific options.
samller order means higher proirity.

[cols="4", options="header"]
|===
| Order | Mode | Option source | Applicability

| 1 | All | Command-line | All options
| 2 | All |Tebako configuration file | All option except ``--tebafile`` (you can not specify new tebafile in a tebafile)
.2+| 3 .2+| All .2+| Environment variable | TEBAKO_PREFIX to set ``--prefix`` option
| LG_VADDR to set ``--with-lg-vaddr`` jemalloc parameter
.2+| 4 | ``runtime`` | Tebako defaults | All options except ``--entry-point`` and ``--root`` that are mandatory
| ``bundle``, ``both``, ``application`` | Tebako defaults | All options

|===


=== Exit codes

The Tebako CLI exits with different exit codes to indicate the status of the
operation. The following table lists the possible exit codes and their meanings.

.Tebako CLI exit codes
[cols="a,a"]
|===
| Code | Condition

| 0    | No error
| 1    | Invalid command line
| 101  | 'tebako setup' configure step failed
| 102  | 'tebako setup' build step failed
| 103  | 'tebako press' configure step failed
| 104  | 'tebako press' build step failed
| 105  | Failed to map MSys path to Windows
| 106  | Entry point does not exist or is not accessible
| 107  | Project root does not exist or is not accessible
| 108  | Package working directory does not exist
| 109  | Invalid Ruby version format
| 110  | Ruby version is not supported
| 111  | Ruby version is not supported on Windows
| 112  | OS is not supported
| 113  | Path to root shall be absolute. Relative path is not allowed
| 114  | Entry point is not within the project root
| 115  | Failed to load Gemfile
| 116  | Ruby version does not satify Gemfile requirements
|===

== Package Ruby version selection rules

During packaging tebako creates its own Ruby execution environment that is independent from the host Ruby environment.
The version of Ruby that is used in the package is defined by the combination of the `--Ruby` option and ruby requirement
specified in Gemfile.

.Package Ruby version selection Rules
[cols="3", options="header"]
|===
| Gemfile requirement\--Ruby option | Specified | Not specified

| Specified | The version specified by --Ruby option if it is supported and satisfies Gemfile requirement; error otherwise | The minimal supported Ruby option that satisfies Gemfile requirement; error otherwise
| Not specified | The version specified by --Ruby option if it is supported; error otherwise | Default Tebako Ruby version (3.2.6).

|===


== Packaging scenarios with Ruby

Tebako for Ruby supports the following packaging scenarios.

This is high-level description of the Tebako Ruby packaging mechanism.

NOTE: These scenarios were inspired by the `ruby-packer` approach.

NOTE: Tebako Ruby is created independently from `ruby-packer`, no line of code
was copied from `ruby-packer`.

Depending on the configuration files that are present in the root project folder, the Tebako Ruby packager supports different packaging scenarios.

These scenarios differ in what files are packaged and where the entry point is located.

Here is a summary of the scenarios:

[cols="a,2a,4a,3a,a,a,a"]
|===
| Scenario | Description | Packaging | Entry point | `*.gemspec` | `Gemfile`  | `*.gem`

| 1
| Simple ruby script
| Copy `<project-root>` with all sub-folders to packaged filesystem
| `<mount_point>/local/<entry_point base name>`
| No
| No
| No


| 2
| Packaged gem
| Install the gem with `gem install` to packaged filesystem
| `<mount_point>/bin/<entry_point base name>` (i.e., binstub is expected)
| No
| No
| One


| 3
| Gem source, no `bundler`
|
. Build the gem using `gem build` command at the host
. Install it with `gem install` to packaged filesystem

| `<mount_point>/bin/<entry_point base name>` (i.e., binstub is expected)
| One
| No
| Any


| 4
| Gem source, `bundler`
|
. Collect dependencies at the host with `bundle install`
. Build the gem using `gem build` command
. Install it with `gem install` to packaged file system

| `<mount_point>/bin/<entry_point base name>` (i.e., binstub is expected)
| One
| One
| Any


| Error
| Error: Two or more `*.gem` files present
| -
| -
| No
| No
| Two or more


| Error
| Error: Two or more `*.gemspec` files present
| -
| -
| Two or more
| Any
| Any

|===

These scenarios determine how the project is packaged and where the entry point is located within the packaged filesystem.

== Run-time options

=== General

Generally Tebako package passes command line options to the packaged application.

[example]
====
For example, if the package was created with the following command

[source,sh]
----
tebako press \
  --root='~/projects/myproject' \
  --entry=start.rb \
  --output=/temp/myproject.tebako
----
running

[source,sh]
----
/temp/myproject.tebako --option --parameter value
----

will be translated by Tebako bootstrap code to

[source,sh]
----
myproject --option --parameter value
----
====

However there are several command-line parameters that are intercepted processed
by Tebako bootstrap code as described below.

=== Running tebako image by tebako runtime (`--tebako-run` option)

Tebako provides an option to an extract its DwarFS filesystem from a package to
a local folder for verification or execution.

[source,sh]
----
$ <tebako-runtime> --tebako-run [<tebako application>]
----

Where,

`<tebako runtime>`::
The tebako runtime in ```runtime``` or ```both``` mode

`<tebako application>`::
The tebako application package created in ```application``` or ```both``` mode

[example]
====
Creating separate runtime and application and running it:

[source,sh]
----
tebako press -m runtime -o tebako-runtime
tebako press -m application -o tebako-application -e hello.rb -r test
tabako-runtime --tebako-run tebako-application Maxim
----
where hello.rb is the Ruby application
[source,Ruby]
----
# frozen_string_literal: true

puts "Hello, #{ARGV[0]}!"
----
And expected output from ```tabako-runtime --tebako-run tebako-application Maxim``` is
[source,sh]
----
Hello, Maxim!
----

====

=== Image extraction (`--tebako-extract` option)

Tebako provides an option to an extract its DwarFS filesystem from a package to
a local folder for verification or execution.

[source,sh]
----
$ <tebako-packaged-executable> --tebako-extract [<root folder for extracted filesystem>]
----

Where,

`<root folder for extracted filesystem>`::
The root folder for the extracted filesystem (optional, defaults to `source_filesystem`)

[example]
====
Extracting Tebako content from the `metanorma` package:

[source,sh]
----
metanorma --tebako-extract temp-image
----
====

The `--tebako-extract` option actually runs the following Ruby script:

[source,ruby]
----
require 'fileutils'
FileUtils.copy_entry '<in-memory filesystem root>', ARGV[2] || 'source_filesystem'
----

=== Mounting host folder to Tebako memfs (`--tebako-mount` option)

Some programs unconditionally use folders located under the application root,
and when processed by Tebako or similar tools, these folders are included in the
packaging.

[example]
====
Rails, for example, does not provide a configuration option to change where
it expects the `tmp` folder to be.

The location is hardcoded in multiple places within the Rails codebase, residing
under the application root, and as a result, it gets included in the read-only
Tebako memfs.

Although patches have been proposed (e.g.,
https://github.com/rails/rails/issues/39583), there is currently no way to
change the paths for temporary files, caches, and sockets.
====

To address this inevitable limitation for Ruby applications,
Tebako provides an option to mount a host folder to the memfs tree.

When using Tebako, consider the packaging scenario mentioned above, as it
defines the layout of the application tree.

The `--tebako-extract` option may be useful for understanding the placement of
files and folders.

[example]
====
The following command starts a `rails.tebako` package with `$PWD/tmp` mounted as
`local/tmp` in the memfs.

Any remaining command-line parameters are passed to the application.

[source,sh]
----
rails.tebako --tebako-mount local/tmp:$PWD/tmp server
----
====

The `--tebako-mount` option has the following syntax:

[source,sh]
----
--tebako-mount <memfs path>:<host path>
----

The `--tebako-mount` option can be repeated multiple times to mount more than
one object. The `memfs path` is relative to the memfs root, and it is
recommended to use absolute paths for host objects. Both directories and files
can be mounted in this way. Tebako allows overlaying existing memfs objects, so
there are no significant limitations.

== Trivia: origin of name

"tamatebako" (玉手箱) is the treasure box given to Urashima Taro in the Ryugu,
for which he was asked not to open if he wished to return. He opened the box
upon the shock from his return that three hundred years has passed. Apparently
what was stored in the box was his age.

This packager was made to store Ruby and its gems, and therefore named after
the said treasure box (storing gems inside a treasure box).

Since "tamatebako" is rather long for the non-Japanese speaker, we use "tebako"
(手箱, also "tehako") instead, the generic term for a personal box.

== Contributing

We welcome contributions! Please see our contributing guidelines for more
information.

== License

Copyright Ribose. All rights reserved.

Tebako is released under the BSD 2-Clause License. See the LICENSE file for details.