Debian Museum

• Releases
• Containers

Important: All container images will be migrated to the tar+zstd format.

Older versions of docker do not support tar+zstd!

docker.io installed with apt is still stuck at v20.10, you may need to install docker-ce manually.

History

version	release
0.01 ~ 0.90	1993-08 ~ 1993-12
0.91	1994-01
0.93R5	1995-03
0.93R6	1995-10-26

---

version	codename	created	release	eol
1.1	Buzz	1993-08-16	1996-06-17	1997-06-05
1.2	Rex	1996-06-17	1996-12-12	1998-06-05
1.3	Bo	1996-12-12	1997-06-05	1999-03-09
2.0	Hamm	1997-06-05	1998-07-24	2000-03-09
2.1	Slink	1998-07-24	1999-03-09	2000-10-30
2.2	Potato	1999-03-09	2000-08-15	2003-06-30
3.0	Woody	2000-08-15	2002-07-19	2006-06-30
3.1	Sarge	2002-07-19	2005-06-06	2008-03-31
4.0	Etch	2005-06-06	2007-04-08	2010-02-15
5.0	Lenny	2007-04-08	2009-02-14	2012-02-06

---

version	codename	created	release	eol	eol-lts	eol-elts
6.0	Squeeze	2009-02-14	2011-02-06	2014-05-31	2016-02-29
7	Wheezy	2011-02-06	2013-05-04	2016-04-25	2018-05-31	2020-06-30
8	Jessie	2013-05-04	2015-04-26	2018-06-17	2020-06-30	2025-06-30
9	Stretch	2015-04-26	2017-06-17	2020-07-18	2022-06-30	2027-06-30
10	Buster	2017-06-17	2019-07-06	2022-09-10	2024-06-30	2029-06-30
11	Bullseye	2019-07-06	2021-08-14	2024-08-14	2026-08-31	2031-06-30
12	Bookworm	2021-08-14	2023-06-10	2026-06-10	2028-06-30	2033-06-30

---

version	codename	created	release
13	Trixie	2023-06-10	2025?
14	Forky	2025-08?	2027?

---

version	codename
unstable	Sid
experimental	RC-Buggy

Unlike the Debian Releases unstable and testing, experimental isn't a complete distribution, it can work only as an extension of unstable.

To enable experimental source, run a sid container and then edit the /etc/apt/sources.list.d/mirror.sources file in the container.

---

See also:

• Ubuntu Museum
• Debian Project History
• distro-info-data/debian.csv
• Wiki/DebianTesting

Virtual Machine

Unlike containers, virtual machine images only support a few newer versions such as jessie, stretch, buster, bullseye, bookworm, and sid.

The primary reason for this is that testing different architectures and versions of virtual machine images is quite exhausting!

Container images even support versions from the last century.

Please feel free to send an issue to tell us which version you are using and which architecture.

DO NOT use PROFANITY or DISCRIMINATORY language in issues!

We're going to probably increase support for the version and the architecture.

Although this isn't the best virtual machine project, nor the most comprehensive, for me, it's the best nogui virtual machine project!

Why not give it a try? You can experience virtual machines for riscv64gc, loongarch64, mipsle, s390x, and other architectures in just a few minutes.

Step0: download the vm image

These VM images are built automatically by github actions.

Go to the Releases page and download file that start with vm-.

Files starting with version numbers are container images (e.g., 8_jessie_), not virtual machine images.

Now let's take bookworm x64 (x86_64) as an example.

mkdir -p vm/12/x64
cd vm/12

url='https://github.com/2cd/debian-museum/releases/download/12/vm-minimal_bookworm_x64.tar.zst'
# If you don't have aria2 installed, then curl will be used.
aria2c -s5 -x5 "$url" || curl -LO "$url"

cd x64
tar -xvf ../vm-minimal_bookworm_x64.tar.zst

Step1: install qemu

Unlike other software, where older versions are installed to maintain stability, for qemu, it is recommended to install a newer version (such as 8.2+) instead of the old one.

Alpine

# If the VM is x64 arch:
apk add qemu-system-x86_64

# If the VM is arm64 arch:
apk add qemu-system-aarch64

# If the VM is rv64 arch:
apk add qemu-system-riscv64

# If the VM is loong64 arch:
apk add qemu-system-loongarch64

Android Termux

# If the VM is x64 arch:
pkg i qemu-system-x86-64-headless

# If the VM is arm64 arch:
pkg i qemu-system-aarch64-headless

ArchLinux

# If the VM is x64 arch:
paru -Sy qemu-system-x86

# If the VM is arm64 arch:
paru -S qemu-system-aarch64

Debian, Ubuntu, Kali, Mint, ...

# run apt as root (e.g., sudo-rs apt update)
apt update
apt install qemu-system

There are other distributions that we haven’t listed one by one. For example, on Fedora, the package name for x64 QEMU is qemu-system-x86. You can use the package manager that comes with your distribution to search for qemu-system and then install the corresponding architecture’s QEMU.

Feel free to explore and set up QEMU based on your distribution!

Step2: fix KVM permissions (Optional)

Please note: Linux Kernel 5.7+ no longer supports arm32 KVM. If you’re using an armv7l device, you can skip this part.

See also phoronix/Linux-5.7-Kill-32-bit-ARM-KVM

Unfortunately, if your system kernel and CPU don’t support KVM or other acceleration schemes (e.g., xen), virtual machine performance will be quite slow. Some people humorously say that QEMU isn’t Quick at all; it’s more like Slow EMU. However, this is a misconception.

For instance:

On devices that support x64 KVM, using a Debian Sid x64 minimal virtual machine, the boot time is only around 4 seconds. But the same image on Android arm64 takes over a minute to boot. Nevertheless, this is already quite fast and comparable to Alpine Linux.

I’ve experienced running Windows arm64 on a device without arm64 KVM support, and it is not an exaggeration to say that you need to wait for hours.

---

Add the current user to the kvm user group.

# run it as root (i.e., +sudo/+sudo-rs/+doas)
usermod -aG kvm "$(id -un)"

If it does not work, change he permissions manually.

# run it as root
chmod 666 -v /dev/kvm

Step3: install zsh

On most distributions, the package name for zsh is zsh, and you can install it using your system's package manager. e.g., apt install zsh

For Linux systems other than Android and embedded distributions, you can also download it from the 2moe/zsh-static-docker repository. (Extract tar.zst and get opt/bin/zsh, finally move it to ${XDG_BIN_HOME:~/.local/bin})

It can also be another "PATH", not necessarily XDG_BIN_HOME.

Step4: expand virtual disk (Optional)

Install qemu-utils, then run qemu-img.

If you have enough disk space then +2G can be changed to +10G

qemu-img resize disk.img +2G

If unsafe-resize-disk.service (systemd) works correctly, it will automatically resize the partitions to utilize the unallocated space.

The disk partition table is automatically backed up to /var/lib/tmm/bak/sfdisk/ before resizing.

Step5: run

This is a Zsh script that doesn’t require root permissions and won’t automatically create files outside of $PWD or ${0:a:h} (i.e., current_exe()?.parent()) directories.

It’s a bit complex, and in the future, it might be combined with WASI (WebAssembly System Interface) to handle the core logic.

./run || zsh run

localhost login: root

Step6: SSH (Optional)

install OpenSSH client

The ssh package name for debian/ubuntu is ssh, or openssh in some distributions.

If you don't need sshd, install the ssh client separately.

Open a new terminal session/window, then run:

# run apt as root (i.e., +sudo/+sudo-rs/+doas)
apt install openssh-client

Run it on your host, not VM.

Then check the ./connect-to-ssh script.

SSH is a very important thing. Before running related scripts, you have to make sure there is no malicious code inside!

No matter what, you need to be cautious about this.

cat connect-to-ssh

Finally, connect to SSH.

./connect-to-ssh || zsh connect-to-ssh

Docker

The following content is prepared for Docker users.

If you want to experience Docker on devices that do not support the Docker kernel, you can use vm-dev_bookworm_x64.tar.zst or vm-dev_bookworm_arm64.tar.zst.

The Dev VM comes pre-installed with Docker, qemu-user-static, and systemd-nspawn.

---

• RUN IT ON zsh.
  • bash is NOT SUPPORTED, NOR is dash
• Just change the values of ver & arch
• What follows may seem complicated, but it's actually quite simple to follow step-by-step.
• Depends:
  • docker-ce | docker-ee
  • zsh (>= 5)
• Recommends:
  • qemu-user-static | qemu-user-static-binfmt

setopt interactive_comments

# On modern Linux distro, running very very old debian containers can be problematic! (e.g., host: 12(bookworm, kernel: 6.x), container: 3.1(sarge), host-arch: amd64, container-arch: amd64)
#
# versions: 8, 9, 10, 11, 12, 13, sid
ver=sid

# architectures: "", "riscv64", "rv64", "amd64", "x86_64", "x64", "arm64", "aarch64", "i386", "loong64", "armhf"
# The architectures supported by different versions are not exactly the same.
# For example, trixie(13) no longer supports mipsle (a.k.a., MIPS 32-bit Little Endian),
# sid(unstable) supports more architectures than stable.
arch=""

# --------------------------
dbg() {
    print >&2 -Pr "%F{blue}[DEBUG]%f $*"
}
# if ver.is_empty()
if ((! $#ver)) {
    ver=sid
}

# Debian sid supports a very large number of architectures, not all of which are listed here.
rv64=riscv64
x86=386
x64=amd64
aa64=arm64
local -A oci_arch_map=(
    # riscv64imafdc
    rv64gc       $rv64
    riscv64      $rv64
    rv64         $rv64
    #
    x64          $x64
    amd64        $x64
    x86_64       $x64
    #
    aarch64      $aa64
    arm64        $aa64
    #
    loong64      loong64
    loongarch64  loong64
    # i386 ~ i686
    x86          $x86
    i386         $x86
    i486         $x86
    i586         $x86
    i686         $x86
    #
    armhf        armv7
    # In fact, just relying on `uname -m` cannot determine `feature = "+vfp3"` (armhf).
    # armv7l       $armv7
    #
    # armv8m is arm32, not arm64
    armv8m       armv7
)

# On alpine, if dpkg is installed, it will output musl-linux-[xxx] (e.g., musl-linux-riscv64), not [xxx] (e.g., riscv64).
# For busybox, the dpkg it contains is a lite version. Therefore, the value of `deb_arch` may be empty.
# If `deb_arch.is_empty()`, use `uname`, not `dpkg`.
get_dpkg_architecture() {
    local dpkg_arch=$(dpkg --print-architecture)

    # dpkg_arch.split('-').last()
    local deb_arch=$dpkg_arch[(ws^-^)-1]

    case $deb_arch {
        ("") uname -m        ;;
        (*)  print $deb_arch ;;
    }
}

# if arch.is_empty()
if ((! $#arch)) {
    # arch = if "dpkg".cmd_exists() { dpkg --print-architecture } else { uname -m }
    arch=$(
        if (($+commands[dpkg])) {
            get_dpkg_architecture
        } else {
            uname -m
        }
    )
}

args=(
    # Run a new container
    run

    # Pull image before running ("always"|"missing"|"never") (default "missing")
    # --pull  always

    # Automatically remove the container when it exits
    --rm

    # ( -i ) Keep STDIN open even if not attached
    --interactive

    # ( -t ) Allocate a pseudo-TTY
    --tty

    # Set environment variables
    --env
    # LANG=?, e.g., C.UTF-8, en_US.UTF-8
    LANG=$LANG
)

# Set timezone env.
#
# if "timedatectl".cmd_exists()
if (($+commands[timedatectl])) {
    # args.push(docker_tz_env)
    args+=(
        --env
        # TZ=?, e.g., UTC, Asia/[CITY], Europe/[CITY]
        TZ=$(timedatectl show --property=Timezone --value)
    )
}

# map: oci_arch_map
#   key: arch
#   value => platform
platform=$oci_arch_map[$arch]

# if platform.is_not_empty()
if (($#platform)) {
    args+=(
        # If you want to run containers from other architectures
        # (e.g., host: arm64, container: riscv64),
        # you need to install `qemu-user-static`.
        --platform  linux/$platform
    )
}

case $ver {
    (sid|unstable) is_sid=true  ;;
    (*)            is_sid=false ;;
}

is_loongarch=false
case $platform {
    # Due to the fact that older versions of Debian (such as Buster)
    # do not support RISC-V 64-bit architecture, it is defined as "sid" here.
    # However, this is not accurate.
    # A more reasonable approach would be to create a `HashMap<version_name, arch_set>`
    # that corresponds to different versions and architectures, and then
    # make the determination based on that.
    (*/riscv64) is_sid=true ;;
    (*/loong64)
        is_sid=true
        is_loongarch=true
    ;;
}

repo="ghcr.io/2cd/debian"

if {$is_sid} {
    repo+=-sid
}
if {$is_loongarch} {
    repo+=:loong64
}
dbg repo: $repo

args+=$repo
dbg args: $args

docker $args

TODO

• +Debian GNU/Hurd VM (a.k.a., Virtual Machine) image

• +Debian kFreeBSD VM image

• +ia64 (a.k.a., Intel Itanium architecture) stage1 container image

  • Since modern qemu does not support emulation of the ia64, there is only stage1 and no full rootfs.