Download the tester image

This is a tester disk image, not an installable ISO and not a signed release. Write the decompressed raw to a USB stick, or boot it as a VM disk. A tester who boots an install ISO from a stick is offered a wipe of their internal disk — qdistro does not ship that ISO.

Signing is post-v1. There is no release signature and no signed manifest for this file. Integrity is the sibling .sha256:

sha256sum -c qdistro-0.1.0-20260902.raw.xz.sha256

Use the names that actually arrived; the 0.1.0-20260902 token is the current kiwi version and snapshot pin, not a promise that a file of that name is hosted here.

Get the file

The intended public host is a GitHub Release asset on github.com/qdistro/qdistro once a tester cut is published. No public download is published yet. Until it is, the bytes come from a kiwi build:

cd qdistro/image
QDISTRO_PROFILE=dev ./build-in-vm.sh
# bundle/qdistro-<version>-<snapshot>.raw.xz
# bundle/qdistro-<version>-<snapshot>.raw.xz.sha256

build-in-vm.sh needs a libvirt session; see image/AGENTS.md. Do not dd a file whose checksum does not match.

Write it to a stick

Identify the stick (lsblk). The of= target is the whole disk, not a partition. This destroys everything on that device.

xzcat qdistro-0.1.0-20260902.raw.xz | sudo dd of=/dev/sdX bs=4M conv=fsync status=progress

Minimum 32 GB. USB 3 or an external SSD. First boot grows the btrfs root onto leftover space (oem-resize on every boot; harmless when there is nothing to grow).

Boot it as a VM

Decompress, then attach the .raw as the VM disk under OVMF / UEFI. Legacy BIOS will not boot. A qcow2 overlay on the raw is the usual shape (that is what image/verify.sh does) so the published bytes stay read-only:

unxz -k qdistro-0.1.0-20260902.raw.xz
qemu-img create -f qcow2 -F raw -b qdistro-0.1.0-20260902.raw overlay.qcow2
# virt-manager: firmware = UEFI (OVMF), disk = overlay.qcow2

Give the guest a few GiB of RAM. Nested KVM is required only for the experimental tier-4/5 path.

First login

greetd presents qdgreeter on tty3. Log in as admin / qdistro. sshd is off; do not expect a network login.

What this image is, and is not

VM-verified: first-boot grow, UUID identity, USB-bus boot, OpenSUSE Secure Boot keys, persist across reboot, greeter login. A real USB stick on real hardware (Secure Boot on, WPA2/WPA3, silos, reboot) is not yet recorded.

Honest limits, so a first boot is not a surprise:

• Not a signed v1. Shared default password, dev profile, SELinux permissive on this image.
• Approvals are local. The broker is installed with its approval paths — the graphical approval-queue app, the qdistro-approvals CLI, and the admin TUI. Grants come only from local admin control-plane peers (those tools, or an admin/root D-Bus client such as busctl); sshd is off on this image, so there is no remote path.
• No isolated print VM. A host print proxy is installed; CUPS in a contained VM is not.
• No installable ISO. This file is a disk image. It is the system, not an installer that targets another disk.
• Tiers 4/5 experimental. The chain installs host launch code only. Build a guest image on the booted system if you have KVM (qdistro-bootstrap.sh --tier4-base).
• No tier-3s paravirt. The experimental gVisor application-kernel mode is an opt-in install (QDISTRO_TIER3S=1 on install-session-manager.sh) and runsc is provisioned on demand — the launch machinery is not on this image (a CreateTier3sSilo call would succeed on this dev-profile image; StartSilo fails without the launch path). Launches are dev-profile only, network=none, and explicit; no KVM needed.
• Polkit agent cannot grant as uid 1000; that path fails closed.

From-source installs (Tumbleweed + bootstrap --profile=dev) stay on the homepage for developers iterating the tree.