AUSTIN (December 20, 2023) — The Adélie Linux distribution recently celebrated its fifth beta release and is making steady progress toward a 1.0 general release. In the 19 months since the 2022 status update, we have stabilized the project to a point where we can onboard developers, rebuild and test project-wide changes, and cut releases with ease.
We are committed to honesty, integrity, and transparency. While this may be a departure from social or business norms, we believe it is important to provide accurate (if candid) information about the state of the project. Consider whether/how other projects do this.
I would like to set realistic expectations. We cannot make service or development guarantees given our limited resources at this point in time. We renamed the long-awaited RC3 to BETA5 because we believe it more accurately reflects the nature of the release.
New Release: 1.0-BETA5
The 1.0-BETA5 release was an important milestone for the project. It’s a tangible product and platform that can be installed on real (or virtual) computers to give real people a viable alternative to other Linux distributions. It’s also our most solid and polished release to date.
Without a way to easily install Adélie, we were inaccessible to all but the most determined hackers. Installable media was an essential milestone because we are now accessible to the general public, meaning critically, potential contributors (users, testers, and developers).
From an integration standpoint, this release proved that our source tree is again stable.
This release fills a few notable holes from the RC2 release: Xfce live media, aarch64 installation media, a full set of armv7 package binaries and installation media, bundled package documentation, and many improvements to our guided installer.
We’ve rolled out some new light and dark themes for some of our desktop environments, which propagates some of our branding and provides a more pleasant experience.
Given industry trends and the popularity of containers and adjacent technologies, we now provide multi-platform Docker images, which are fully compatible with Podman.
Docker sponsors our membership to the Docker-Sponsored Open Source Program, which gives us another tool for vulnerability analysis and other features that mainly benefit developers, including rate limiting removal and usage statistics.
We will soon release pure OCI-compliant images, as well as upstream some patches to the Proxmox Virtual Environment platform, so that Adélie can be a first class citizen in popular self-hosted resource management systems.
New Initiatives
We are aiming to tighten our release schedule to occur at quarterly (3-month) intervals, which is complementary to our effort to publish nightly development images.
With BETA5 out the door, we are turning our attention to focus on updating key components of our platform, namely compiler toolchains like GCC and Rust, for cases where upstreams add modern Rust as a dependency or transition to C++20.
For BETA6, we are aiming for the following milestones:
- Fix most (or all) of the known G3- and G4- boot problems
- Fix outstanding Horizon (installer) bugs, particularly with Mac partitioning
- Improve compatibility/integration with VirtualBox and other hypervisors
- Bootstrap Java fully (we currently rely on old binaries from RC2)
- Rust toolchain update to current
- Fix test suites, including xvfb, so more KDE components can run their tests
- Update Qt to the latest 5.15 patchlevel and KDE to latest 5 branch
- Support booting of 32-bit ARM media
For BETA7:
- musl 1.2.4 (or newer, if there is another release)
For BETA8:
- GCC 13 (or newer, depending on requirements)
- OpenSSL 3.x
- Boot performance tuning
- Zig
These are just a few highlights and should not be considered exhaustive. You can follow some more of our ongoing initiatives here.
New Tools
At the time of writing the 2022 status update, while we had been making amazing and rapid progress, we had not yet successfully built the source tree for a single target architecture since before my time as project lead. The last time that this happened was around three years ago for the RC2 release, with a gamut of manual processes and less automation.
We are designing and implementing new tools and automatic processes to address this. Naturally, we will be using Adélie containers for some of this infrastructure.
For example, we’ve developed a new tool to build the complete matrix of ISO and tarball media with a single click. This tool can be operated from just one machine of any CPU architecture (using emulation), rather than one machine of each architecture as before.
This is the same tool that produces our nightly development images, so there won’t be any surprises or discrepancies between testing and release media. It also means anyone can reproduce our media at home, no fuss, if they’d prefer to customize/bake it themselves.
To further assist with reproducibility and stability, we have implemented a “snapshot” tool to capture the state of our binary package mirror, and a corresponding service to make the read-only, point-in-time and tag-based snapshot data accessible to you.
It uses btrfs under the hood and is considered an experimental service for now, as it remains under active development while we integrate with other ongoing initiatives.
This will be an important step toward automating developer-facing infrastructure, including pipelines and various endpoints (such as a PPA-inspired platform), to accelerate developer onboarding and productivity. We anticipate production-facing benefits to be a side-effect.
The idea is to layer your /etc/apk/repositories
file with URLs that point to immutable snapshots of our repositories, plus any personal repositories you may be working on.
This offers an air-tight solution to package version or repository pinning, a potential issue for reproducibility and stability in contexts such as containers and debugging.
Porting Adélie to other CPU architectures requires a serious investment of time and attention to detail. The amount of work varies from “a lot” to “holy penguin” depending on which components (i.e. compilers, kernel, package compatibility, etc.) need to be modified.
Zach is leading an internal effort to develop a tool to “bootstrap” a minimal Adélie environment, sufficient to begin building packages, with minimal human intervention.
New Hardware
ARM is no longer a bottleneck for development, testing, or releasing!
We now host a tricked-out Wiwynn SV328R, on “permanent loan” from one of our donors! This beast, with Ampere Altra CPUs, is able to build packages faster than our non-ARM builders. It replaces our LX2160A from SolidRun, which is now a development machine.
We were also granted “permanent” access to two remotely-hosted Ampere eMag 8180 machines, courtesy of OSUOSL. These are for non-production use and internal ARM testing.
SPARC is on the official roadmap!
Several months ago, we began sponsoring and hosting a shared SPARC T4-2 machine for the Compile Farm. We were excited to continue working on our musl sparc64 port, which is a requisite step toward bringing up Adélie on SPARC. This remains a lower priority effort.
Due to a hardware failure, we are replacing this machine with a T5-2 loaded with 512GB of memory and making it available to projects like Compile Farm and Debian as before.
We hope that this public infrastructure will continue to be useful to all sorts of upstream projects; we are routinely being asked to “hurry up” because of various bugs/regressions.
RISC-V is on the official roadmap!
While we had been playing with a HiFive Unmatched board for our riscv64 port, it leaves much to be desired in terms of performance. Poof, an olive branch appears!
RISC-V International is donating two Milk-V Pioneer boards, each with 64-cores and 128GB memory, to our project and two more to Compile Farm. We will manage and host the latter.
Our colocation facility has been expanded to accommodate some of this new hardware.
A kind Internet stranger has subsidized nearly the full cost of a CN7800-EVB development board, which features at least two dozen Cavium OCTEON III MIPS64 cores.
Marvell has generously provided access to their SDK, but we do not currently have the spare time to prioritize any MIPS-related work (the SDK is required to bring up the board).
We will be replacing some of our network equipment in the coming weeks to enable better (read: easier) management and control over our internal network topology. Bye, Cisco!
Data center costs across the country have been rising rapidly over the past few years, and we are not immune to this. This is well in excess of a typical 10% price hike cap and is fueled by a rise in demand for hosting spurred by artificial intelligence companies.
We do not yet have enough income to cover all of our expenses, but we are close. We are in need of financial contributions, but the situation is not dire in terms of remaining online.
One of our sponsors of off-site infrastructure is shutting down within a few months. They have shared that their direct costs have increased by 15%, 10%, and 18% in Q1, Q2, and Q3 of this year, respectively, for a YoY increase of 53%! We will lose a U.S. mirror server.
On a more positive note, we have a tentative agreement with OSUOSL to host some of our development-only machines. If this works out, we should have a sustainable path for future growth if finances continue to be tight. Production systems will remain in Austin, Texas.
To wrap up hardware news, A. Wilcox is working on a “Retro Lab” for vintage platforms that we actively support. It is dubbed the “Tulsa Platform Compatibility Lab,” and will serve as a testing lab. We are always looking for older Apple and Pentium machines.
New Partners
Spack, the package manager for supercomputers, is highly receptive to bug fixes and nicely models the type of upstream-downstream communication that helps projects thrive.
They do not currently target any big-endian platforms, nor do they have hardware for building or testing it. We are working with the Spack project to set up a big-endian system, which we’ll host, in exchange for their giving some attention to finding and fixing issues with their upstream projects. This could unlock new sources of funding.
Patches would be upstreamed or shared with other distributions. It’s a team effort.
As a community engagement incentive, we are offering financial rewards (bounties) for certain types of issues that one might encounter while testing the Adélie platform. If you might be interested, get in touch first. This program will be formally announced with 1.0.
New Year
We look forward to 2024 and hope to address several important housekeeping needs.
Documentation continues to be a sore spot for our users and we’re making progress to create, improve, and reorganize it. Moving forward, we’ll be migrating most of our documentation to GitLab. In some cases, it will make sense to choose another medium.
You can also look forward to a manual pages platform like that of Arch Linux. This will improve user and developer experience in furtherance of our mission.
We will be introducing a “Kernels” repository, intended to provide alternatives to our stable kernel, to support Apple Silicon hardware, offer mainline features, or better performance.
Other noteworthy items include:
- Formalization of project roadmap and issue tracking/management
- Formalization of CI, development and release pipelines, and internal automation
- Accessibility initiative, as mentioned in the 2022 status update
- Up-to-date OpenJDK
- Up-to-date Rust
- Up-to-date LibreOffice
- Up-to-date Firefox and Thunderbird
- Updates to gcompat to support more software, and tools to help with testing
- Out-of-the-box support for ARM SBCs (Raspberry Pi, ODROID, etc.)
New You, New U
Adélie insists on correctness. This is baked into our DNA and is a core tenet of the project. A smooth user experience is not defined by standards, but by subjective feedback.
In regards to our Desktop and Server campaigns, we hope 2024 will be the Adélie Year of Graphics. This means gaming, media, VFX, drivers, and the like.
On behalf of the Adélie family, thank you for another wonderful year!