#linux
5034 posts · Last used 1m
Talos Linux
Link: https://www.siderolabs.com/talos-linux
Discussion: https://news.ycombinator.com/item?id=50543343
#linux
Every November I go through photos from the past year and select the best ones for a hardcover photo album that I gift to my family (husband, parents, in-laws). I've been doing this since the birth of my child.
I may create a second album for myself about my tech journey: I found many pics documenting the process of jailbreaking my Kindles, doing mesh radio experiments and tweaking my VPS with CLI. This week: getting my old MacBook Air ready for #Linux.
✨ Loving every step of this journey ✨
Two Independent Projects Want to Bring MATE to Wayland, Each in its Own Way
The big Linux desktop environments have largely made up their minds about Wayland.
GNOME finished removing X11 from its login manager in March, and last month, KDE Plasma retired support for starting X11 sessions. Cinnamon is a step behind, but Linux Mint 27, due around Christmas, is set to ship a Wayland session that is no longer experimental. Though X11 stays the default.
Smaller desktops are at different points on the same journey.
LXQt dropped the experimental label from its Wayland session in v2.3.0, and Budgie in the Ubuntu Budgie 26.04 release went Wayland-only with labwc.
MATE, on the other hand, has moved in bits and pieces. Since MATE 1.26, apps like Atril, System Monitor, Pluma, and Terminal can run on Wayland, while the official mate-wayland-session is a simple, experimental setup that relies on Wayfire.
What’s still missing is a Wayland-native window manager for MATE. Its own, Marco, has no Wayland mode, and the following two projects, Polo and MATE Wayland Desktop, now aim to replace it in different ways.
Polo
Polo comes from developer Kerem Söke, and it is a compositor, window manager, and session manager for people who want MATE to look and behave on Wayland the way it does today.
It is written in Vala and C on top of libweston 16, follows Marco’s window management rules, and retains the settings, themes, and keybindings MATE users already have. Apps that still need X11 get a window manager of their own under XWayland.
Polo handles the session too, using the D-Bus interface of mate-session-manager, and the panel, Caja, and the screensaver can talk to it as they would on X11.
It’s first stable release was published on 12 November with support for tiling, workspaces, screen sharing through xdg-desktop-portal, and layer shell compatibility for mate-panel.
A DRM backend is responsible for handling multiple monitors and remembering screen layouts. Effects like wobbly windows and a desktop cube are also included (fans of Compiz, rejoice!).
You can build Polo from source on Fedora 44 or 45, Debian 14, or Ubuntu 26.04 or 26.10. Once installed, a “MATE (Polo)” entry should show up in LightDM.
Kerem has also said that he may host his own .deb repository if the Debian and Ubuntu MATE teams do not ship Wayland-enabled packages.
MATE Wayland Desktop
MATE Wayland Desktop uses an existing compositor instead of writing one. It swaps Marco for labwc and ships as an installable ISO built on Ubuntu 26.04 LTS, with Linux 7.0 and MATE 1.29 components as the foundations.
Eight of those MATE packages are rebuilt from newer upstream code, and one of them is mate-panel 1.29.2 with Wayland support turned on. Apps that only speak X11 run through XWayland.
Installation is handled by Calamares, and it offers optional LUKS2 full-disk encryption.
The project is also clear about the rough edges in it.
The keyboard, mouse, and keyboard shortcuts panels in Control Center crash when running natively, so MATE Wayland Desktop opens them through XWayland instead.
MATE’s own Displays dialog does not work on Wayland either. The “Displays” entry in the launcher opens wdisplays instead.
During installation and usage, Secure Boot has to be off, and the ISO boots up with the open source Nouveau GPU driver. NVIDIA’s own driver can be installed afterward with ubuntu-drivers.
Seeing that this is a new offering, I highly recommend you test it on a spare machine before going for a full installation.
You can get started with its ISO on x86_64 systems by heading over to the official download portal.
Canonical is Now Shipping AI-Written Security Patches to Ubuntu Pro Users
If you’ve followed security news this year, you know the number of reported vulnerabilities has gone up sharply. AI-assisted bug hunting means flaws are being found faster than maintainers can patch them, and every distro is feeling the pressure.
Canonical’s answer is to fight fire with fire. They are going to use AI to fix what AI is finding. At the recently concluded Ubuntu Summit 26.10, Canonical introduced Surge, a multi-agent AI system that triages vulnerabilities and writes backported fixes for Ubuntu packages.
Don’t worry. None of this goes into your regular Ubuntu security updates automatically. Surge’s fixes go to a new opt-in repository called the Edge archive, and only reach the normal updates after human-led testing.
The CVE problem Canonical is trying to solve
The announcement came in the security segment of the Summit’s opening remarks, presented by Pierre Guillemin, VP Engineering.
He started with numbers. According to the slides, the “fix velocity” in Ubuntu increased by 119% in one year. The “fix volume by quarter” chart shows roughly 1,400 fixes in early 2025 and about 3,100 a quarter by mid-2026.
That’s not it. Canonical says October 2026 alone had almost as many CVEs as all of 2025.
October 2026 had almost as many CVEs as all of 2025
And Canonical doesn’t expect this to slow down. Guillemin said the company expects 2027 to bring “another order of magnitude growth” in CVEs.
How do you keep up with this pace while sticking to the review and testing standards that stable Ubuntu releases are known for? Hiring and working harder cannot practically compete with this pace.
Meet Surge
Surge is “a multi-agent orchestration framework for remediating vulnerabilities in Ubuntu.”
In simpler words, it’s a pipeline where several specialised AI agents handle different steps of fixing a CVE, from finding out whether Ubuntu is affected to producing a patch.
Surge is a multi-agent orchestration framework
Guillemin says it builds on Canonical’s 20+ years of Ubuntu security work. The stated goal is to absorb a big influx of CVEs without creating backlogs. Or, as he put it, “fight AI with AI.”
Surge only covers Ubuntu userspace packages, meaning the applications and libraries in the archive. Kernel CVEs are not part of it, as kernel fixes need a different backporting approach and follow their own release cycle.
How does Surge work?
The pipeline has six stages, and the interesting part is that it doesn’t trust a single AI model.
Surge pipeline slide at Ubuntu Summit 26.10
It starts with intake and sourcing, which takes in vulnerability reports, maps them to packages, and checks whether the affected package is in Ubuntu at all. Then code forensics fetches the upstream patch, analyses the source, and tries to reproduce the vulnerability to confirm Ubuntu’s version is actually affected.
In the backporting engine, instead of one model writing the patch, several competing models each produce a candidate fix.
These candidates then go to consensus judging, where a separate panel of “judge” models evaluates them and has to agree on the right patch. If confidence is low or the judges deadlock, the case goes to human escalation, where Ubuntu security engineers review it manually.
Finally, the patch goes through the same build, verify and publish process as any other Ubuntu package, including deterministic tests, before it is published.
The Edge archive: AI fixes, but only if you ask for them
Surge-generated fixes are not pushed to your system through the usual security updates. Instead, Canonical is launching the Edge archive, an opt-in preview of fixes that are still in development.
Edge gives you machine-generated patches before they’ve gone through Canonical’s full qualification for general availability (GA). So you get fixes at “machine speed,” but with the compromise on the fact they haven’t had the full testing yet.
Canonical stresses that this is not a change in its security policy. It “will not automatically upload Surge-generated fixes to our stable releases.” Your regular security updates stay human-validated and keep coming through the usual channels.
Will you opt-in?
If you’re a regular Ubuntu desktop or server user, nothing changes unless you choose to. Your security updates continue to arrive the way they always have.
If you want to try it, Edge is available through Ubuntu Pro, which remains free for personal use. Canonical has published a guide on how to enable the Edge archive.
I wouldn’t put Edge on a production server, yet. But on a regular desktop machine or a homelab, it’s a good way to help Canonical find problems early.
To be honest, the idea of AI writing security patches will make some of you uneasy, and that’s fair. But we have to “fight AI with AI” because CVEs are not going to stop or reduced. Am I right?
(ignore skirt and socks) Mobian running natively on Pixel 3 with NuraOS kernel+Mobian
Not usable, it can detect modem and sim card but can’t pick up service, all sensors are not working touch stop responding if left in lock screen for 2h or more also no cameras (front work in NuraOS), GPU, speakers, buttons are working fine
AV Linux 25.3 for Content Creators Returns to Xfce by Default
AV Linux 25.3, a Debian-based Linux distro for multimedia content creators, has been released, marking the return of Xfce as a fully integrated desktop environment alongside Enlightenment. AV Linux now boots into Xfce by default, but users can switch to Enlightenment from the LightDM login screen.
According to developer Glen MacArthur, the decision to bring back Xfce was driven by user feedback. Some users were reluctant to adopt Enlightenment after the distribution’s earlier desktop transition, and the new release offers both environments without abandoning the existing Enlightenment integration.
The two desktops share ConnMan for network connectivity, the Xfce application menu, similar panel layouts, and matching keyboard shortcuts. Moreover, a new Mosaic theme provides a consistent appearance across both environments.
Another change is the introduction of a dedicated package repository, replacing the previous web-based package archive. It is divided into three components: Desktop Enlightenment for AV Linux themes and configurations, Desktop Moksha for MX Moksha-specific packages, and Main for multimedia applications and plugins shared by both distributions.
The release also introduces in-house packaging of the Enlightenment stack, including EFL 1.29.0, Enlightenment 0.28, and associated applications and modules. Updated applications include the Evisum system monitor, Elaro IRC client, Eclair music player, and Equation calculator. MX Moksha, meanwhile, receives Moksha 4.2, along with fixes for suspend and hibernate functionality.
Under the hood, AV Linux 25.3 ships with the Liquorix kernel 7.3.4. The updated multimedia software selection includes Ardour 9.10, REAPER 7.85 as a demo, Audacity 4.0.2, and Cinelerra-GG. New audio plugins, including Pitchnet, are also available through the repository.
Several refinements have been made to the distribution’s file actions, including crop and preview functionality for the video upscaler. A new MX Linux utility for verifying ISO signatures is included as well.
One important consideration for existing users is that the extensive package restructuring requires a fresh installation to obtain the complete AV Linux 25.3 experience, including the new desktop configurations and repository integration.
However, users running AV Linux or MX Moksha 25 through 25.2 can enable the new Main repository without reinstalling. This provides access to updated multimedia applications, plugins, and file actions, although the new themes and desktop configurations will not be included.
Finally, the developer has discontinued XLibre ISO builds, citing the additional maintenance requirements and limited VPS storage capacity. Future package updates will also be distributed through the new repository rather than the old web archive.
For more information, see the announcement.
Rocky Linux 10.3 Released with Kernel 6.12, GCC 16, and Security Updates
Rocky Linux 10.3 (Red Quartz) has been officially released following the Red Hat Enterprise Linux 10.3 release on November 11. The new version arrives alongside AlmaLinux 10.3, also released today, giving users two updated free RHEL-compatible options.
As expected, Rocky Linux 10.3 is built around Linux kernel 6.12, specifically version 6.12.0-264.16.1. Several kernel components have been updated beyond that baseline. The eBPF subsystem and perf tools are rebased to Linux 6.19, while memory management improvements include newer CXL support and changes to compressed swap handling.
Hardware support has also been improved, particularly for Intel platforms, while AMD’s amdxdna driver for Neural Processing Units (NPUs) is now available as a Technology Preview.
On the development side, Rocky Linux 10.3 introduces GCC Toolset 16, featuring GCC 16.2 and Binutils 2.46, alongside LLVM 22.1.8. The update also includes Rust 1.97, Go 1.27.3, and GDB 17.2, giving developers access to newer language features, debugging capabilities, and compiler improvements.
Regarding security, Libreswan 5.4 introduces post-quantum key exchange support through ML-KEM-768 and ML-KEM-1024, while SSH connections operating in FIPS mode now support and prefer hybrid post-quantum key exchange algorithms.
Other security-related updates include Keylime 7.14.3, which improves remote attestation capabilities, and fapolicyd 2.0.1, which introduces parallel policy processing and transactional updates. OpenSC 0.27.1 adds PKCS #11 3.2 support, while the updated Audit 4.2.1 provides additional capabilities for secure remote logging.
System monitoring also receives updates, with Grafana 12.4.3 introducing a redesigned dashboard architecture and improvements to unified alerting. Performance Co-Pilot 7.1.5 adds automatic archive repair and additional performance metrics.
For administrators working with system images, the image-builder command-line interface is now generally available. The web-based image builder can generate bootc-compatible images from locally available base images, while Kickstart gains additional certificate-handling capabilities during installation.
There are several compatibility changes worth mentioning. WEP wireless security support has been removed, along with the ZBUD compressed page allocator. Grafana also drops legacy alerting and AngularJS plugin support, which could affect existing monitoring configurations.
Regarding supported architectures, Rocky Linux 10.3 is available for x86-64-v3, ARM64, RISC-V, POWER10 and newer (ppc64le), and IBM z15 and newer. As with previous Rocky Linux 10 releases, older x86-64-v2 processors are not supported.
Existing Rocky Linux 10 users can upgrade to version 10.3 using the standard sudo dnf -y upgrade command or graphical update tools. However, upgrades from Rocky Linux 8 or 9 to version 10 are not officially supported, with a fresh installation recommended instead.
Installation ISOs, container images, cloud images, and live images are available from the Rocky Linux downloads page. For more information, see the official announcement.
Print settings vs print dialogues
I feel like I’m missing something, and possibly have been gaslighting myself for a while.
I just got a Brother mono printer and logged into its web config to set it for double sided printing and a few other things. I always thought this was better than using the os printer config gui to set defaults.
When I went to print from LibreWriter (from mint) I needed single sided instead. Oddly the print dialogue said single sided. So I hit print expecting single sided, but it came out double sided.
There were similar issues with the old printer i.e. things not coming out as sent from whatever application.
How do I make sure that what I’m seeing in a print dialogue (from any given application) is what will come out.
Since when did Linux held 8% of the global marketshare???
I was looking at statcounter when I came across that Linux once held 8% of the desktop marketshare worldwide. Since when did it get this far in popularity???
Boosted by @welcome@friends.deko.cloud
Hello Mastodon! 👋
Thought I'd introduce myself!
I'm Mike, living in North Yorkshire, UK. By day I'm a Church of England vicar, but this account is mostly for indulging my geeky side!
I'm a Linux enthusiast, running an OMV home server with Docker, Nextcloud, Home Assistant, Immich and various other bits and pieces. I've recently moved to a Jolla phone running Sailfish OS, and I'm enjoying exploring life beyond Android and Google.
I'm interested in open source, self-hosting, privacy and digital independence. I enjoy tinkering, learning new things and occasionally breaking stuff before figuring out how to fix it!
#Introduction #Introductions #Linux #SelfHosting #OpenSource #SailfishOS #Jolla #OpenMediaVault #Docker #HomeAssistant #Nextcloud #DeGoogle #Fediverse
AlmaLinux OS 10.3 Released as Free Alternative to Red Hat Enterprise Linux 10.3
The AlmaLinux OS Foundation announced today the release and general availability of AlmaLinux OS 10.3 (codename Mauve Lion) as the second point release of the latest AlmaLinux OS 10 series, based on Red Hat Enterprise Linux 10.3.
Coming five and a half months after AlmaLinux OS 10.2, the AlmaLinux OS 10.3 release is based on Red Hat Enterprise Linux 10.3 and brings back the ModSecurity web application firewall and the OWASP Core Rule Set (CRS). ModSecurity 2.9.14 with CRS 4.22 is included in the release.
According to the devs, CRS 4 reorganized the rule set, so if you’re bringing rules or tuning over from an AlmaLinux 9 server, plan to review them before you put AlmaLinux OS 10 in front of production traffic.
AlmaLinux OS 10.3 also ships with the Podman 6 container tool, which moves the vendor default configuration from /etc/containers to /usr/share/containers. Moreover, rootful and rootless storage settings now live in separate drop-in directories.
This release also introduces performance enhancements, updated development tools, and improved security. The toolchain has been updated as well to GCC 16, LLVM 22, Rust 1.97, Go 1.27, GDB 17.2, and Node.js 26. Security updates include fapolicyd 2.0, Keylime 7.14, audit 4.2, and libreswan 5.4.
You can download the AlmaLinux OS 10.3 release right now from the official website as installation ISO images for 64-bit (x86_64 and x86_64_v2), AArch64 (ARM64), PowerPC 64-bit Little Endian (ppc64le), and IBM System Z (s390x) architectures. Live and cloud images will be available as well, along with containers.
As expected, this release targets new users and those who want to perform a fresh installation. Existing AlmaLinux OS users can upgrade their installations to the new version by running the sudo dnf update command in a terminal emulator or virtual console.
Also today, the AlmaLinux OS Foundation released AlmaLinux OS 9.9 “Chartreuse Bobcat” as a free alternative to Red Hat Enterprise Linux 9.9, which keeps the ALESCo-approved kernel fix shipped in AlmaLinux OS 9.8, which stops systemd and ps from using excessive CPU during task cleanup. Check out the release notes for more details.
Manjaro Linux Has a Non-Profit Now
The Manjaro team has registered its own non-profit in Germany. Called Manjaro Project e.V., it has seven founding members and is independent of the commercial-oriented Manjaro GmbH & Co. KG entity.
You might remember earlier this year, in March, a group of team members published a manifesto calling for the project to be split from the company, with a strike following soon after.
While this was announced on the Manjaro Linux forum and not on the main project webpage, the registration is already visible in the German register portal, under the Berlin court.
An e.V.?
An e.V. (eingetragener Verein) is a registered association under German law. It is entered in a register kept by the local court, and its purpose cannot be running a commercial business. Registering gives it its own legal identity, separate from the people in it.
That identity matters to the developers because it lets them hold things as a group.
The manifesto asked for equal ownership among members and open voting on big decisions instead of everything being centered on one person. It now means there’s a body that could own the assets if they’re handed over.
Manjaro GmbH & Co. KG is a different kind of body. A GmbH & Co. KG is a limited partnership where a GmbH takes on the personal liability, with the GmbH’s managing directors handling the running of the business.
Stating the obvious, Manjaro now has two separate bodies behind it, and it’s interesting to see how the project ended up in such a situation.
How Manjaro got here
Manjaro has been a German company since 2019. Way back in April 2024, that company announced a management change when Bernhard Landauer stepped away from the company to focus on other work and a new CTO joined to replace him.
Then came the Manjaro 2.0 Manifesto in March, where signatories said the leadership’s priorities didn’t line up with what the developers and community wanted.
Project lead Philip Müller replied, saying he had no personal objection to an association but wouldn’t help set it up and that any asset transfer would have to be discussed closely with the company.
The team has since managed to push out the most recent Testing and Stable updates without Müller being involved.
What now?
The manifesto suggested handing over assets like the manjaro.org domain, the forum, and community donations. According to the team, no transfer of company assets has happened yet.
Before that, the e.V. has some groundwork to do, such as tidying up its statutes, bringing in more members, and opening its own bank account. Talks with the company come after.
For end users, I don’t expect them to be affected in any major way. This kind of feels like a takeover, but a gentle one so far. The situation for donors who had set up recurring payments is a bit hazy.
You see, donations still go to accounts under the company’s control, possibly Müller’s too, and the e.V. hasn’t settled on whether to start fresh accounts or take over the current ones.
Basically, there’s no new recipient to switch to yet, and the team hasn’t said when that gets decided. A second announcement confirming that assets have moved is what I’d watch for, since that’s when the split becomes more than a register entry.
Then there’s the Orange Pi Neo, a handheld Orange Pi built with Manjaro that was put on hold in March because of skyrocketing memory and SSD prices. The announcement doesn’t say who handles partner projects like it, so that’s an open question.
Boosted by @trending@homestead.social
Pretty much 98% of the "tips" we're getting to LWN now are projects that have been largely cobbled together using LLMs. Probably 75% of the project tips are for things that have existed for all of a week and have like 1,000 commits over three days.
I really wish that projects that are entirely human-created and involve multiple people were as eager to promote themselves as the people prompting Claude who want to brag that they "made" something.
Perhaps LLM-driven development is inevitable. I hope not, but perhaps. But even if it is... LLM-first projects will never be as interesting as actual community driven projects with real human work behind them.
It's like the difference between knitting your own sweater or socks versus buying some. If you have made yourself some unique fiber art clothing, I'm down to hear about how you did it and to admire it. I might even be a bit jealous, because some hand-knit clothing is just absolutely gorgeous.
If it's store-bought, though? Congrats. You ... picked something out and paid for it. Nice. It might be worth a "oh, nice sweater" but you don't get bragging rights.
#OpenSource #Linux #LLMs #AI
Boosted by @joe@f.duriansoftware.com
Most Git tutorials stop after add, commit, push.
New post on the parts that come after: worktrees for parallel branches, cherry-pick -x for backports, fixup commits with autosquash, --update-refs for stacked branches, the reflog as your undo history, and signed commits from an OpenPGP smartcard via oct-git.
Plus one trap I tested: a plain rebase silently drops your signatures.
https://blog.hofstede.it/git-beyond-the-basics-worktrees-cherry-picks-and-signed-commits/
#Git #FOSS #Linux #FreeBSD #OpenPGP
NoisyTCP: Linux Kernel TCP Pre-auth Remote Corruption
Fwupd 2.1.10 Brings Fixes, New Synaptics Touchpad Supported
Lead Fwupd/LVFS developer Richard Hughes of Red Hat kicked off the new week by releasing Fwupd 2.1.10 as the newest stable release for performing firmware updates on Linux.
Fwupd 2.1.10 ships with a wide variety of fixes. This takes care of various bugs reported by users as well as AI/LLM detected issues in the codebase. Nothing too particularly exciting on the fixes side unless you were personally impacted by one of the bugs.
On the new hardware support side, the only addition with Fwupd 2.1.10 is introducing support for the Synaptics 96F1 touchpad.
The full list of Fwupd 2.1.10 changes can be found via GitHub.
Boosted by @welcome@friends.deko.cloud
Geophysicist here. I've spent 25+ years analyzing large-scale geophysical data, which is data science by another name. Now I want to own my working environment, so I'm taking a detour through FreeBSD. 🐡
Debian is my daily driver. The lab: a 2011 Sony Vaio netbook (AMD E-350, 6 GB RAM, 120 GB SSD) getting FreeBSD 15.1 + MATE.
Goal: get my geoscience stack (Python, GMT, QGIS) running on it.
New to FreeBSD, so tips are welcome!
#FreeBSD #Linux #DataScience #Geophysics #introduction
Librem 5 second-gen $1599 RISC-V Linux phone - OMG! Ubuntu
Purism is crowdfunding a new version of its Librem 5 Linux smartphone, this one powered by RISC-V rather than Arm.
Librem 5 v2 will use an octa-core SpacemiT K1 RISC-V SoC, replacing the quad-core ARM64 NXP CPU This platform shift will provide much better performance and longer battery life than the existing Librem 5 – and not run as hot in use.
Memory is increased to 8 GB RAM, up from 3 GB in the first-generation handset, and storage leaps to 128 GB from 32 GB.
SD card reading speeds are boosted, and on-board hardware encoding means the Librem 5 v2 can take 13 megapixel photos with and shoot high-definition video for the first time.
The rest of the phone is familiar to the first. The same case design, 5.7-inch (720×1440) display, removable 4,500mAh battery, audio jack and physical kit switches. This is intentional as it will make repairs and upgrades between generations easier.
Wireless connectivity also stays modular, though the v2 handset will ship with an improved 4G modem. A 5G module is planned.
Librem 5 v2 will run PureOS, Purism’s Debian-based operating system using Phosh as its UI.
Librem 5 v2 costs from $1599
“RISC-V is the future, and Librem 5 v2 is the perfect product to bring it to a mainstream audience”, Jonathon Hall, CTO at Purism, says of the launch.
“Freeing us from the control of Big Tech starts with the foundation, and RISC-V’s foundation frees us from the oligarchy. The Librem 5 v2 puts you in control of your data, combining that with the ease of use we expect in the modern era”.
Given that general feedback to the first-gen Librem 5 – beset by delays, price rises and patchy build quality – was mixed best. But, to it credit, Purism is still out there making, selling and supporting the phones (and their other devices and services).
Transitioning to RISC-V for the follow-up won’t be trivial, mind. Graphics acceleration, new drivers and power management (which on a phone is rather critical) will be big undertakings.
And more work is needed to make the day-to-day software stack editable, like browser performance (sluggish on the newest and fastest RISC-V chips) and providing an Arm64 app compatibility layer to run software not made for RISC-V.
Available to back until February 2027
If you’ve got at least $1599 burning a hole in your pocket and you don’t need anything to show for it before 2028, you can go pledge to the campaign. It runs until February 2027, so don’t feel pressured to decide just yet.









