Hace un tiempo que os vine a recomendar el podcast «La era de las distros» un proyecto del gran Lorenzo Carbonell que junto a los protagonistas correspondientes repasa lo que fue la punta de lanza del desembarco de Software Libre en España: las distribuciones educativas autonómicas. Dado que quiero que este proyecto llegue al máximo número de pesonas posibles voy a ir promocionando episodio a episodio porque bien vale la pena escucharlos. Bienvenidos al Episodio 5 del Podcast «La era de las distros», que lleva por título «Software Libre para las Personas. LinEx y la Comunidad» donde nos relatan como LinEx no solo fue software sino Comunidad de la mano deFernanda Jaramillo, Jesús Rubio, Jesús González-Barahona y Lorenzo Carbonell.

Software Libre para las Personas. LinEx y la Comunidad – Episodio 5 del Podcast «La era de las distros»

Como sabréis los que me siguen, he sido docente de secundaria en la Comunidad Valenciana desde el 2000, así que viví los años de la irrupción informática en las aulas en primera persona: los primeros ordenadores en las salas de profesores, las primera aplicaciones para la gestión de los colegios, los tímidos intentos de acercar las TIC al alumnado, las primeras escaramuzas para controlar los móviles en las aulas, etc.

En ese panorama, algunas comunidades autónomas del estado español decidieron que era el momento de dar una oportunidad al Software Libre, tanto por razones económicas como por tener la soberanía digital que tanta importancia está adquiriendo ultimamente.

Contar la historia de estas iniciativas es el propósito del Podcast de «La era de las distros» que en este quinto episodio vemos el lado más humano de LinEx, pero mejor que nos lo describan los protagonistas:

El software libre suele asociarse a programadores encerrados en habitaciones oscuras, líneas de código interminables y debates técnicos indescifrables. Pero hubo un tiempo y un lugar donde el software libre significó algo muy distinto: libertad, comunidad y orgullo. Ese lugar fue Extremadura y el protagonista fue LinEx.

En el último episodio de nuestro podcast, nos sumergimos en el «rostro humano» de esta revolución tecnológica. No lo hacemos solos; nos acompañan Fernanda Jaramillo, coordinadora del Plan de Alfabetización Tecnológica, y Jesús Rubio, jefe de servicio durante aquella etapa dorada.

Los participantes del podcast son:

• Fernanda Jaramillo. Durante esos años fue Coordinadora del Plan de Alfabetización Tecnológica de Extremadura.
• Jesús Rubio. Durante esos años fue Jefe de Servicio en la Dirección General de Sociedad de la Información de la Junta de Extremadura.
• Jesús González-Barahona. Catedrático del área de Ingeniería Telemática en la Universidad Rey Juan Carlos (Móstoles, Madrid).
• Lorenzo Carbonell

Con este quinto episodio se cierra el ciclo de audios dedicados a LinEx, la primera distribución autonómica, la que rompió el hielo y que facilitó el camino a las que iban llegando.

Os puedo asegurar que vale la pena escucharlo para que no caiga en el olvido el increíble trabajo que realizaron las personas involucradas en el proyecto y para aprender a gestionar el futuro, aprendiendo de las cosas que no se pudieron hacer bien en el pasado.

Puedes escuchar el podcast en cualquiera de las siguientes plataformas:

• 
  Cómo Extremadura cambió el Software Libre para siempre – Episodio 2 del Podcast «La era de las distros»
  Bienvenidos al Episodio 2 del Podcast «La era de las distros», que lleva por título «Cómo Extremadura cambió el Software Libre para siempre» donde nos relatan el nacimiento de Linex, la distribución extremeña que inició la revolución.
• 
  Lanzado LinEx Colegios 2010
  Y otro lanzamiento más. En esta ocasión otra distribución más para colegios por parte de una administración pública, concretamente de la comunidad autónoma de Extremadura y su Linex. LinEx Colegios… Leer más: Lanzado LinEx Colegios 2010
• 
  Anuncios sobre Linux (V): LinEx
  Esta vez le toca a LinEx, la distribución Linux de Extremadura, la cual también se decanta por Gnome ya que se basa en Debian (gracias env)(la moda de Ubuntu es… Leer más: Anuncios sobre Linux (V): LinEx

La entrada Software Libre para las Personas. LinEx y la Comunidad – Episodio 5 del Podcast «La era de las distros» se publicó primero en KDE Blog.

A few weeks ago I picked up my first real bug in the Open Build Service (OBS) codebase, part of the openSUSE project: a production crash that had been showing up on Errbit every time a GitHub workflow config with the “wrong” character in it got synced. Here’s what broke, why it broke, and how the fix closed it out.

The bug

OBS lets you wire up a GitHub repository so pushes trigger a workflow, things like package rebuilds via webhooks. When that webhook fires, Workflows::YAMLDownloader pulls down the workflow's YAML config file from GitHub and writes it to a Tempfile so OBS can parse it.

The crash looked like this in production:

"\xE2" from ASCII-8BIT to UTF-8

Thrown from IO#write, inside Workflows::YAMLDownloader#create_temp_file.

Why it happened

GitHub’s API returns file contents Base64-encoded. OBS decodes that with Base64.decode64, which hands back a plain Ruby string, but that string is tagged with the ASCII-8BIT encoding (Ruby's way of saying "this is just raw bytes, no assumptions about text"). That's correct and expected; Base64.decode64 has no way to know what encoding the original bytes represent.

The problem shows up one step later. Rails’ Tempfile defaults to expecting UTF-8 when you write to it. So when OBS calls .write with an ASCII-8BIT string, Ruby tries to transcode it, converting the bytes from ASCII-8BIT to UTF-8 on the fly. That's fine for plain ASCII content. But the moment the YAML file contains a multi-byte UTF-8 character, an em dash, a smart quote, anything outside the ASCII range, Ruby has no valid transcoding rule to apply, and it raises Encoding::UndefinedConversionError.

In short: a perfectly valid UTF-8 file was crashing because Ruby thought it was starting from ASCII-8BIT and tried to “translate” bytes that didn’t need translating at all.

The fix

The core of it comes down to one call:

decoded_content.force_encoding('UTF-8')

added right before the Tempfile write. The full diff landed at three lines changed across two files: the fix itself in yaml_downloader.rb, plus a matching update to the spec covering that code path, so the regression is actually tested going forward and not just patched.

The key distinction with the fix itself is force_encoding versus a real transcode. force_encoding doesn't touch a single byte; it just relabels the string, telling Ruby "trust me, these bytes are already UTF-8." Since the bytes coming out of Base64.decode64 genuinely were valid UTF-8 all along (they just carried the wrong encoding tag), relabeling is exactly the correct fix. No transcoding attempt means no UndefinedConversionError, because there's nothing left to convert. Ruby writes the bytes exactly as they are.

What I’d tell someone hitting this themselves

If you’re decoding Base64 content in Ruby and writing it somewhere that expects text, check what encoding the decoded string is actually tagged with (str.encoding) before you assume it matches what the destination wants. Base64.decode64 will hand you ASCII-8BIT regardless of what the original content was; it's on you to correctly re-tag it if you know better, and force_encoding is the right tool for that when you're confident the bytes are already valid in the target encoding. Reach for a real transcode (.encode) only when the bytes genuinely aren't in the target encoding yet.

Getting the PR in

This was my first contribution to OBS, and I want to call out something small that made a real difference: I asked the maintainer, hennevogel, if it was okay for me to take the issue before opening a PR. His reply was simple: “no need to ask me. Just PR small things :)”, which told me a lot about how this project, and the openSUSE community around it, actually wants newcomers to engage. Low friction, just ship the fix.

The PR (#20236) was small, three lines changed across two files, reviewed, and merged. Small fix, but a good first real look at how OBS’s Rails frontend talks to external Git hosts, and that boundary, it turns out, is where interesting bugs tend to live.

Executar inteligência artificial localmente exige mais do que escolher um bom modelo. É preciso aproveitar o hardware disponível, administrar memória e reduzir o tempo de resposta. As novidades do OpenVINO 2026.4.0 avançam nessas frentes, com ampliação do suporte a modelos, novas técnicas de aceleração e melhorias para aplicações que utilizam CPUs, GPUs e NPUs Intel.

A atualização reúne recursos para linguagem, visão, reconhecimento de fala e geração de imagens. Para quem desenvolve soluções de IA local, embarcada ou corporativa, isso amplia as possibilidades de construir aplicações com diferentes modalidades dentro do mesmo ecossistema.

Mais modelos e possibilidades de aplicação

A relação de novos modelos suportados inclui Gemma-3n em CPU e uma lista abrangente para execução em CPU e GPU: Kokoro-82M, Qwen3-VL-4B com EAGLE-3, Qwen3-ASR, Muse Glimmer 30B, Qwen3.8 27B, Gemma 4 12B, Hy-MT2-1.8B, DeepSeek OCR-2 e Granite 4.0 H Micro.

Essa expansão aproxima diferentes etapas de uma aplicação: reconhecer fala, interpretar imagens e documentos, traduzir conteúdo, gerar respostas e sintetizar voz.

Nas NPUs, os destaques são FLUX.2-Klein 4B e Kokoro-82M, ampliando as opções para geração de imagens e síntese de fala nesse acelerador. O suporte anunciado deve ser considerado junto aos requisitos de cada modelo e plataforma: a presença de uma NPU, por si só, não garante que qualquer configuração consiga executar todas essas cargas.

Também chegam, como versões iniciais para CPU e GPU, Qwen-Image, Z-Image-Turbo, Granite 4.0 H Tiny, Fun-ASR-Nano, LFM2.5-8B-A1B, MiniCPM5-2B, RF-DETR, BGE Reranker-V2-M3 e BGE M3. São novas opções para experimentação, com validação necessária antes da adoção em produção.

Geração de texto acelerada com previsão de múltiplos tokens

Uma das novidades mais relevantes do OpenVINO GenAI é o suporte à decodificação especulativa com Multi-Token Prediction, ou MTP, para Gemma 4, Qwen3.5 e Qwen3.6 em CPUs e GPUs.

A ideia é antecipar múltiplos tokens e submetê-los à validação, aproveitando as previsões aceitas para acelerar a geração. Segundo as notas da versão, o objetivo é aumentar a vazão e reduzir a latência sem sacrificar a precisão.

O ganho efetivo depende do modelo, do hardware e da carga de trabalho. Por isso, vale medir tanto o tempo até o primeiro token quanto a velocidade de geração e o tempo total da resposta.

O release também apresenta, em prévia, aceleração DFlash para modelos Qwen em GPUs e suporte a tokens visuais voltado à redução de latência em pipelines GenAI nos processadores Intel Core Ultra Series 3.

Mais desempenho em aplicações multimodais

O EAGLE-3 passa a contar com Tree Drafting, utilizando Top-K. Em vez de explorar uma única sequência de candidatos, como no Chain Drafting com Top-1, essa abordagem permite considerar alternativas em uma estrutura de árvore.

A proposta é aumentar a vazão em pipelines de modelos de visão e linguagem, que combinam informações visuais e textuais.

Outra melhoria está nas otimizações para os gráficos integrados Xe3. Elas buscam melhorar a inferência de modelos Gemma 4 com entradas de contexto longo nos Intel Core Ultra Series 3, um cenário relevante para aplicações que precisam processar grandes volumes de informação antes de responder.

CPU, GPU e NPU na mesma ferramenta de análise

O OpenVINO estende à NPU o suporte de instrumentação e rastreamento ITT. Com isso, o Intel VTune Profiler passa a oferecer uma ferramenta consistente para analisar a execução em CPU, GPU e NPU.

Esse avanço ajuda a investigar gargalos e fundamentar decisões sobre a distribuição das cargas de trabalho. Para a computação heterogênea, observar o comportamento de cada dispositivo é tão importante quanto conseguir executar o modelo nele.

Também em prévia, o suporte a formas dinâmicas com limites definidos nas NPUs avança para tarefas de visão, como super-resolução de imagens, com validação no modelo ESPCN. É um passo relevante para entradas de dimensões variáveis, dentro dos limites suportados, sem representar suporte irrestrito a qualquer forma dinâmica.

Reconhecimento de fala diretamente no Node.js

O OpenVINO GenAI adiciona suporte ao ASRPipeline em Node.js. Desenvolvedores JavaScript passam a contar com uma API para reconhecimento automático de fala, incluindo modelos como Whisper e Qwen3-ASR, com streaming e métricas de desempenho.

A interface segue uma abordagem semelhante à disponível em C++ e Python. Isso facilita a integração de transcrição em aplicações e serviços que já utilizam JavaScript.

Servidores com melhor aproveitamento de memória

O OpenVINO Model Server recebe, em prévia, gerenciamento de modelos ociosos. A funcionalidade permite descarregar modelos que não estão em uso para reduzir o consumo de memória.

Esse recurso pode ser especialmente útil em serviços que disponibilizam vários modelos com demandas intermitentes. Na avaliação prática, será importante considerar também o tempo necessário para recarregá-los quando uma nova solicitação chegar.

O Model Server amplia ainda o suporte a modelos voltados a aplicações com agentes, incluindo Muse Glimmer 30B e Qwen3.8 27B.

Para quem trabalha com IA local e infraestrutura própria, o OpenVINO 2026.4.0 oferece novas oportunidades de aproveitar o hardware disponível. O próximo passo é transformar suporte anunciado em resultados medidos: validar compatibilidade, qualidade das respostas, consumo de memória e latência nas condições reais de uso. É assim que novas capacidades se tornam soluções úteis no dia a dia.

Fonte: https://github.com/openvinotoolkit/openvino

There were a few software package updates for openSUSE Tumbleweed ARM this month with snapshots reaching the ARM port.

September’s snapshots of the ARM port carried the Linux kernel to a security-heavy 7.2.5. There was a jump to glibc 2.44 with its large AArch64 push, and QEMU reaching 11.1.1. Raspberry Pi owners get a firmware fix that clears up a Wi-Fi scan error, the Networking Stack received security patches, and GIMP 3.2.6 brings AArch64-specific fixes.

Raspberry Pi

This month’s raspberrypi-firmware update prevents the spurious “-52” error message that could appear during Wi-Fi scans (bsc#1215134), so scanning for networks is quieter on boards where the firmware reported a phantom failure. The Linux kernel 7.2.4 update fixed a memory leak in the brcmfmac driver’s SDIO control path, which reduces memory pressure over long uptimes on the onboard Cypress Wi-Fi chip used across recent Pi models.

Rockchip & Other SBCs

Owners of Rockchip and other single-board computers get kernel fixes for the MediaTek Wi-Fi chipsets that many boards ship with, including headroom fixes for USB/SDIO transfers on mt7925 and EEPROM size validation on the mt7915 and mt7996 drivers, which protects against malformed device data (bsc#1012628). kernel-firmware-qcom adds DSP firmware for the Qualcomm x1e80100, nord and qcs8300 platforms, improving modem, DSP and graphics support. dracut now includes leds-qcom-lpg in the aarch64 DRM module set, keeping RGB indicator LEDs working when booting from an initrd. Finally, wireless-regdb enables 320 MHz channels for Hong Kong, updates South Africa’s rules, and drops the 60 GHz DFS flag for Togo.

i.MX & NXP

Developers working with NXP hardware gain the imx8mp-evk machine type in QEMU 11.1.1, which emulates the i.MX 8M Plus Evaluation Kit and makes it easier to test software for that platform without physical hardware. The kernel 7.2.3 update fixed the mxs-dcp crypto engine’s source scatterlist length access and improved the Data Co-Processor security block on i.MX System on Chips that use it for accelerated encryption. These changes make NXP hardware slightly more accessible to both emulation and crypto workloads.

aarch64 Virtualization, UEFI & Toolchain

QEMU 11.1.1 is the virtualization highlight of the month, advancing the emulator with Universal Flash Storage emulation for Write Booster and Host-Initiated Defragmentation based on the UFS 4.1 specification, plus vhost-host-user support for offloading real-time clock handling through virtio-rtc. The ARM targets gain new architectural CPU features, and the virt board can now specify cache topology. The ARM-specific fixes include a regression test and boundary-case fix for the SVE2 whilewr/whilerw instructions and a fix for arm_gicv3 kconfig selection, and openSUSE disables GCS linker validation in the spec to keep the aarch64 build working.

glibc 2.44 delivers the strongest AArch64 additions in months. On targets with the Guarded Control Stack extension, glibc now locks all GCS operations, including status, write-on-shadow-stack and push-to-shadow-stack, immediately after enabling GCS with an ENFORCED or OVERRIDE policy, closing a window where a process could still mutate its own shadow stack. Special-case paths for log, exp, sin, cas, sinh, cosh, asinh, acosh and atanh are vectorized for both SVE and AdvSIMD, and vector powr variants are added, which speeds up floating-point-heavy code on capable Cortex-X and Neoverse cores.

ovmf advances to edk2-stable202608, refreshing the UEFI firmware that backs the qemu-uefi-aarch64 subpackage. The update adds AArch64 host test coverage across SecurityPkg, CryptoPkg, PrmPkg, DynamicTablesPkg, MdeModulePkg and MdePkg, plus UnitTestFrameworkPkg support for GCC AArch64, which strengthens firmware quality assurance for arm64 virtual machines. The EFI Memory Attributes Protocol workaround is dropped because GRUB2 now supports the protocol, so booting stays compatible with the cleaner firmware behavior.

sdbootutil improved its systemd-boot handling with a new status command, parallel test execution, a --repair parameter for cleanup, and better reporting of entries with missing files. The update prevents duplicate entries on non-snapper systems, keeps the exit status of sdbootutil calls intact, and avoids writing the recovery PIN to the journal, which is a meaningful privacy improvement for systems using PCR-based secure boot.

GTK4 4.22.5 fixes a build failure on 32-bit ARM and repairs 32-bit Vulkan image builds, so the toolkit and gskvulkanimage both compile cleanly on armv7. A Wayland session crash in gdk_wayland_toplevel_remove_from_session() and a memory buffer fix round out the release for desktop users on ARM.

LLVM 23.1.1 arrives as a bugfix release for the 23.1.0 series, remaining API and ABI compatible while dropping the obsolete i586 build patch.

GIMP 3.2.6 carries the first AArch64-specific wins for the image editor. The clipboard brush and pattern maximum size is bumped to 8192 on AArch64 as well, matching the ceiling that x86-64 and PPC64 have had since the 3.2.0 release candidates, so larger brushes and patterns survive clipboard round-trips on ARM desktops.

shotwell 33.0 makes the jump from the 0.32 series with a full port to GTK4 (requiring at least 4.22), a reworked printing flow, and fixes for the fullscreen toolbar auto-hide and disappearing slideshow icons. The release replaces its dedicated authentication helper with a small localhost web server for publishing, adds a “peek password” toggle to the Piwigo flow and fixes setting desktop backgrounds.

kbd 2.10.0 extends the console stack with XKB support: loadkeys can now generate console keymaps from XKB data, import XKB compose tables and handle XKB group switching, modifier handling, virtual console switching and keypad/editing remaps.

Snapshots at a Glance

ARM Resources

For ARM-specific discussions and support, subscribe to the openSUSE ARM mailing list and check the openSUSE ARM wiki for device support information, image downloads, and documentation. New and existing ARM users can also join the openSUSE ARM IRC channel for real-time help.

Sigo la serie con las nuevas funcionalidades que nos ofrece la Comunidad KDE después de lanzar su conjunto de aplicaciones revisadas y mejoradas. Finalizamos con las novedades de Minuet de KDE Gear 26.08 edición «Enjoy Shiny Stuff» la aplicación de educación musical de KDE.

Las novedades de Minuet de KDE Gear 26.08, edición «Enjoy Shiny Stuff»

Minuet es la aplicación de KDE para la educación musical cuyo objetivo es ayudar a estudiantes y músicos a entrenar el oído mediante ejercicios de intervalos, acordes, escalas y ritmos.

Para este KDE Gear 26.08, Minuet estrena una nueva interfaz diseñada para funcionar a la perfección tanto en pantallas de escritorio como en dispositivos móviles.

Para lograrlo, la página de inicio y el panel de navegación facilitan la localización de las categorías de ejercicios, mientras que el explorador de actividades presenta ahora cada ejercicio en forma de tarjeta con una breve descripción. La categoría actual permanece resaltada y el nuevo campo de búsqueda permite filtrar los ejercicios por sus nombres y descripciones traducidos.

Del mismo modo, las páginas de los ejercicios se han reorganizado para aprovechar mejor el espacio disponible en ventanas estrechas y teléfonos.

En resumen, un lavado de cara para que el uso de esta aplicación sea lo más satisfactorio posible.

Todo ello mientras se prepara el 30 aniversario de un proyecto que según más de uno es el I+D+I de un Software de todo el mundo para todo el mundo.

Y, recuerda, todo este software es gratuito y sin publicidad en todos los sentidos: no te cuesta ni un euro y no se cobra en en forma de datos personales. No obstante, si quieres ayudar a su desarrollo siempre puedes participar en su campaña de recaudación de fondos.

• 
  Las novedades de Minuet de KDE Gear 26.08, edición «Enjoy Shiny Stuff»septiembre 16, 2026
  Las novedades de Minuet de KDE Gear 26.08, edición «Enjoy Shiny Stuff»
• 
  Las novedades de Kdenlive de KDE Gear 26.08, edición «Enjoy Shiny Stuff»septiembre 15, 2026
  La novedades de Kdenlive de KDE Gear 26.08, edición «Enjoy Shiny Stuff»
• 
  Las novedades de Konsole de KDE Gear 26.08, edición «Enjoy Shiny Stuff»septiembre 2, 2026
  Las novedades de Konsole de KDE Gear 26.08, edición «Enjoy Shiny Stuff»

La entrada Las novedades de Minuet de KDE Gear 26.08, edición «Enjoy Shiny Stuff» se publicó primero en KDE Blog.

author: Douglas DeMaio date: 2026-09-16 14:00:00+02:00 layout: post image: /wp-content/uploads/2025/07/sbc.png license: CC-BY-SA-3.0 title: Tumbleweed ARM Updates in September categories:

• Announcements
• openSUSE
• Tumbleweed
• Slowroll
• MicroOS
• arm tags:
• openSUSE
• Tumbleweed
• Developers
• sysadmin
• user
• Open Source
• rolling release
• gamers
• superuser
• distrowatch
• Linux
• kernel
• kernel-source
• Mesa
• graphics
• CVE
• Power Users
• Superuser
• ARM
• AArch64
• Raspberry Pi
• Rockchip
• SBC
• Jetson
• QEMU
• glibc
• UEFI
• snapper
• GIMP
• shotwell
• kbd

There were a few software package updates for openSUSE Tumbleweed ARM this month with snapshots reaching the ARM port.

September’s snapshots of the ARM port carried the Linux kernel to a security-heavy 7.2.5. There was a jump to glibc 2.44 with its large AArch64 push, and QEMU reaching 11.1.1. Raspberry Pi owners get a firmware fix that clears up a Wi-Fi scan error, the Networking Stack received security patches, and GIMP 3.2.6 brings AArch64-specific fixes.

Raspberry Pi

This month’s raspberrypi-firmware update prevents the spurious “-52” error message that could appear during Wi-Fi scans (bsc#1215134), so scanning for networks is quieter on boards where the firmware reported a phantom failure. The Linux kernel 7.2.4 update fixed a memory leak in the brcmfmac driver’s SDIO control path, which reduces memory pressure over long uptimes on the onboard Cypress Wi-Fi chip used across recent Pi models.

Rockchip & Other SBCs

Owners of Rockchip and other single-board computers get kernel fixes for the MediaTek Wi-Fi chipsets that many boards ship with, including headroom fixes for USB/SDIO transfers on mt7925 and EEPROM size validation on the mt7915 and mt7996 drivers, which protects against malformed device data (bsc#1012628). kernel-firmware-qcom adds DSP firmware for the Qualcomm x1e80100, nord and qcs8300 platforms, improving modem, DSP and graphics support. dracut now includes leds-qcom-lpg in the aarch64 DRM module set, keeping RGB indicator LEDs working when booting from an initrd. Finally, wireless-regdb enables 320 MHz channels for Hong Kong, updates South Africa’s rules, and drops the 60 GHz DFS flag for Togo.

i.MX & NXP

Developers working with NXP hardware gain the imx8mp-evk machine type in QEMU 11.1.1, which emulates the i.MX 8M Plus Evaluation Kit and makes it easier to test software for that platform without physical hardware. The kernel 7.2.3 update fixed the mxs-dcp crypto engine’s source scatterlist length access and improved the Data Co-Processor security block on i.MX System on Chips that use it for accelerated encryption. These changes make NXP hardware slightly more accessible to both emulation and crypto workloads.

aarch64 Virtualization, UEFI & Toolchain

QEMU 11.1.1 is the virtualization highlight of the month, advancing the emulator with Universal Flash Storage emulation for Write Booster and Host-Initiated Defragmentation based on the UFS 4.1 specification, plus vhost-host-user support for offloading real-time clock handling through virtio-rtc. The ARM targets gain new architectural CPU features, and the virt board can now specify cache topology. The ARM-specific fixes include a regression test and boundary-case fix for the SVE2 whilewr/whilerw instructions and a fix for arm_gicv3 kconfig selection, and openSUSE disables GCS linker validation in the spec to keep the aarch64 build working.

glibc 2.44 delivers the strongest AArch64 additions in months. On targets with the Guarded Control Stack extension, glibc now locks all GCS operations, including status, write-on-shadow-stack and push-to-shadow-stack, immediately after enabling GCS with an ENFORCED or OVERRIDE policy, closing a window where a process could still mutate its own shadow stack. Special-case paths for log, exp, sin, cas, sinh, cosh, asinh, acosh and atanh are vectorized for both SVE and AdvSIMD, and vector powr variants are added, which speeds up floating-point-heavy code on capable Cortex-X and Neoverse cores.

ovmf advances to edk2-stable202608, refreshing the UEFI firmware that backs the qemu-uefi-aarch64 subpackage. The update adds AArch64 host test coverage across SecurityPkg, CryptoPkg, PrmPkg, DynamicTablesPkg, MdeModulePkg and MdePkg, plus UnitTestFrameworkPkg support for GCC AArch64, which strengthens firmware quality assurance for arm64 virtual machines. The EFI Memory Attributes Protocol workaround is dropped because GRUB2 now supports the protocol, so booting stays compatible with the cleaner firmware behavior.

sdbootutil improved its systemd-boot handling with a new status command, parallel test execution, a --repair parameter for cleanup, and better reporting of entries with missing files. The update prevents duplicate entries on non-snapper systems, keeps the exit status of sdbootutil calls intact, and avoids writing the recovery PIN to the journal, which is a meaningful privacy improvement for systems using PCR-based secure boot.

GTK4 4.22.5 fixes a build failure on 32-bit ARM and repairs 32-bit Vulkan image builds, so the toolkit and gskvulkanimage both compile cleanly on armv7. A Wayland session crash in gdk_wayland_toplevel_remove_from_session() and a memory buffer fix round out the release for desktop users on ARM.

LLVM 23.1.1 arrives as a bugfix release for the 23.1.0 series, remaining API and ABI compatible while dropping the obsolete i586 build patch.

GIMP 3.2.6 carries the first AArch64-specific wins for the image editor. The clipboard brush and pattern maximum size is bumped to 8192 on AArch64 as well, matching the ceiling that x86-64 and PPC64 have had since the 3.2.0 release candidates, so larger brushes and patterns survive clipboard round-trips on ARM desktops.

shotwell 33.0 makes the jump from the 0.32 series with a full port to GTK4 (requiring at least 4.22), a reworked printing flow, and fixes for the fullscreen toolbar auto-hide and disappearing slideshow icons. The release replaces its dedicated authentication helper with a small localhost web server for publishing, adds a “peek password” toggle to the Piwigo flow and fixes setting desktop backgrounds.

kbd 2.10.0 extends the console stack with XKB support: loadkeys can now generate console keymaps from XKB data, import XKB compose tables and handle XKB group switching, modifier handling, virtual console switching and keypad/editing remaps.

Snapshots at a Glance

ARM Resources

For ARM-specific discussions and support, subscribe to the openSUSE ARM mailing list and check the openSUSE ARM wiki for device support information, image downloads, and documentation. New and existing ARM users can also join the openSUSE ARM IRC channel for real-time help.

With GNOME 51 out the door, it's wallpaper reveal season again. This time around it's evolution, not revolution — the set sticks to its geometric roots.

The default is really just a stylistic touch up of the 50 hexagons. The subtle rim highlight received a spotlight and now shines extra bright.

I do keep hoping landscape nature photos eventually join the lineup, but capturing the same scenery under different conditions so the light and dark variants actually make sense is trickier than it sounds. That one's still on the wishlist.

As usual, plenty of concepts didn't make the cut. For every wallpaper that ships, there's a good pile of experiments that never got past the "that's kinda neat" stage.

One thing we keep struggling with is performance in the Appearance panel. The images now include an embedded small thumbnail, so hopefully a faster way to build the initial cache of thumbnails is on the horizon.

We've also started embedding attribution and license straight into the images themselves, so the credits travel with the file instead of living only in the repo. And with Loupe displaying the metadata nicely, you get to see it conveniently.

Con este script en Bash puedes descargar y volver a ver en tu equipo cualquiera de las 1144 tiras que su creador Danigm ha publicado desde 2004 hasta 2026

El pasado 10 de septiembre de 2026. La tira de Linux Hispano llegaba a su fin. Su creador, Danigm, cerraba así un círculo que empezó en 2004 y le llevó a publicar 1144 tiras en sus 22 años de vida.

Un gran ejemplo de constancia y de empeño. Mi agradecimiento a Danigm por todos esos años. Mi lector de feed tendrá un hueco cada semana.

Pero las tiras están publicadas bajo una licencia libre, así que podremos volver a revivirlas en nuestro equipo cuando queramos. Para eso he creado un repositorio y un script en Bash que te mostrará una tira en tu equipo utilizando feh (así que esta es la única dependencia).

Lo primero que hice fue descargar todas las tiras desde la web del creador. Mas 2 tiras que desde esa página no pude descargar, que estaban en formato jpg y las descargué desde la web de Linux Hispano.

El script en python para descargar las imágenes es el siguiente:

#!/usr/bin/env python3

import os
import re
import requests
from bs4 import BeautifulSoup
from urllib.parse import urljoin

BASE_URL = "https://tira.danigm.net/"
OUTPUT_DIR = "tiras"

# Meses en español -> número
MESES = {
    "enero": "01",
    "febrero": "02",
    "marzo": "03",
    "abril": "04",
    "mayo": "05",
    "junio": "06",
    "julio": "07",
    "agosto": "08",
    "septiembre": "09",
    "octubre": "10",
    "noviembre": "11",
    "diciembre": "12",
}

HEADERS = {
    "User-Agent": "Mozilla/5.0"
}

os.makedirs(OUTPUT_DIR, exist_ok=True)

session = requests.Session()
session.headers.update(HEADERS)

print("Leyendo listado...")
response = session.get(BASE_URL)
response.raise_for_status()

soup = BeautifulSoup(response.text, "html.parser")

total = 0
descargadas = 0
existentes = 0
errores = 0

for link in soup.find_all("a", href=True):

    texto = link.get_text(" ", strip=True)

    # Ejemplo:
    # Eof.Png (08 de septiembre de 2026)
    match = re.match(
        r"^(.+?)\.Png\s+\((\d{1,2})\s+de\s+([a-záéíóú]+)\s+de\s+(\d{4})\)$",
        texto,
        re.IGNORECASE
    )

    if not match:
        continue

    nombre = match.group(1)
    dia = match.group(2).zfill(2)
    mes_nombre = match.group(3).lower()
    año = match.group(4)

    if mes_nombre not in MESES:
        print(f"Mes desconocido: {mes_nombre}")
        errores += 1
        continue

    mes = MESES[mes_nombre]
    fecha = f"{dia}-{mes}-{año}"

    # La página individual, por ejemplo:
    # https://tira.danigm.net/all/1143-EOF.html
    pagina_url = urljoin(BASE_URL, link["href"])

    total += 1

    print(f"[{total}] {nombre} ({fecha})")

    try:
        pagina_response = session.get(pagina_url)
        pagina_response.raise_for_status()
    except requests.RequestException as e:
        print(f"    ERROR página: {e}")
        errores += 1
        continue

    pagina_soup = BeautifulSoup(pagina_response.text, "html.parser")

    # Buscamos la imagen de la tira
    imagen = None

    for img in pagina_soup.find_all("img", src=True):
        src = img["src"]

        if "_images/" in src:
            imagen = urljoin(pagina_url, src)
            break

    if not imagen:
        print("    ERROR: no se encontró la imagen")
        errores += 1
        continue

    # Siempre guardamos como PNG
    archivo = f"{nombre}|{fecha}.png"
    destino = os.path.join(OUTPUT_DIR, archivo)

    if os.path.exists(destino):
        print(f"    Ya existe: {archivo}")
        existentes += 1
        continue

    try:
        imagen_response = session.get(imagen)
        imagen_response.raise_for_status()

        with open(destino, "wb") as f:
            f.write(imagen_response.content)

        print(f"    OK -> {archivo}")
        descargadas += 1

    except requests.RequestException as e:
        print(f"    ERROR descargando: {e}")
        errores += 1

print()
print("=" * 50)
print(f"Entradas encontradas : {total}")
print(f"Descargadas           : {descargadas}")
print(f"Ya existentes          : {existentes}")
print(f"Errores                : {errores}")
print("=" * 50)

Esto me descargó las imágenes con el formato nombre|fecha_publicación.png

Esas imágenes las he subido a un repositorio de GitHub. He utilizado GitHub, para no sobrecargar de capacidad ni de peticiones ningún servidor comunitario como Codeberg o similares ni quería utilizar el servidor del propio creador. GitHub tiene dinero suficiente para mantener en sus servidores estas imágenes y capacidad para gestionar peticiones de descarga.

Ahora a ese repositorio de GitHub le hacemos una petición para que nos descargue de manera aleatoria una de las tiras y nos muestre su nombre y fecha de publicación utilizando feh.

El script en Bash en cuestión es el siguiente:

#!/usr/bin/env bash

set -euo pipefail

REPO="https://api.github.com/repos/victorhck/tiras-linux-hispano/contents/images"
RAW="https://raw.githubusercontent.com/victorhck/tiras-linux-hispano/main/images"

tab="                                              "

# Elegir una imagen PNG aleatoria del repositorio
file=$(
    curl -fsSL "$REPO" |
        grep '"name":' |
        sed -E 's/.*"name": "([^"]+)".*/\1/' |
        grep -Ei '\.png$' |
        shuf -n 1
)

[[ -n "$file" ]] || exit 1

# Extraer nombre y fecha
base="${file%.png}"
name="${base%%|*}"
date="${base#*|}"

echo "Nombre: $name"
echo "Fecha de publicación: $date"
echo "Creada por Danigm"
echo "Ctrl+c para cerrar"

# Mostrar la imagen directamente desde GitHub
feh \
    -x \
    --auto-zoom \
    --info "echo '$tab $date'" \
    "$RAW/$file"

Guardamos es script con un nombre y le damos permisos de ejecución y ya podremos ejecutarlo para volver a disfrutar de las tiras de Linux Hispano creadas por Danigm.

Podremos hacer un repaso de algunas de las noticias en estos 22 años de vida de la tira rememorando sucesos que casi habíamos olvidado.

Para terminar basta con pulsar Ctrl+C. Espero que te guste. Y de nuevo agradecer al autor su trabajo.

Two years ago, I created a syslog-ng repo for Amazon Linux 2023. As I received zero feedback, I left it alone until last week. The repo is now updated to the latest version of syslog-ng: 4.12.

According to Fedora Copr download stats, my Amazon Linux 2023 repo has many downloads. However, for over two years, I received zero feedback from the community, no matter how many GitHub, e-mail or social media posts I used to ask users about their experiences and requirements. As any kind of package maintenance needs some effort, I left the repo alone without updates.

Fast forward to last week: someone responded to my post on GitHub, asking for an update. So, once I got some idle time during a longer task, I updated the repo to syslog-ng version 4.12. During another break, I disabled Java support and slog, just as in any other repository we maintain.

As usual, you can find my Amazon Linux 2023 repo at https://copr.fedorainfracloud.org/coprs/czanik/syslog-ng-amazon23/ and share your feedback at https://github.com/syslog-ng/syslog-ng/discussions/4965

Originally published at https://www.syslog-ng.com/community/b/blog/posts/syslog-ng-repo-for-amazon-linux-2023-updated-to-syslog-ng-version-4-12