freenode

← digests

Emacs LLM rules, graphics kernel work, KDE malware

Desktop & Graphics2026-11-16

Emacs developers spent the day arguing whether machine-written code may enter the project, while kernel graphics lists carried several large patch series and KDE warned about a malicious third-party extension.

Emacs developers debate a permanent ban on LLM contributions

A large emacs-devel thread is debating a permanent ban on contributions produced by large language models. The arguments go beyond copyright and cite review burden and harm to the community. The outcome would set a standing rule for what may be accepted into GNU Emacs.

Procfs series aims to remove a userspace pinning attack surface

A resent seven-patch series proposes moving procfs toward the approach kernfs already uses, with the goal of eliminating a userspace pinning attack surface. Linus and core maintainers are actively reviewing it. If it lands, the change would alter a core kernel interface that user-space tools, including graphics stacks, rely on.

ELPA spellchecker submission reopens the LLM policy fight

An ELPA submission for vibe-spell, an LLM-written spellchecker, has triggered a heated policy debate. Discussion centers on maintenance and naming, not only on whether the code itself is acceptable. The submission is a concrete test of the same questions raised in the broader Emacs ban thread.

DMA-buf mapping types proposed for VFIO, IOMMUFD, and P2P

A 23-patch v2 series introduces a DMA-buf mapping types framework, along with SGT conversions and driver updates. The stated use cases are VFIO, IOMMUFD, and peer-to-peer buffer sharing. The work would give drivers a shared way to describe how DMA-buf mappings are established.

KDE reports malware disguised as a Plasma wallpaper extension

The KDE security team reports malware found disguised as a third-party Plasma extension, the Neonova wallpaper. Discussion has turned to whether code from the store matches what is published on GitHub. Anyone installing third-party Plasma extensions should treat store listings and upstream repositories as potentially divergent.

Blender gathers feedback on Material Lighting Nodes

A long thread on Blender's developer forum is collecting feedback on new Material Lighting Nodes. Commenters are focused on limits for non-photorealistic lighting and on planned AOV support. The volume of response suggests the design is still open to change before it settles.

AMDGPU SVM support reaches v10 on the shared DRM core

A 22-patch v10 series adds SVM support to amdgpu on top of the shared DRM GPUSVM core. This phase is limited to a single GPU with XNACK on, and review is ongoing. It is the first step toward AMD using the common DRM SVM path rather than a driver-only mechanism.

RFC would fold DRM linux-next trees into one branch

An RFC proposes replacing most DRM subsystem trees in linux-next with a single drm-for-linux-next tree, described as one drm-tip branch. The stated aim is to reduce merge conflicts. Integrators who track linux-next would see a simpler intake path for DRM if the change is adopted.