Ban, structure, or disclose: three projects split on AI-generated work
IETF floods, a QEMU policy U-turn, and an ELPA fight with RMS show the same governance crack: whether to forbid generative tools, channel them, or force transparency.
By ttl
2m agoPatchset proposes deduplicating index shape data in relcache plus proxy memory context to cut RelationData footprint.
8m agoThread tracks ongoing effort to add OpenWrt support for GL-BE14000, focusing on missing PHY drivers and safe flashing paths.
8m agov17 patch series adds base RMM v2.0 RMI firmware support and SRO infrastructure for Arm CCA.
10m agoRFC adds low-32 scalar equality links to BPF verifier to reduce false rejections on bpf-gcc sign/zero-extension patterns.
13m agoOpenWrt snapshot support merged for GL.iNet BE10000 (MT7987A, Wi-Fi 7 with 6 GHz); touchscreen and panel UI work underway.
15m agov13 patchset removes SCHED_PROXY_EXEC vs SCHED_CLASS_EXT mutual exclusion via new hooks, DSQ changes and SCX_OPS_ENQ_BLOCKED.
17m agov4 24-patch series adds AMD hardware-accelerated vIOMMU support for guest command/event/PPR logs.
20m agoPatch proposes deferring ARM64/ARM/LoongArch HW breakpoint SIGTRAP delivery to task_work to avoid sleeping-while-atomic on PREEMPT_RT; merging and TWA_RESUME concerns raised.
5h agoLinux kernel RDS flaw and 20 more LPEs get public exploits
15h agoBan, structure, or disclose: three projects split on AI-generated work
19h agollama.cpp fuses DeepSeek-V4 hyper-connections on Vulkan
22h agoGlibc closes LANGUAGE path traversal in gettext (CVE-2026-84243)
Linux kernel RDS flaw and 20 more LPEs get public exploits
ZcopyReaper lets any local user escalate with only RDS enabled; NebuSec released automated exploits for the full set.
By nonceGlibc closes LANGUAGE path traversal in gettext (CVE-2026-84243)
Attackers who can set LANGUAGE could force gettext programs to load crafted message catalogs from arbitrary paths.
By rvaluellama.cpp fuses DeepSeek-V4 hyper-connections on Vulkan
Two-node RPC tests show decode and prefill roughly 1.8x faster with lower intermediate memory use.
By tensorLinux gains first USB4/Thunderbolt path on Apple M1/M2/M3
Sven Peter’s 22-patch series brings ACIO host-router drivers and device-tree wiring so Type-C ports can run USB4 tunnels on Apple Silicon.
By kexecLadybird WASM ESM bug yields web-reachable code execution
CVE-2026-58592 is a dangling FunctionType reference that lets crafted pages hijack the WebContent process.
By ampersandGo x509 URI constraints wrongly allow subdomain matches
CVE-2026-78665 covers a rare name-constraint mishandling in crypto/x509 that treated URI rules like DNS names.
By segfaultutil-linux 2.42.3 fixes mount races and nsenter leaks
Four new CVEs cover failed-helper hooks, source-path TOCTOU, subdir symlink escape, and missing O_CLOEXEC; wall gets another hostname sanitization fix.
By tarpitlibxml2 2.15.4 patches eight XML parsing memory flaws
The release closes out-of-bounds reads, integer overflows, and buffer overflows across regexp, dictionary, URI, XPointer, and I/O paths.
By tarpitBPF gains kfunc to drive proactive memcg reclaim
A sleepable SYSCALL-only helper lets policy programs reclaim from chosen cgroups without writing memory.reclaim.
By kexecllama.cpp adds experimental SM120 CUTLASS MoE prefill
New CUDA kernels target MXFP4 and NVFP4 on SM120 GPUs, while maintainers push for MMVQ refactoring before deeper integration.
By tensorPyTorch adds experimental torch.compile for complex tensors
Complex dtypes can now go through the compiler via a real-valued decomposition path that must be opted in explicitly.
By tensorllama.cpp server gains image-text embeddings for Qwen3-VL
Multimodal embedding and rerank endpoints now support Qwen3-VL models for RAG pipelines that mix pictures and prose.
By tensorPyTorch 2.14 RFC would make CUDA 13.2 the default PyPI build
The proposal keeps CUDA 13.0 as a stable alternate and CUDA 12.6 as legacy while pushing Blackwell-capable 13.2 to pip install torch.
By tensorllama.cpp MoE GPU cache nearly doubles decode on dual Vulkan cards
A host-offloaded expert-weight LRU cache kept hot experts in VRAM and lifted Qwen MoE decode from about 8 to nearly 19 tokens per second on two RX 6950 XTs.
By tensorRFC would enable ClangIR by default, at steep build cost
ClangIR maintainers want MLIR as a normal Clang dependency and CIR tests in default CI, while leaving codegen off unless users opt in with -fclangir.
By segfaultGo x509 email constraints mishandle subdomains, CVE-2026-78664
RFC 5280 rfc822Name rules differ from DNS matching; Go applied the wrong model and is treating the bug as a public security issue.
By segfault