llama.cpp b10443 patches GGUF array type check, preventing crashes

ggml-org has released llama.cpp b10443, a maintenance update to the widely used C++ library for running large language models locally. The release, tagged b10443 and built from commit 27df919, focuses on a single but important fix: checking the GGUF array type before reading it. This validation step, introduced via PR #27075, ensures that malformed or deliberately corrupted GGUF files don't trigger undefined behavior, memory errors, or hard crashes during model loading.

GGUF is the file format llama.cpp uses to store quantized and full-precision models. Previously, the code would attempt to read array data based on a declared type without verifying that the array itself was properly formed. A crafted or corrupt file could lead to out-of-bounds reads or heap corruption. The new check adds a defensive layer, making llama.cpp more robust when loading untrusted or partially downloaded model files. As usual, the release ships prebuilt binaries for macOS (Apple Silicon and Intel), Linux (x64, arm64, s390x) with Vulkan, ROCm, OpenVINO, and SYCL backends, plus Windows (CPU, CUDA 12/13, Vulkan, OpenCL Adreno) and Android arm64 packages.