llama.cpp Release b11570 Optimizes Test Build System Dependencies
Release b11570 of llama.cpp refactors the test build system to ensure that individual test suites link only against the specific libraries they require, rather than inheriting broad default dependencies. Specifically, helper functions now support a LINK keyword that allows call sites to explicitly declare required targets like ggml or llama. Decoupling test dependencies streamlines the development workflow by reducing unnecessary compilation coupling and improving build efficiency for contributors. This infrastructural cleanup helps maintain a clean, modular codebase as the project scales across diverse hardware backends. The update introduces a default LINK keyword (defaulting to llama-common) in llama_build() and llama_build_and_test(), drops common.h inclusion from testing.h by inlining string_join, and enforces explicit C++17 standards since tests no longer inherit it from shared interfaces.
## BACKGROUND
llama.cpp is a popular open-source software library designed for efficient local inference of large language models on consumer-grade hardware. It is built on top of GGML, a tensor library optimized for machine learning operations.