Comparing build2 performance against Ninja for C++ project compilation

Ninja as a Build System Benchmark

Ninja is often considered a performance benchmark for build systems due to its minimalist design and reliance on generated build logic. It provides minimal functionality, particularly in change tracking, compared to modern build systems.

The Challenge of Comparison

Directly comparing build systems like build2 with Ninja is difficult for complex projects. When packaging substantial projects for build2, intra-dependencies often need to be restructured, leading to different intermediate build artifacts. This makes an 'apples-to-apples' comparison challenging.

Xerces-C++ as a Test Case

To ensure a fair comparison, a simpler project was chosen where the same object files and binaries could be produced with identical compile and link options. Xerces-C++, an XML parser/serializer for C++, was selected. This project includes 299 C++ translation units linked into a shared library, making it a suitable, non-trivial test case.

Initial Ninja Performance

For a full, from-scratch build of Xerces-C++, Ninja completed the process in an average of 3.429 seconds across 10 runs. This establishes the baseline performance against which build2 will be measured.