Part V Compiler and Toolchain

• Chapter 15 The Compiler Pipeline
• Chapter 16 Tooling and Observability
• Chapter 17 Modules and Dependencies

This sentence comes from Thompson’s Turing Award lecture. He went on to point out that no amount of source-level verification or scrutiny protects against using untrusted code, because programmers inspect the source but run the compiler’s output. Source code becomes a runnable program only after passing through an entire toolchain; the compiler, the linker, the build system and the transitive closure of dependencies each influence the program’s behavior, and whether it can be trusted. One of Go’s responses to this problem is that, since Go 1.21, the toolchain is fully reproducible, so an officially released toolchain can be rebuilt from source and compared (3.3).

This part discusses Go’s toolchain in three chapters. Chapter 15 follows the compiler pipeline and examines how source code becomes machine code through lexical and syntactic analysis, intermediate representations and optimization, along with mechanisms at the boundary between compiler and runtime such as escape analysis and cgo. Chapter 16 covers tools for correctness and performance: deadlock detection, race detection, execution tracing, testing, benchmarking and profiling, runtime statistics, and gopls, which provides language services to editors. Chapter 17 covers modules and dependency management: how Go uses minimal version selection to obtain reproducible builds, and how go.sum and the checksum database ensure that everyone downloads the same code.