PyTorch simplifies CMake build with 9-stage overview in PR #193615

PyTorch, the 102k-star deep learning framework, merged a developer-experience PR that adds a high-level map of its build system. The root CMakeLists.txt—a sprawling orchestration file—previously had no entry point explaining what happens when. Pull request #193615, authored with help from Claude Code, inserts a comment-only overview at the top listing nine build stages in execution order: bootstrap and env-var forwarding, pre-build submodule/NCCL setup, options, dependencies, codegen, the main build, export plus optional binaries/JNI, Python packaging, and the configuration summary. Each stage names the specific cmake/ module or subdirectory that implements it, so a contributor can jump straight to the relevant code instead of reading the entire file.

The commit also updates the `spin develop` and `spin install` docstrings to reference both the new overview and `cmake/EnvVarForwarding.cmake`, which is the canonical place to look for environment variables that configure the build. Notable exceptions are called out: third-party paths like `MKLROOT` and `ROCM_PATH`, plus a few niche flags read directly at their usage sites. The changes were verified with `BUILD_CONFIG spin develop`, `spin develop --help`, `spin install --help`, and `spin lint`. Since it's comment-only, there's zero impact on compile behavior, but the payoff is real: new contributors and even veteran maintainers can now understand the build pipeline in minutes rather than digging through thousands of lines of CMake logic.