feat(pathfinder): centralize CUDA env var handling, prioritize `CUDA_PATH` over `CUDA_HOME`#1801

Draft

rwgk wants to merge 12 commits intoNVIDIA:mainfrom

rwgk:CUDA_PATH_CUDA_HOME_cleanup

Commits on Mar 20, 2026

Squash-merge of PR #1519 (rparolin/env_var_improvements) rebased onto main.

rwgk
and
rparolin
committed
replace _get_cuda_paths() with _get_cuda_path() using pathfinder
rwgk
committed
treat empty env vars as undefined in get_cuda_home_or_path()
rwgk
committed
fix(pathfinder): clear get_cuda_home_or_path cache in test fixtures
rwgk
committed
fix(core): update test_build_hooks for _get_cuda_path rename, drop multi-path test
rwgk
committed
refactor(core): use get_cuda_home_or_path() in test conftest skipif
rwgk
committed
refactor(core): use get_cuda_home_or_path() in examples
rwgk
committed
rename get_cuda_home_or_path -> get_cuda_path_or_home
rwgk
committed
make get_cuda_path_or_home a public API, privatize CUDA_ENV_VARS_ORDERED
rwgk
committed
docs(pathfinder): manually edit 1.5.0 release notes, fix RST formatting (Cursor)
rwgk
committed
Add 1.5.0, 1.4.3, 1.4.2 in cuda_pathfinder/docs/nv-versions.json
rwgk
committed
docs: clarify that CUDA_PATH/CUDA_HOME priority comes from pathfinder
rwgk
committed