Which method is preferred for managing compiler overrides in shared HPC environments?

Spack utilizes compiler.json files to store paths, flags, and versioning information for detected compilers. While spack compiler find automates the creation of these entries, HPC environments often feature compilers installed in non-standard locations or specific module-loaded paths that auto-detection may overlook.

In shared environments where multiple users require different compiler variants or specific flag overrides, there is uncertainty regarding the optimal balance between global configuration and user-specific overrides. Relying solely on spack compiler add can lead to fragmented configurations across different user home directories.

• Does Spack support a hierarchical inheritance model for compiler.json to allow site-wide defaults with user-level overrides?
• What is the recommended approach for ensuring consistency in compiler specifications across a shared cluster without requiring every user to manually run detection?