nx_incclosure_lib.nx

nx_incclosure_lib.nx -- ONE OWNER for import-closure NAME->PATH resolution (2026-08-25, rung LN10). WHY THIS IS A LIB AND NOT A SECOND COPY. nx_inc_compile (the per-module incremental build cache) must answer, for each module named in a recorded closure, "is the file the compiler would read still the same bytes?" -- and it can only ask that of a PATH. nx_closurehash already owns that resolution and the order is load-bearing: runtime/_hdl_build/<name> FIRST (buildrun compiles that twin when both exist) runtime/<name> SECOND runtime/<subdir>/<name> a getdents64 sweep, looped until the kernel says stop A SECOND RESOLVER THAT DISAGREES BY ONE PATH HASHES THE WRONG FILE, REPORTS A CACHE HIT, AND SERVES A STALE ARTIFACT. That is a silent miscompile -- the worst class this estate tracks, and the one a cache exists to risk. So the resolver is EXTRACTED here and BOTH organs call it: disagreement is impossible by construction rather than by discipline, which is the estate's own standing answer whenever two organs must agree. nx_closurehash keeps its function names and DELEGATES, so its callers are untouched and its neutrality is provable by nx_behaveprobe (live vs staged) rather than asserted. license_tier: ORIGINAL No hw writes (Rule 26).