nx_buildmode_lib.nx

nx_buildmode_lib.nx -- THE COMPILE MODE THE BUILD LANE USES, AS DATA (LN44, 2026-09-03). THE GAP THIS CLOSES, measured. No build lane in this estate passes `--mode` at all: a grep for it across the build-lane sources returns one hit and it is a GATE, not a builder. So every organ ships compiled in DEFAULT, which by deliberate contract has every declared mode OFF -- including `--bckelide`, LN7 sound bounds-check elision, which is already written, already promoted, and already proven by nx_bck_elide_gate 13/13 against the live compiler with both out-of-bounds legs still trapping 71. Cost of that omission, measured two independent ways: the published h2h receipt reads default 7105 us against f1 4189 us on the same workload (1.70x), and a 15-round interleaved re-measure on a quiet host reads 8275 us against 5210 us (1.59x). On a divide-dominated loop the same pair is 45807 against 44851, i.e. NEUTRAL -- the mode wins where it can and costs nothing where it cannot. WHY THIS IS A BUILD-LANE POLICY AND EMPHATICALLY NOT A COMPILER-DEFAULT CHANGE. nx_compile_x86 states the contract in its own source: default builds stay BYTE-IDENTICAL BY CONSTRUCTION, which is exactly what every drift, contentdiff and artifact-comparison instrument in the estate rests on. Flipping the language default would re-baseline ALL of those instruments in the same instant -- so the undo path would run straight through the comparison surface needed to judge whether the change was safe. That is the static-stability violation this estate already has doctrine about. The mode therefore moves PER TARGET, declared here, provable one organ at a time. THE SAFETY PROPERTY, AND IT IS THE WHOLE DESIGN: an ABSENT conf, an absent row, or an empty flag list yields ZERO flags AND contributes ZERO bytes to the build-cache key. Adopting this lib therefore cannot change one byte of one artifact until somebody writes a row. There is no flag day. ⚠ THE CACHE KEY IS PART OF THE CONTRACT, NOT AN AFTERTHOUGHT. nx_sov_build_run caches artifacts under sha256(expanded closure ++ compilerSha ++ asmSha ++ 'g'|'n'). A mode that changed the emitted code WITHOUT entering that key would serve a stale artifact compiled under the other mode -- silently, and looking like a cache hit. bm_key_token exists so the mode enters the key by construction, and the gate asserts that two different modes produce two different tokens. Row grammar, one per line, first match wins: target|<name>|<flags...> the mode for exactly this build target all|<flags...> the mode for every target that has no row of its own `#` and `;` start a comment. Flags are passed to the compiler verbatim and are exactly the tokens its own driver already parses -- this lib can never invent a switch the compiler does not have.