nx_dwline_roundtrip_gate.nx

nx_dwline_roundtrip_gate.nx -- DOES OUR DWARF ENCODER PRODUCE A TABLE OUR DECODER READS BACK? WHY THIS EXISTS. DDR-002 shipped .debug_line (compiler emits .file/.loc -> nxasm builds the line program -> ELF carries the section) and named its own falsification condition: "If nx_addr2line cannot round-trip a table nx_dwline_* produced, the two organs disagree about the format and the arc is blocked on reconciling them." That condition was NEVER TESTED. The capability had no gate at all, and by this estate's own doctrine an ungated capability cannot be honestly graded on /compare -- whose subtitle promises "Nishi cells measured (gates with dates)". This file is that measurement. WHY IT IS A REAL TWO-SYSTEM CHECK. nx_dwarf_line.nx (encoder) and nx_addr2line_lib.nx (decoder) share NO code and deliberately repeat the DWARF opcode constants separately. So agreement here is evidence, not a system agreeing with itself. Do not "clean up" that duplication. WHY THESE ADDRESSES. 0x400096 -> line 5 and 0x4000AD -> line 9 are the exact pairs an EXTERNAL oracle (GNU readelf) resolved from a real sovereign ELF built from nx_probe_dbgtiny.nx. Pinning the gate to externally-confirmed values means a regression cannot be hidden by changing both sides at once. CALL SEQUENCE IS COPIED FROM PRODUCTION (nxasm_x86_main.nx: new -> header -> set_address -> record* -> end_sequence -> finish). A gate that exercised a different sequence would be measuring something other than what ships -- the instrument-subject mismatch this estate keeps paying for. license_tier: ORIGINAL expect_exit: 0