Fix compile-validation (shared evidence):
- All five fix.diff files apply cleanly: `git apply --check -p1 < fix.diff` => OK.
- A single `make -j6 nativekernel KERNCONF=X86_64_GENERIC` on the audit source
  tree with ALL FIVE fixes applied completed with `=== NK_DONE rc=0 ===` and
  ZERO compiler errors across the kernel proper + the full module set
  (incl. the entire amdgpu DC stack and the radeon module).
- Translation-unit compile confirmations (grep of the build log):
    isp.o      (DF-1125, in kernel proper)        -> compiled
    if_bwn.o   (DF-1131, in kernel proper)        -> compiled
    gfx_v8_0.o (DF-1133 + DF-1134, amdgpu module) -> compiled
    si_dpm.o   (DF-1136, radeon module)           -> compiled
- Targeted module-TU compiles (independent, -Werror clean):
    `make gfx_v8_0.o` in sys/dev/drm/amd/amdgpu -> gfx_v8_0.o built (116648 B)
    `make si_dpm.o`   in sys/dev/drm/radeon     -> si_dpm.o   built (86880 B)
- The harness "WITH FIX" pass for each finding shows 0 OOB cases remain with the
  corrected logic (clamp / bound / validate), confirming the fix closes the
  primitive at the algorithm level.
- Boot-test of the stripped patched kernel could not be completed: copying
  kernel.stripped over /boot/kernel/kernel produced a loader-format error
  ("Unable to load /kernel/kernel / don't know how to load module 'kernel'"),
  a build/install artifact unrelated to the small validation-check fixes
  (the bug paths require FC HBA / Broadcom WiFi / AMD GPU hardware absent on
   the QEMU guest and cannot be exercised regardless).
=> fix_status = not_testable (HW-gated PoC + stripped-kernel boot artifact);
   apply-check + compile (rc=0) + harness fix-demo + source trace all pass.