2026-09-01 - sysctlbyname implemented; kern.osrelease answers (obSCEne osrelease passes)

Next obSCEne divergence (D446 list): 135-sysctl/osrelease failed because sysctlbyname was unimplemented and the stub refused - and a refused osrelease is what turns firmware detection off in a title that asks. Implemented sysctlbyname + a pure sysctl_value: kern.ostype="FreeBSD" (the citable FreeBSD-derived fact), kern.osrelease=the configured machine().kernel_release (empty until a profile sets it - not invented; a console measured "0.0-prototype"), everything else refused (kern.version/hw.ncpu are a measured banner and a core count orbistoun does not carry honestly). POSIX contract, never overruns the given length; too-small buffer → ENOMEM (added errno::NO_MEMORY=12). Oracle: 135-sysctl/osrelease → pass (as on console); 135-sysctl/names → partial (answers ostype+osrelease, honestly refuses the other two). Distinct fails 5→4, no regressions. Unit test pins sysctl_value. Kernel/core tests pass, clippy clean, fmt-clean (D447).

Left: 137-kernelcall/system-version skips on hardware (its getpid gadget didn't resolve there) but fails under orbistoun (getpid resolves, reaches the raw syscall) - a resolving/syscall divergence, not a knob. Remaining fails: 110-modules (one-module gap), 900-surface/control (resolver reports a non-existent symbol present).