nx_footcheck.nx

nx_footcheck.nx -- ★THE DISTAL-LIMB STRUCTURE JUDGE (beyond-metahuman program). WHY THIS EXISTS, and it is a correction of my own work. I filed seq1038 claiming the toes had NO part header of their own and that the emitter therefore lofted ONE continuous ribbon weaving foot->toe1->toe2. That was WRONG. Reading the GENERATED canon (knowledge/canon_gx34r.dat) shows six part headers in the foot's rotated frame -- one tube plus five separate toe tubes -- exactly as emitDigitP intends. I had misread my own probe output: the probe printed the OWNING part beside each ring, and because all six headers are BYTE-IDENTICAL (`P 1 90 48 40 -41 0`, since every toe inherits the foot frame and carries its spread in the ring xoff) I read six identical headers as one repeated header. ★★★THE LAW THAT COST ME: READING THE GENERATED DATA IS NOT ENOUGH IF THE THING YOU ARE COUNTING IS INDISTINGUISHABLE FROM ITS NEIGHBOUR. Print the INDEX, not just the row. A judge that a human eye has to disambiguate is a judge that will be misread -- so the answer is this organ: the count is MEASURED. ★AND THE REAL DEFECT, which the numbers give up immediately once you subtract them: foot tube runs y 0 -> 145 toes are based at y 118 and reach y 152 (hallux) so 27 of each toe's 34 units are BURIED INSIDE THE FOOT TUBE. Only 7 units of the hallux clear the tube at all, and the fifth toe ends at 128 -- seventeen units SHORT of the tube's own tip, entirely invisible. That is the "blunt hoof with pale nubs" the render shows. nx_body_proc even carries the correct instruction in a comment -- "toes must REPLACE the flat toe-plate, not extend past it" -- and only the second half was ever applied: the toes were shortened to fit inside a plate that should have been deleted. ★A HALF-APPLIED FIX IS INDISTINGUISHABLE FROM NO FIX, AND IT LOOKS LIKE A DONE ONE IN THE SOURCE. A second, smaller defect falls out of the same subtraction. Read along the foot the quantity (zoff + rb) -- the sole side of every ring -- and the tube holds a CONSTANT 40: a flat sole, correct. The toes give 43. The toes hang three units BELOW the sole plane, so the figure's ground contact is its toe tips. WHAT IT MEASURES (all from the canon text, no rendering, no mesh, no third party): burial tube_end - toe_base must be <= 0: the tube must END where the toes BEGIN protrusion foot_len - tube_end must be >= 20% of foot length -- toes are a fifth of a foot sole break max(zoff+rb) toes - same tube must be <= 0: no toe may pierce the sole plane foot length max y over the WHOLE group ★INVARIANT -- silhouette-critical, must not move when fixed The last one is the regression guard the source comment asked for and never got: a previous round grew the toes to y 174, the height-normalised side silhouette fell 831 -> 723, and the lesson was written down as prose instead of as a test. Prose does not fail a build. nx_footcheck check <canon.dat> [rot] -> the measured report + verdict for a real artifact nx_footcheck selftest -> gv gate: proves the JUDGE discriminates, on fixtures license_tier: ORIGINAL expect_exit: 0 No hw writes (Rule 26).