[nicolas-maman]

Kneeling by a point on the ground to work at it with both hands, and up again (#825's 5). Fixing it showed that the keyed ways asked the box man's joints for more than they have, and stepped feet badly, so this also fixes that for every keyed way (#841, in part).

Kneeling

motion.kneel(body, at, facing) kneels a figure by a point on the ground. kneel_spot says where to stand for it, and kneeling, knelt and work_at report on it.

• Down: the figure steps back onto its left knee, then brings the right knee down beside it, upright, and only then leans over the work. The hands move over the work as hands treating a wound do.
• Up: stand_up rises the way the get-up's half kneel does.

tests/test_kneel.ae:

	measured	held to
knelt, the left knee's flesh over the ground	1.9 cm	under 2 cm
the hands' middle from the work	3.0 cm	5 cm
toes sliding while they bear, all the way	4.0 mm	5 mm
the lowest drawn flesh	0.7 cm in (a standing foot)	1 cm
the worst body turn in a step	5.4°	6°

Every keyed way

• Joint ranges. A solve keeps the knee or elbow inside its range, and a stepped foot inside its ankle's cone. The plans had asked for a 153° knee against the box man's 130°. The bodies stopped at their limits and dragged a foot 26 cm under a foot drawn floating.
• The stage eased onto the animation. It now puts each leg's end back where its solve had it, keeping the knee hinge's lie. Eased bone by bone, a back leg's ankle dipped under both of its keys, 8.3 cm into the ground.
• Starting from the drawn pose. A way begun while acting starts from what is drawn, not from the bodies. Standing up from kneeling, the neck snapped 58° and the arms 29° on the first frame.
• Stepped feet:
  • they lift, travel at an even pace and set down level;
  • once off the ground they ride their shins;
  • while carried they stay 3 cm clear;
  • where the leg can't lift them, they turn their toes up and swing out round the folded knee, both planned over the stage.
• Settling after a way up. The stance shifts at 4 mm a step, under the 5 mm slide bound.
• Getting up from the front. The kneeling feet tuck their toes under. The drawn feet went 8.8 cm into the ground and now go 0.7 cm. tests/test_get_up_feet.ae fails on main.
• The knee's range is 149 degrees, up from 130. That is a person's bend crouched or kneeling; 130 is their bend standing. At 130, the get-up's crouch over the feet asked the knees past their range, and the knee clamp pushed the planted feet 6.7 cm into the ground. A crouch lifted to fit 130 turned a forearm 6.2 degrees a step, and leaning it instead 7.1, against main's 4.6. At 149 the crouch fits as keyed, and face up's worst turn is 5.9 degrees. A crouch that still asks for more raises its hips to fit (crouch_within).

Holds and follow-ups

• get-up: the drawn feet go up to 8.8 cm into the ground; test_get_up holds only the bodies #841 stays open. Up from the back, the first stage still draws the feet 3.6 cm into the ground, as main does. Face down, that stage draws the forearm 8 cm in, which the feet had hidden. Both come from the unplanned first stage's world slerp.
• The pins stay at Aether 0.834.0 and aephysics f0bd6d6. On 0.837.0 every program compiles except three. A program's own function shadows a module's import of the same name, and test_taa, test_dlss and test_ssr hit it. The repro is with the Aether session for batch 5. The migration waits on branch feat/aether-0838, and ci.yml says so.
• motion.hold: the drawn hands are on the grips but the forearm bodies are 13 to 50 cm off them (held off by the thighs) #843, the held forearm bodies off their grips, comes next.

The motion gate passed all 41 suites, kneel included.