The previous commit found+drew contacts but they landed off-radar. Debugged with
BT_MAP_LOG entity/blip tracing: the reads were CORRECT all along -- the enemy mech
(cls 0xBB9, own=0) sits at dsq=14400 (exactly 120u) from the player (own=1, dsq=0),
and the grid also carries ~311 cls-0x5E vehicle props per cycle (the map's
scattered vehicles) which the 0xBB9 blip filter correctly skips.
The bug was the PROJECTION: worldToView is a camera (view->world) matrix and
Point3D::Multiply bakes a scale that is wrong for a point transform -- a 120u
contact at ppm=0.366 projected to |167px| instead of |44px| (~3.8x), and with a
spurious rotation. (The earlier "377-828m" reading was this wrong scale amplifying
the varying player-drift distance, not a distance error.) Fixed by projecting
delta = (entity - viewpoint) DIRECTLY: rotate by -heading (from the orientation
quaternion's yaw, top-down heading-up) and scale by pixelsPerMeter. Now |screen|
== |delta|*ppm exactly (blip @(0,44) for the 120u enemy), on-radar inside the view
wedge. Blip is a cross + box so it reads as a distinct contact marker.
Verified live (BT_DEV_GAUGES): the enemy contact renders as a box marker INSIDE
the FOV wedge (render-confirmed via a 4x radar crop); combat un-regressed (TARGET
DESTROYED after 8 hits), heap-clean under BT_HEAPCHECK. BT_MAP_LOG traces blip
coords + scale + delta. Remaining map polish: blip-vs-wedge rotation convention
(the enemy reads ahead/in-wedge, which is correct); the authentic pip symbol/name
infrastructure (still stubbed); SetTargetRange (radar zoom, 1km default).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
