mirror of
https://github.com/Kelsidavis/WoWee.git
synced 2026-08-01 10:21:34 -04:00
2d9d7576ed
Cars in the same 3-car train seed their client-animation phase from
nowEpochMs() % their own path's durationMs. Two of the six DBC path
entries (176080, 176084) report a durationMs ~102ms longer than their
four siblings; taking modulo of a huge wall-clock epoch value by two
close-but-different divisors produces effectively uncorrelated
remainders, so those two cars end up on the opposite side of the loop
from siblings registered a second earlier ("solo cars"). Round the
divisor to the nearest 500ms for Deeprun tram seeding only, so sibling
cars' seeds stay correlated by real elapsed registration time again.
Also:
- Widen the boarding capture radius from 7x6 back to 9.5x8 (7x6 was
overcorrected - stopped attaching almost entirely) and log the
nearest-miss distance on rejection so future tuning has real numbers
instead of guessing blind again.
- Clamp (rather than fully flatten) the tram's pitch on grade changes:
full flattening fixed the earlier nose-dive complaint but left the
car level while its position still follows the real track elevation,
visually clipping into the sloped tunnel floor on hills.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>