erik/acdream
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

revert(render): near plane back to 1.0m pending #110 - 0.1 correlated with missing indoor textures

Browse source 
Bisect (user-gated): two consecutive runs on 0.1 lost indoor textures; the 1.0 bisect run rendered clean. #105 tripwires silent on the bad runs (GL-side). No known mechanism links the near plane to texturing - #110 filed to investigate (RenderDoc / flip-testing) before re-landing retail's znear=0.1, which the corner see-through fix depends on. Comments on all four cameras point at #110 so the retail value is not re-landed blind.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
Erik 2026-06-10 11:02:55 +02:00
parent 137b4f2d25
commit 8bd3492612
5 changed files with 40 additions and 13 deletions
Download patch file Download diff file Expand all files Collapse all files

25
docs/ISSUES.md
View file

@ -3768,6 +3768,31 @@ stable now; this is a draw-order/depth oscillation localized to the door surface
---
## #110 — Near plane 0.1 m (retail znear) correlates with missing indoor textures; corner see-through blocked on it
**Status:** OPEN
**Severity:** HIGH (blocks the §4 corner see-through fix)
**Component:** render / camera projection (+ possibly texture upload / #105 interaction)
Retail runs `Render::znear = 0.1` (decomp :342173, initializer :1101867); ours is 1.0 m.
Because the camera-collision sphere holds the eye 0.3 m from walls, the 1.0 m near plane
clips away any wall the camera presses against — the §4 "camera clipping into the wall"
corner background (user-gated 2026-06-10, unchanged by the flood fix `dac8f6a`).
Landing 0.1 (`137b4f2`) fixed the geometry relationship but correlated with MISSING INDOOR
TEXTURES on two consecutive runs; the 1.0 bisect run immediately rendered clean, and the
change was reverted same-day. The #105 dat-miss tripwires were SILENT on the bad runs
(GL-side per the #105 protocol). No mechanism is known by which the near plane affects
texturing — candidates: (a) coincidence with the intermittent #105 (2/2-then-1/1 is a small
sample), (b) a depth-precision interaction (near 0.1 + far 5000 = 50k ratio) breaking a
depth-dependent pass indoors, (c) something in the visible-set/prepare path scaling with
the wider near frustum. Investigation: re-land 0.1 locally, reproduce, RenderDoc the
missing-texture frame; or flip-test 0.1↔1.0 over N runs to settle (a) statistically.
User question pending: did the missing textures render WHITE (upload path) or INVISIBLE
(visibility/depth)?
---
---
# Recently closed