vwemirror/BT411
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Gitea #48: arm the artifact tripwires BY DEFAULT + surface detections through the matchlog

Browse Source 
The user's ask: if the translation-table fix did not quite catch it, will the
next playtest LOG the artifact?  Before this commit, no -- BT_PLANE_AUDIT was
opt-in via env var, and playtest machines set none.

Changes (all engine L4VB16):
  * BT_PLANE_AUDIT is now DEFAULT-ON, =0 opts out.  Armed cost is a couple of
    integer compares per DRAW CALL (not per pixel); confirmed zero log noise
    on a clean default-env run post-fix.
  * A SECOND trap class: out-of-bounds draw STARTS, checked always-on in
    buildDestPointer (the funnel every primitive builds its pointer through).
    Release builds compile Verify() to nothing, so a wild start previously
    computed a silent rogue pointer into (or past) the shared buffer.  Now
    logged ([plane] OOB) and CLAMPED in-bounds.
  * First detection of each class writes a GLITCH matchlog record
    (kind=plane / kind=oob with position+color+mask), so the evidence arrives
    with the round logs the operator already collects -- unattended.
  * Both existing plane-leak trap levels (display primitives + the port-level
    pixmap table scan) flipped to the same default-on gate.

Honest coverage note (in gauges-hud.md): an in-buffer EXTENT overrun from a
valid start (a blit walking past the right/bottom edge into later rows of the
SAME plane) is not trapped -- per-pixel walk checks are too hot.  The plane
trap still catches any such overrun whose color leaks planes, which is the
class the night-3 artifacts belonged to.

Verified: full clean rebuild -- 40 pre-existing LNK2019 only (the
engine->game BTMatchLog linkage resolved, /FORCE-trap checked); default-env
probe run: 0 [plane] lines, game unaffected.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
arcattack
2026-07-25 17:28:13 -05:00
co-authored by Claude Fable 5
parent 6bb03aed0b
commit 1d31af61d7
3 changed files with 69 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
context/decomp-reference.md
+1 -1
View File
@@ -419,7 +419,7 @@ default-ON (`'0'` disables).
| `BT_GAUGE_SKIP_LOG` | `[gskip] <primitive>` per unregistered gauge widget the parse skips |
| `BT_VALVE_LOG` | condenser valve flow distribution |
| `BT_BAYTEST=<frame>` | #46 rig: send message 1 (crit-induced cook-off arm) to the first AmmoBin at the given sim frame — the 10 s fuse then runs live (`scratchpad/baytest.py` / `baypurge.py`) |
| `BT_PLANE_AUDIT` | #48 forensics: log every gauge draw whose color carries bits outside its port's plane mask (the cross-display leak trap; `[plane] LEAK/PORT` lines) |
| `BT_PLANE_AUDIT` | #48 forensics -- **DEFAULT-ON** (`=0` opts out): logs any gauge draw whose color carries bits outside its port's plane mask (`[plane] LEAK/PORT`) AND any out-of-bounds draw start (`[plane] OOB`, clamped). First detection of each also writes a `GLITCH` **matchlog** record, so playtest evidence arrives with the round logs even unattended |
| `BT_LOOP_AUDIT` | `[loop-audit]` per-sound `sample-flag × source-render-type -> AL_LOOPING` (Gitea #51; rig `scratchpad/loopaudit.py`) |
| `BT_LOOP_LEGACY=1` | restore the OLD always-loop rule (`AL_LOOPING = sample != ForceStatic`) for a field A/B — see the loop-flag note in [[wintesla-port]] |
| `BT_AUDIO_DUMP` | once-a-second `[playing]` dump of every playing AL source + gain/pitch/**loop** — catches a stuck looping source red-handed |