PROCITY/tools/flags_check.py
m3ultra 3b86b0c5b3 Lane F R6: ≤350 interior draw gate + fix interior renderer.info accumulation
Added an interior draw-budget sweep to flags_check.py (decision #2): enter all 9 shop types,
assert <=350 draws/room, dig on so it also confirms C1's batching kept ?dig=1 riffling.

The sweep caught a real bug in interior_mode.js (F-owned): the interior/dig render never called
renderer.info.reset() while the shell runs autoReset=false (for the composer), so
renderer.info.render.calls ACCUMULATED every frame — inflating any interior draw read (and the
HUD were it shown). Fixed: reset per interior frame before both the dig and the walk render.

True per-frame result after the fix: all 9 room types <=350 draws (worst dept 96) — C1 batching
(35981ad) verified. flags_check GREEN (regression + interior budget + 4 flag smokes + all-on combo).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-07-14 23:10:11 +10:00

254 lines
13 KiB
Python

#!/usr/bin/env python3
"""PROCITY Lane F — v2 flag enforcement harness (round-6 F1).
The machine version of what the R5 review did by hand. Three things, headless, <2 min:
1. FLAGS-OFF REGRESSION — the enforcement arm of the prime law. Boot with no flags → the plan
fingerprint must equal the golden 0x3fa36874, the settled street view must stay within the v1
budget (≤300 draws / ≤200k tris) AND within tolerance of the v1.1 snapshot, every v2 subsystem
must report inactive, and there must be ZERO console errors. A v2 flag leaking into the default
boot trips this.
2. PER-FLAG SMOKE — for each landed flag (winmap, dig, roster, +plansrc when A lands): boot flag-on,
cross ≥2 chunks, enter one interior (dig: also open+close a riffle), 0 console errors.
3. ALL-ON COMBO (flag-composition law, decision #4): every landed flag on at once, same smoke.
Run: cd web && python3 -m http.server 8130 (or let this script spawn its own)
tools/.venv/bin/python tools/flags_check.py
Wired into tools/qa.sh as a WARN-level gate this round (strict in round 7). Exit 0 = all pass;
non-zero = a check failed (qa.sh reports it as a warning this round).
"""
import sys, json, time, socket, subprocess, pathlib
try:
from playwright.sync_api import sync_playwright
except ImportError:
sys.exit("playwright not installed — python3 -m venv tools/.venv && "
"tools/.venv/bin/pip install playwright && tools/.venv/bin/python -m playwright install chromium")
ROOT = pathlib.Path(__file__).resolve().parent.parent
PORT = 8130
HOST = f'http://127.0.0.1:{PORT}'
SEED = 20261990
GOLDEN_HASH = 0x3fa36874 # frozen determinism anchor (Lane A selfcheck)
BASE_DRAWS, BASE_TRIS = 163, 19000 # flags-off DBG.shot('street_noon') snapshot (v1.1, measured); ±40% warn tolerance
STREET_DRAWS_MAX, STREET_TRIS_MAX = 300, 200_000
KNOWN_FLAGS = ['winmap=1', 'dig=1', 'roster=stream'] # plansrc=osm auto-appended if Lane A landed it
fails, warns = [], []
def FAIL(m): fails.append(m); print(f" \033[31m✗ FAIL\033[0m {m}")
def WARN(m): warns.append(m); print(f" \033[33m! WARN\033[0m {m}")
def OK(m): print(f" \033[32m✓\033[0m {m}")
def head(m): print(f"\n\033[1m{m}\033[0m")
def port_up(port):
with socket.socket() as s:
s.settimeout(0.4)
return s.connect_ex(('127.0.0.1', port)) == 0
def ensure_server():
if port_up(PORT):
return None # reuse a running dev server
proc = subprocess.Popen(['python3', '-m', 'http.server', str(PORT)],
cwd=str(ROOT / 'web'), stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
for _ in range(50):
if port_up(PORT): return proc
time.sleep(0.1)
proc.terminate(); sys.exit(f"could not start http.server on :{PORT}")
# ── page helpers ─────────────────────────────────────────────────────────────
def new_page(p):
b = p.chromium.launch()
pg = b.new_page(viewport={'width': 1280, 'height': 720})
errs = []
pg.on('console', lambda m: errs.append(m.text) if m.type == 'error' else None)
pg.on('pageerror', lambda e: errs.append(str(e)))
return b, pg, errs
def boot(pg, query):
pg.goto(f'{HOST}/index.html?seed={SEED}&dbg=1' + (('&' + query) if query else ''))
pg.wait_for_function("window.DBG && window.DBG.ready === true", timeout=20000)
pg.evaluate("() => { const o=document.getElementById('pc-start'); if(o) o.style.display='none'; }")
try:
pg.wait_for_function("() => window.PROCITY && (!window.PROCITY.fleet || window.PROCITY.fleet.ready)", timeout=12000)
except Exception:
pass
def plansrc_landed(p):
"""Boot ?plansrc=osm; landed iff the shell exposes an 'osm' plan source (else the flag is ignored)."""
b, pg, _ = new_page(p)
try:
boot(pg, 'plansrc=osm')
src = pg.evaluate("() => (window.PROCITY.plan && (window.PROCITY.plan.source||window.PROCITY.plan.planSource)) "
"|| window.PROCITY.planSource || null")
return src == 'osm'
except Exception:
return False
finally:
b.close()
# ── the three checks ─────────────────────────────────────────────────────────
def flags_off_regression(p):
head('FLAGS-OFF REGRESSION (prime-law enforcement)')
b, pg, errs = new_page(p)
try:
boot(pg, '')
state = pg.evaluate("""() => {
const P=window.PROCITY, D=window.DBG;
D.shot('street_noon'); // fixed pose → comparable draws/tris
const i = D.info();
const c = P.citizens || {};
return { draws:i.drawCalls, tris:i.tris, mode:i.mode,
winmap: !!(P.winmap && P.winmap.active),
digActive: !!(P.interiorMode && P.interiorMode.digActive),
streaming: !!(c.streaming || c.mode==='stream') };
}""")
# determinism anchor — run Lane A's selfcheck in-process (golden fingerprint)
r = subprocess.run(['node', 'web/js/citygen/selfcheck.js'], cwd=str(ROOT),
capture_output=True, text=True)
hash_ok = (f'{GOLDEN_HASH:#010x}'.replace('0x', '0x') in r.stdout) or (hex(GOLDEN_HASH) in r.stdout) \
or ('3fa36874' in r.stdout)
if hash_ok and r.returncode == 0: OK(f'plan fingerprint == golden {hex(GOLDEN_HASH)} (selfcheck green)')
else: FAIL(f'plan fingerprint != golden {hex(GOLDEN_HASH)} — determinism/prime-law broken')
if state['winmap'] or state['digActive'] or state['streaming']:
FAIL(f"a v2 subsystem is ACTIVE on default boot: {state}")
else: OK('all v2 subsystems inactive on default boot (winmap/dig/roster off)')
if state['draws'] <= STREET_DRAWS_MAX and state['tris'] <= STREET_TRIS_MAX:
OK(f"flags-off street view within v1 budget (draws {state['draws']}{STREET_DRAWS_MAX}, tris {state['tris']}{STREET_TRIS_MAX})")
else: FAIL(f"flags-off street view OVER v1 budget: draws {state['draws']}, tris {state['tris']}")
dd = abs(state['draws'] - BASE_DRAWS) / BASE_DRAWS
dt = abs(state['tris'] - BASE_TRIS) / BASE_TRIS
if dd <= 0.40 and dt <= 0.40:
OK(f"draws/tris within 40% of v1.1 snapshot ({state['draws']} vs {BASE_DRAWS}, {state['tris']} vs {BASE_TRIS})")
else:
WARN(f"draws/tris drifted >40% from v1.1 snapshot ({state['draws']} vs {BASE_DRAWS}, {state['tris']} vs {BASE_TRIS}) — investigate")
if errs: FAIL(f"{len(errs)} console error(s) on default boot; first: {errs[0][:140]}")
else: OK('0 console errors on default boot')
finally:
b.close()
def smoke(p, label, query, is_combo=False):
head(f'SMOKE: {label} (?{query})')
b, pg, errs = new_page(p)
try:
boot(pg, query)
# cross ≥2 chunks by teleporting between separated street nodes
chunks = pg.evaluate("""() => {
const P=window.PROCITY, D=window.DBG, seen=new Set();
const ns = P.plan.streets.nodes;
for (const idx of [0, (ns.length/3|0), (2*ns.length/3|0), ns.length-1]) {
const n = ns[idx]; D.teleport(n.x, n.z, 0);
for (let k=0;k<10;k++) (P.citizens&&P.citizens.update&&P.citizens.update(0.05));
const c = D.info().chunk; if (c!=null) seen.add(String(c));
}
return seen.size;
}""")
if chunks >= 2: OK(f'crossed {chunks} chunks streaming (flag active, no crash)')
else: WARN(f'only {chunks} distinct chunks seen (teleport/stream probe)')
# enter one interior (open shop), exit
entered = pg.evaluate("""() => {
const P=window.PROCITY, D=window.DBG;
D.setSegment(2);
const s=(P.plan.shops||[]).find(x=>x.type==='record' && P.isOpen(x)) || (P.plan.shops||[]).find(x=>P.isOpen(x));
if(!s) return {ok:false};
D.enterShop(s.id);
const inMode = D.info().mode==='interior';
return {ok:inMode, shop:s.name};
}""")
if entered.get('ok'): OK(f"entered interior '{entered.get('shop')}' (mode=interior)")
else: FAIL('could not enter an interior')
# dig-specific: open one riffle + close, leak-free
if 'dig=1' in query:
dug = pg.evaluate("""() => {
const P=window.PROCITY, THREE=P.THREE, room=P.interiorMode.current;
if(!room) return {ok:false, why:'no room'};
const bins=[]; room.group.traverse(o=>{ if(o.userData&&o.userData.kind==='bin') bins.push(o); });
if(!bins.length) return {ok:false, why:'no bins'};
const bp=new THREE.Vector3(); bins[0].getWorldPosition(bp);
P.camera.position.set(bp.x,1.6,bp.z+1.8); P.camera.lookAt(bp.x,bp.y+0.4,bp.z); P.camera.updateMatrixWorld();
window.dispatchEvent(new KeyboardEvent('keydown',{code:'KeyE'}));
const opened = P.interiorMode.digActive;
const btn=document.querySelector('.pcdg-out'); if(btn) btn.click();
return { ok: opened, closed: !P.interiorMode.digActive };
}""")
pg.wait_for_timeout(300)
if dug.get('ok') and dug.get('closed'): OK('dig: riffle opened (E on bin) + closed (WALK OUT)')
elif dug.get('ok'): WARN(f"dig: opened but close not confirmed ({dug})")
else: FAIL(f"dig: riffle did not open ({dug.get('why','?')})")
pg.evaluate("() => window.DBG.exitShop && window.DBG.exitShop()")
if errs: FAIL(f"{label}: {len(errs)} console error(s); first: {errs[0][:140]}")
else: OK(f'{label}: 0 console errors')
finally:
b.close()
def interior_budget(p):
"""Decision #2: interior draw budget is law — ≤350 draws worst room. Sweep every shop type
(dig on, so this also confirms batching kept ?dig=1 riffling). C enforces via its sweep; F pins it."""
head('INTERIOR DRAW BUDGET (≤350 draws/room law — decision #2)')
b, pg, errs = new_page(p)
try:
boot(pg, 'dig=1')
pg.evaluate("() => window.DBG.setSegment(2)") # midday → shops open
types = ['record', 'opshop', 'toy', 'book', 'video', 'pawn', 'milkbar', 'dept', 'stall']
rows, seen = [], set()
for t in types:
picked = pg.evaluate("""(t) => {
const P=window.PROCITY, D=window.DBG;
const s=(P.plan.shops||[]).find(x=>x.type===t && P.isOpen(x));
if(!s) return null; D.enterShop(s.id); return {id:s.id, name:s.name};
}""", t)
if not picked:
continue
pg.wait_for_timeout(500) # let async GLB fittings load → settled worst-case draws
info = pg.evaluate("() => { const i=window.DBG.info(); return {draws:i.drawCalls, tris:i.tris, mode:i.mode}; }")
rows.append({'type': t, 'shop': picked['name'], **info})
pg.evaluate("() => window.DBG.exitShop && window.DBG.exitShop()")
pg.wait_for_timeout(150)
over = [r for r in rows if r['draws'] > 350]
for r in over:
FAIL(f"interior {r['type']} '{r['shop']}': {r['draws']} draws > 350")
if rows and not over:
worst = max(rows, key=lambda r: r['draws'])
OK(f"all {len(rows)} room types ≤350 draws (worst: {worst['type']} {worst['draws']}, tris {worst['tris']})")
elif not rows:
WARN('no interiors sampled (no open shops of the swept types)')
if errs:
FAIL(f"interior sweep: {len(errs)} console error(s); first: {errs[0][:140]}")
finally:
b.close()
def main():
srv = ensure_server()
try:
with sync_playwright() as p:
flags = list(KNOWN_FLAGS)
if plansrc_landed(p):
flags.append('plansrc=osm'); print("· plansrc=osm detected as landed — included")
else:
print("· plansrc=osm not landed (Lane A) — skipping that flag")
flags_off_regression(p)
interior_budget(p)
for fl in flags:
smoke(p, fl.split('=')[0], fl)
smoke(p, 'ALL-ON combo', '&'.join(flags), is_combo=True)
finally:
if srv: srv.terminate()
head('VERDICT')
print(json.dumps({'fails': len(fails), 'warns': len(warns)}, indent=0))
if fails:
print(f"\033[31m● flags_check RED\033[0m — {len(fails)} failure(s), {len(warns)} warning(s).")
for f in fails: print(' FAIL:', f)
sys.exit(1)
print(f"\033[32m● flags_check GREEN\033[0m — {len(warns)} warning(s).")
for w in warns: print(' warn:', w)
sys.exit(0)
if __name__ == '__main__':
main()