fix(PPO_Bot): SIGSEGV crash fixes + static buffers for thread safety

- bullets: seq[InFlightBullet] → array[4, InFlightBullet] + bulletCount
  (eliminates cross-thread heap realloc under ORC)
- hasFired: edge-triggered (cleared after state build, not level-triggered)
- round_counter parseInt: wrapped for empty/torn file → 0
- Static SVG + intent buffers to kill cross-thread heap realloc
- Tick-local alive/bulletData also fixed arrays

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

PPO_Bot/tests/test_state.nim

@@ -84,7 +84,7 @@ block testStateVectorLength:
     arenaWidth: 800.0, arenaHeight: 600.0,
   )
   let sv = buildStateVector(bot, t)
-  check sv.shape == [44], "state vector has 44 elements"
+  check sv.shape == [56], "state vector has 56 elements"
 
 block testStateVectorRange:
   var t = initEnemyTracker()
@@ -95,7 +95,7 @@ block testStateVectorRange:
     arenaWidth: 800.0, arenaHeight: 600.0,
   )
   let sv = buildStateVector(bot, t)
-  for i in 0 ..< 44:
+  for i in 0 ..< 56:
     check sv[i] >= -2.0f32 and sv[i] <= 2.0f32,
       &"sv[{i}]={sv[i]} out of [-2,2] range"
 
@@ -158,5 +158,28 @@ block testHistoryPaddedWhenEmpty:
   # indices 42-43 (goto inputs) default to 0 when not provided
   check sv[42] == 0.0f32, "sv[42] (remainingGotoDistance) should default to 0"
   check sv[43] == 0.0f32, "sv[43] (remainingGunAngle) should default to 0"
+  # indices 44-55 (bullet slots) default to 0 when no bullets provided
+  for i in 44 ..< 56:
+    check sv[i] == 0.0f32, &"bullet slot {i} should be 0 when no bullets"
+
+block testBulletSlots:
+  var t = initEnemyTracker()
+  t.update(400.0, 300.0, 0.0, 0.0, 100.0)  # enemy at (400,300)
+  let bot = BotStateData(
+    x: 200.0, y: 200.0, direction: 0.0, speed: 0.0, energy: 100.0,
+    gunDirection: 0.0, gunHeat: 0.0,
+    arenaWidth: 800.0, arenaHeight: 600.0,
+  )
+  # Bullet at (300,250), power 1.0 → speed = 20-3 = 17
+  # relX = 300-400 = -100, relY = 250-300 = -50
+  # dist to enemy = sqrt(100^2+50^2) ≈ 111.8, ticks ≈ 111.8/17 ≈ 6.58
+  let b = BulletData(x: 300.0, y: 250.0, power: 1.0)
+  let sv = buildStateVector(bot, t, 0.0, 0.0, [b], 1)
+  check abs(sv[44] - (-100.0/800.0).float32) < 0.001f32, "bullet relX"
+  check abs(sv[45] - (-50.0/600.0).float32) < 0.001f32,  "bullet relY"
+  check abs(sv[46] - (17.0/20.0).float32) < 0.001f32,    "bullet speed norm"
+  # second slot should be zero-padded
+  for i in 48 ..< 56:
+    check sv[i] == 0.0f32, &"unused bullet slot {i} should be 0"
 
 echo "All tests passed"

PPO_Bot/tests/test_training.nim

@@ -149,4 +149,27 @@ block testPpoUpdateNormalReward:
   check m.valueLoss == m.valueLoss, "valueLoss NaN on normal-reward round"
   check m.gradNorm  == m.gradNorm,  "gradNorm NaN on normal-reward round"
 
+# ── logStd ceiling: raw param must never drift above the collection clamp ─────
+# Regression for the train/collection std mismatch: logStd starting above the
+# ceiling (as in the warm-start snapshot) must be clamped back to [floor, ceiling]
+# by the first Adam step, so recomputed logP matches the acting policy's std.
+
+block testLogStdCeilingClamp:
+  randomize(45)
+  var ac = initActorCritic()
+  ac.logStd = newTensor[float32](ACTION_DIM).map(proc(v: float32): float32 = 3.0'f32)
+  var buf = initTrajectoryBuffer()
+  for _ in 0..<16:
+    let s = randomNormalTensor[float32](STATE_DIM)
+    let a = randomNormalTensor[float32](ACTION_DIM)
+    let lp = ac.computeLogProb(s, a)
+    buf.add(Transition(state: s.stateToArr, action: a.actionToArr, logProb: lp,
+                       reward: 0.1'f32, value: 0.5'f32))
+  var adam: ACAdamStates
+  discard ppoUpdate(ac, buf, lastValue = 0.0'f32, adamStates = adam,
+                    epochs = 1, miniBatchSize = 16)
+  for v in ac.logStd:
+    check v <= logStdCeiling + 1e-6'f32, "logStd above ceiling after ppoUpdate"
+    check v >= logStdFloor - 1e-6'f32, "logStd below floor after ppoUpdate"
+
 echo "All tests passed"