fix(training): make policy evaluation portable

TRAINING.md

@@ -139,6 +139,14 @@ appends to `training/eval_history.json` — the long-term progress record.
 Evaluate each new candidate against the previous promoted bot and a fixed
 early reference to see absolute progress over time.
 
+The evaluator launches Godot with the explicit headless display driver,
+Compatibility renderer, dummy audio driver, and a temporary writable log path
+so evaluation is reproducible on machines where the default renderer or
+`user://` log location is unavailable. A two-seed smoke check on 2026-09-01
+completed 10 paired episodes per seed for the current Stage 6 export versus
+`hard.json` (both runs finished 3–5 with 2 draws); this is only a runtime
+smoke and is not promotion evidence for the Stage 6 gate.
+
 If a model was trained with the locomotion mask on (curriculum stages 1, 2,
 and 5 — see below), pass `--grounded-a`/`--grounded-b` for whichever side it's on.
 The eval otherwise runs `AIShipController` fully unmasked regardless of how a

training/evaluate.py

@@ -17,6 +17,7 @@ import json
 import os
 import pathlib
 import subprocess
+import tempfile
 
 TRAINING_DIR = pathlib.Path(__file__).resolve().parent
 GAME_DIR = TRAINING_DIR.parent / "Game"
@@ -37,6 +38,14 @@ def run_half(
 ) -> dict:
     cmd = [
         godot_bin,
+        "--display-driver",
+        "headless",
+        "--rendering-method",
+        "gl_compatibility",
+        "--audio-driver",
+        "Dummy",
+        "--log-file",
+        str(pathlib.Path(tempfile.gettempdir()) / "cosmic-clash-evaluate-godot.log"),
         "--path",
         str(GAME_DIR),
         TRAINING_SCENE,

training/test_evaluate.py

@@ -55,6 +55,14 @@ class EvaluatePairTests(unittest.TestCase):
         command = run_process.call_args.args[0]
         self.assertIn("--eval_team_size=2", command)
 
+    @patch("evaluate.subprocess.run")
+    def test_run_uses_portable_headless_renderer_and_writable_log(self, run_process) -> None:
+        run_process.return_value.stdout = 'EVAL_RESULT {"episodes": 2, "goals_a": 1, "goals_b": 0, "draws": 1}\n'
+        evaluate.run_half("godot", "a", "b", 2, 16, 9)
+        command = run_process.call_args.args[0]
+        for option in ("--display-driver", "headless", "--rendering-method", "gl_compatibility", "--audio-driver", "Dummy", "--log-file"):
+            self.assertIn(option, command)
+
     @patch("evaluate.run_half")
     def test_2v2_evaluation_preserves_side_swap_and_team_size(self, run_half) -> None:
         run_half.side_effect = [