Preserve bounded validation evidence and historical failures

tests/validation/multigrid/shear_v0_2_evidence/run_once.py

@@ -0,0 +1,74 @@
+"""Bounded execution recorder; refuses to overwrite evidence or repeat a run."""
+import datetime
+import hashlib
+import importlib.metadata
+import json
+import os
+from pathlib import Path
+import platform
+import resource
+import shlex
+import subprocess
+import sys
+import time
+
+evidence = Path(__file__).resolve().parent
+base = evidence.parent
+script = base / "shear_diagnostic_v0_2.py"
+reference = base / "cpu_reference.py"
+expected_env = {
+    "OPENBLAS_NUM_THREADS": "1",
+    "PYTHONDONTWRITEBYTECODE": "1",
+    "PYTHONNOUSERSITE": "1",
+    "PYTHONOPTIMIZE": "0",
+    "PYTHONPATH": str(base),
+    "CPU_REFERENCE_PATH": str(reference),
+}
+limits = {
+    "address_space_bytes": (resource.RLIMIT_AS, 268435456),
+    "output_file_bytes": (resource.RLIMIT_FSIZE, 2097152),
+    "core_bytes": (resource.RLIMIT_CORE, 0),
+}
+assert all(os.environ.get(k) == v for k, v in expected_env.items())
+assert all(resource.getrlimit(key) == (value, value) for key, value in limits.values())
+assert sys.executable == "/home/frank14f/anaconda3/envs/pycuda_3_10/bin/python"
+assert hashlib.sha256(script.read_bytes()).hexdigest() == "a8b1cf2f7abd00efc4d068bbd27bb2317144a6ca4c545d9b2e0dcf421eadebc6"
+assert hashlib.sha256(reference.read_bytes()).hexdigest() == "bba2a06a9f6517b32079ffff15e4060d8c3b7ed35f16ca134e1877226c9e4e1e"
+log_path = evidence / "stdout_horsea.txt"
+result_path = evidence / "results_horsea.json"
+assert not log_path.exists() and not result_path.exists(), "Existing execution evidence; do not rerun"
+prefix = ["/usr/bin/prlimit", "--as=268435456", "--fsize=2097152", "--core=0", "--", sys.executable, "-B"]
+command = prefix + [str(script)]
+report = {
+    "started_utc": datetime.datetime.now(datetime.timezone.utc).isoformat(),
+    "cwd": str(base),
+    "parent_command": ["env"] + [f"{k}={v}" for k, v in expected_env.items()] + prefix + [str(Path(__file__).resolve())],
+    "script_command": command,
+    "environment": expected_env,
+    "python_version": sys.version,
+    "numpy_version_from_installed_metadata": importlib.metadata.version("numpy"),
+    "platform": platform.platform(),
+    "hard_limits": {name: resource.getrlimit(key) for name, (key, _) in limits.items()},
+    "planned_runs": 7,
+    "coarse_steps_per_run": 32,
+    "attempt": 1,
+}
+with log_path.open("xb", buffering=0) as log:
+    log.write(("RECORDER " + json.dumps(report) + "\nCOMMAND " + shlex.join(command) + "\n").encode())
+    start = time.monotonic()
+    child = subprocess.run(command, cwd=base, stdout=log, stderr=subprocess.STDOUT, check=False)
+    report.update(
+        returncode=child.returncode,
+        elapsed_seconds=time.monotonic() - start,
+        child_peak_rss_MiB=resource.getrusage(resource.RUSAGE_CHILDREN).ru_maxrss / 1024,
+        parent_peak_rss_MiB=resource.getrusage(resource.RUSAGE_SELF).ru_maxrss / 1024,
+        completed_utc=datetime.datetime.now(datetime.timezone.utc).isoformat(),
+    )
+    log.write(("\nRUN_RESULT " + json.dumps(report) + "\n").encode())
+report["stdout_bytes"] = log_path.stat().st_size
+with result_path.open("x") as result:
+    json.dump(report, result, indent=2)
+    result.write("\n")
+print(json.dumps(report, indent=2))
+assert report["stdout_bytes"] < 2097152
+sys.exit(child.returncode if child.returncode >= 0 else 1)