{
  "schema_version": 1,
  "experiment": "php-runtime-after-the-graph-fix-v1",
  "date": "2026-09-20",
  "task": "pre-push-safety-net 13.2",
  "claim_scope": "What remains of the PHP blocker once the malformed-CPG bug is excluded. Measured on Paid Memberships Pro, 673 files.",
  "the_graph_bug_is_fixed": {
    "issue": "joernio/joern#6281, this project's own report",
    "defect": "a PHP `global` inside a closure gave a node two AST parents, so the dataflow overlay threw and one file cost the whole graph",
    "filed_against": "4.0.623",
    "fixed_in": "4.0.625, by joernio/joern#6270",
    "the_image_pins": "4.0.625, since before any measurement recorded under this task",
    "verified_today": "Not on a transcription of the repro, which does not reproduce, but on the real trigger: PMPro's includes/fields.php, the file the original bisection found, whose add_filter callback carries a `global $submenu_file`. It builds and applies the default overlays cleanly on BOTH 4.0.625 and 4.0.631: files=2, methods=158, no ReachingDefPass assertion.",
    "consequence": "What remains under this task is a runtime failure and must not be re-diagnosed as the graph bug."
  },
  "what_remains_is_runtime": {
    "subject": {
      "files": 673,
      "regions": 4461,
      "php_regions": 4425,
      "questions_asked": 2526
    },
    "runs": {
      "pmpro": {
        "questions_total": 2526,
        "questions_completed": 0,
        "seconds": 627.7,
        "build_seconds": 144.04,
        "query_seconds": 363.41,
        "degradations": [
          "cross_partition_semantics_unresolved",
          "partition_file_census_incomplete",
          "process_timeout",
          "query_failed",
          "scratch_cleanup_incomplete",
          "shard_query_incomplete"
        ]
      },
      "pmpro-joern631": {
        "questions_total": 2526,
        "questions_completed": 0,
        "seconds": 619.1,
        "build_seconds": 134.54,
        "query_seconds": 367.32,
        "degradations": [
          "cross_partition_semantics_unresolved",
          "partition_file_census_incomplete",
          "process_timeout",
          "query_failed",
          "scratch_cleanup_incomplete",
          "shard_query_incomplete"
        ]
      },
      "pmpro-631-longceiling": {
        "questions_total": 2526,
        "questions_completed": 0,
        "seconds": 2096.5,
        "build_seconds": 113.24,
        "query_seconds": 1874.15,
        "degradations": [
          "cross_partition_semantics_unresolved",
          "partition_file_census_incomplete",
          "process_timeout",
          "query_failed",
          "scratch_cleanup_incomplete",
          "shard_query_incomplete"
        ]
      }
    },
    "reading": "A taint batch of 2,526 questions over a 673-file PHP graph does not complete in thirty minutes. Raising the ceiling six-fold, from 300 to 1,800 seconds, changed nothing except how long the failure took: 363 seconds of query time became 1,874, and zero questions completed in both. The eight-way split retry did not rescue it either, finishing its parts in about three minutes against a deadline with 3,300 seconds still on it, which is worth a look on its own.",
    "engine_comparison": "4.0.625 and 4.0.631 are indistinguishable on this subject: 627.7s against 619.1s, 363.4s against 367.3s of query time, zero completed in both. The six releases between them change nothing here."
  },
  "on_the_engine_pin": {
    "decision": "leave the image pinned at 4.0.625",
    "why": "4.0.631 was built and measured and is equivalent on this subject, so there is no demonstrated benefit to weigh against re-verifying every JavaScript and Python measurement on a new engine.",
    "what_a_bump_would_cost": "Three values in ops/frontend-retention/install.py, not a port. The upstream AstGenRunner.scala is byte-identical in 4.0.631, sha256 9963f949360d7e17baed1bd8875b717b2855745cf7260eef8a87d17533a54268, and its patch anchor still appears exactly once. Only the jar path, the jar checksum (f770d929d192a7da40932ee8c22df76ca635af216f643acf2f46a0fbe5747ef1 for io.joern.jssrc2cpg-4.0.631.jar) and the source URL tag need changing."
  },
  "limits": [
    "One PHP subject. The other three plugins available hold no deserialization site and were not run.",
    "The split retry's fast failure is unexplained and may be hiding a second cause.",
    "Both engines were measured with the same region budget, 500, and the same driver."
  ]
}