test(m3): PHP still not established; blocker is the dataflow execution pass

Retried the PMPro transfer with batch splitting, a 4500-second deadline and a 900-second query ceiling. Stopped after 26 minutes with no case completed and no verdict, so PHP transfer remains unestablished. A probe narrowed the cause and corrected my earlier reading of it. The evidence pass answers all 4,425 PMPro PHP regions in ONE batch in 29.7 seconds with a 681 KB response, exit 0, and its cost is fixed rather than per request: one request took 40.2 s and 4,425 took 29.7 s. Neither request count nor response size is anywhere near a limit, so batch size and serialization are not the cause as I first supposed from a single extrapolated data point. What fails is the execution pass, which runs reachableByFlows per candidate. The probe's own scope is recorded with it: it sent evidenceOnly=true and therefore measured the structural pass, not the failing one. Those numbers would otherwise read as evidence that PMPro's taint work is cheap, which they are not. Splitting stays, because it is correct on its own terms and recovers a batch that fails whole. It is worth noting it also makes an unrecoverable batch slower to abandon, since each of the eight parts burns its own share of the query ceiling. M3 remains two of three: Node established in M1b, Python established, PHP blocked on dataflow cost for a 1,274-file repository. Co-Authored-By: Claude Opus 5 (1M context) <[email protected]> Claude-Session: https://claude.ai/code/session_01EmM1j46qgKMh3zqD1QMwEo
2026-10-17 20:19:29 +00:00 · 2026-09-18 13:22:05 +02:00
parent aa66429a82
commit 537483657d
4 changed files with 102 additions and 7 deletions
@@ -140,7 +140,7 @@
    "sha256": "6753e7e53695016da04f0092ad54fea5c5fc7470d4de0c5e284721d68ca22337",
    "files": 10
  },
-  "outcome": "PHP TRANSFER NOT ESTABLISHED",
+  "outcome": "PHP TRANSFER NOT ESTABLISHED; blocker narrowed to the dataflow execution pass",
  "what_failed": "All three comparisons exhausted a 3000-second deadline and produced no artifact, so there is no target evidence at all. The engine reports `cpg batch taint failed` twice per case on the 1,274-file PMPro tree: first with an exception whose trace runs through ujson array transformation, then by timeout on the retry.",
  "budgets_tried": [
    {
@@ -166,5 +166,22 @@
  "reading": "This is a runtime blocker on a large repository, not evidence about whether the design transfers to PHP. The pins, source receipts and declared target are all correct, so the question was asked properly and the engine could not answer it within any budget tried. PHP transfer stays unestablished and M3 stays unmet.",
  "limits": [
    "No detection, precision or admission claim for PHP: nothing completed to claim anything about."
-  ]
+  ],
+  "retry_with_batch_splitting": {
+    "date": "2026-09-18",
+    "change": "A batch that fails whole is retried in eight parts; each request is independent.",
+    "budgets": {
+      "deadline_seconds": 4500,
+      "query_timeout_seconds": 900
+    },
+    "result": "stopped by the maintainer after 26 minutes with no case completed and no verdict produced",
+    "reading": "Splitting did not rescue the execution pass within a time anyone would wait. It also makes an unrecoverable batch slower to give up on, because each of the eight parts burns its own share of the query ceiling before the next is tried, so one timeout becomes several."
+  },
+  "narrowed_by_probe": {
+    "evidence_pass": "All 4,425 PMPro PHP regions answer in ONE batch in 29.7 s with a 681 KB response, exit 0.",
+    "cost_shape": "Fixed, not per request: 1 request took 40.2 s and 4,425 took 29.7 s.",
+    "conclusion": "Neither request count nor response size is near a limit. The blocker is the execution pass, which runs reachableByFlows per candidate, and not batch size or serialization as first supposed.",
+    "instrument_note": "That probe sent evidenceOnly=true, so it measured the structural pass and not the failing one. Recorded because those numbers would otherwise read as evidence that PMPro's taint work is cheap.",
+    "evidence": "benchmarks/measurements/2026-09-18-php-batch-diagnosis.json"
+  }
 }
@@ -0,0 +1,72 @@
+{
+  "experiment": "php-taint-batch-v1",
+  "root": "/cache/pmpro/1a900be3f09f",
+  "attempts": [
+    {
+      "requests": 1,
+      "seconds": 40.16467795101926,
+      "exit": 0,
+      "stdout_bytes": 36400,
+      "answered": true,
+      "answer_keys": 2,
+      "stderr_head": "Picked up JAVA_TOOL_OPTIONS: -Xmx2048m\nPicked up JAVA_TOOL_OPTIONS: -Xmx2048m\nexecuting /app/src/openultrasast/cpg/queries/taint.sc with params=Map(cpgFile -> /results/php-probe/cpg.bin, requestsFile -> /results/php-probe/requests-1.json)\nCreating project `cpg.bin` for CPG at `/results/php-probe/cpg.bin`\nCreating working copy of CPG to be safe\nLoading base CPG from: /results/php-probe/workspace/cpg.bin/cpg.bin.tmp\nOverlay dataflowOss already exists - skipping\nThe graph has been modified. You may want to use the `save` command to persist changes to disk.  All changes will also be saved collectively on exit\nclosing/saving project `cpg.bin`\n",
+      "stderr_bytes": 646
+    },
+    {
+      "requests": 10,
+      "seconds": 39.16813864791766,
+      "exit": 0,
+      "stdout_bytes": 37689,
+      "answered": true,
+      "answer_keys": 11,
+      "stderr_head": "Picked up JAVA_TOOL_OPTIONS: -Xmx2048m\nPicked up JAVA_TOOL_OPTIONS: -Xmx2048m\nexecuting /app/src/openultrasast/cpg/queries/taint.sc with params=Map(cpgFile -> /results/php-probe/cpg.bin, requestsFile -> /results/php-probe/requests-10.json)\nCreating project `cpg.bin` for CPG at `/results/php-probe/cpg.bin`\nProject with name cpg.bin already exists - overwriting\nCreating working copy of CPG to be safe\nLoading base CPG from: /results/php-probe/workspace/cpg.bin/cpg.bin.tmp\nOverlay dataflowOss already exists - skipping\nThe graph has been modified. You may want to use the `save` command to persist changes to disk.  All changes will also be saved collectively on exit\nclosing/saving project `cpg.bin`\n",
+      "stderr_bytes": 702
+    },
+    {
+      "requests": 50,
+      "seconds": 30.821678715990856,
+      "exit": 0,
+      "stdout_bytes": 43445,
+      "answered": true,
+      "answer_keys": 51,
+      "stderr_head": "Picked up JAVA_TOOL_OPTIONS: -Xmx2048m\nPicked up JAVA_TOOL_OPTIONS: -Xmx2048m\nexecuting /app/src/openultrasast/cpg/queries/taint.sc with params=Map(cpgFile -> /results/php-probe/cpg.bin, requestsFile -> /results/php-probe/requests-50.json)\nCreating project `cpg.bin` for CPG at `/results/php-probe/cpg.bin`\nProject with name cpg.bin already exists - overwriting\nCreating working copy of CPG to be safe\nLoading base CPG from: /results/php-probe/workspace/cpg.bin/cpg.bin.tmp\nOverlay dataflowOss already exists - skipping\nThe graph has been modified. You may want to use the `save` command to persist changes to disk.  All changes will also be saved collectively on exit\nclosing/saving project `cpg.bin`\n",
+      "stderr_bytes": 702
+    },
+    {
+      "requests": 200,
+      "seconds": 31.992039224831387,
+      "exit": 0,
+      "stdout_bytes": 65147,
+      "answered": true,
+      "answer_keys": 201,
+      "stderr_head": "Picked up JAVA_TOOL_OPTIONS: -Xmx2048m\nPicked up JAVA_TOOL_OPTIONS: -Xmx2048m\nexecuting /app/src/openultrasast/cpg/queries/taint.sc with params=Map(cpgFile -> /results/php-probe/cpg.bin, requestsFile -> /results/php-probe/requests-200.json)\nCreating project `cpg.bin` for CPG at `/results/php-probe/cpg.bin`\nProject with name cpg.bin already exists - overwriting\nCreating working copy of CPG to be safe\nLoading base CPG from: /results/php-probe/workspace/cpg.bin/cpg.bin.tmp\nOverlay dataflowOss already exists - skipping\nThe graph has been modified. You may want to use the `save` command to persist changes to disk.  All changes will also be saved collectively on exit\nclosing/saving project `cpg.bin`\n",
+      "stderr_bytes": 703
+    },
+    {
+      "requests": 4425,
+      "seconds": 29.67125784815289,
+      "exit": 0,
+      "stdout_bytes": 681199,
+      "answered": true,
+      "answer_keys": 4426,
+      "stderr_head": "Picked up JAVA_TOOL_OPTIONS: -Xmx2048m\nPicked up JAVA_TOOL_OPTIONS: -Xmx2048m\nexecuting /app/src/openultrasast/cpg/queries/taint.sc with params=Map(cpgFile -> /results/php-probe/cpg.bin, requestsFile -> /results/php-probe/requests-4425.json)\nCreating project `cpg.bin` for CPG at `/results/php-probe/cpg.bin`\nProject with name cpg.bin already exists - overwriting\nCreating working copy of CPG to be safe\nLoading base CPG from: /results/php-probe/workspace/cpg.bin/cpg.bin.tmp\nOverlay dataflowOss already exists - skipping\nThe graph has been modified. You may want to use the `save` command to persist changes to disk.  All changes will also be saved collectively on exit\nclosing/saving project `cpg.bin`\n",
+      "stderr_bytes": 704
+    }
+  ],
+  "build": {
+    "seconds": 51.444085616152734,
+    "exit": 0,
+    "bytes": 12062695
+  },
+  "regions": {
+    "total": 4461,
+    "php": 4425
+  },
+  "milestone": "M3 PHP diagnosis",
+  "instrument_scope": "This probe sent evidenceOnly=true, so it measured the EVIDENCE pass, which is structural and does no dataflow. That is not the pass that fails in production. Stating it plainly because the numbers would otherwise read as evidence that PMPro's taint work is cheap, which they are not.",
+  "what_it_establishes": "The evidence pass scales on PMPro without difficulty: all 4,425 PHP regions answered in one batch in 29.7 seconds with a 681 KB response, exit 0. Cost is fixed overhead, not per request: one request took 40.2 s and 4,425 took 29.7 s. So neither request count nor response size is near any limit in that mode.",
+  "what_it_narrows": "The production failure is therefore specific to the EXECUTION pass, which runs reachableByFlows per candidate. Batch size and serialization are not the cause on their own; the dataflow work is.",
+  "limits": [
+    "Evidence mode only, one PHP repository, one machine. No detection or latency claim."
+  ]
+}