feat(pre-push-safety-net): exclude vendor inputs and partition frontend coverage

.kiro/specs/pre-push-safety-net/brief.md

@@ -1,6 +1,6 @@
 # Brief: pre-push-safety-net
 
-Status: requirements, design and tasks approved on 2026-09-12; groups 1 and 2 plus tasks 3.1–3.3 verified complete (10/27 subtasks); task 3.4 blocked on frontend test retention; groups 4–8 pending. Language: English.
+Status: requirements, design and tasks approved on 2026-09-12; groups 1–3 verified complete (11/27 subtasks), including the approved contributor-scan frontend-retention prerequisite; groups 4–8 pending. Language: English.
 
 ## Problem
 

.kiro/specs/pre-push-safety-net/implementation-group-3.md

@@ -15,10 +15,10 @@ coverage limits.
 
 ## Verification state
 
-Tasks 3.1–3.3 are independently approved and verified (10/27 tasks overall).
+Group 3 is independently approved and VERIFIED (TASK scope): tasks 3.1–3.4 are
-Task 3.4 is blocked on a confirmed frontend retention prerequisite; group 3 is
+complete, bringing the pre-push spec to 11/27 tasks. The approved contributor-scan
-not complete. The proposed foundation repair is in
+frontend-retention prerequisite 2.18 is also verified. Groups 4–8 remain pending;
-`../contributor-scan/frontend-retention-prerequisite.md`.
+this is not feature GO, capability admission or a hook latency claim.
 
 ## Baseline
 
@@ -130,7 +130,7 @@ line enforced security, that the defect is new, or that a hook meets its latency
 
 Ten of 27 tasks are now complete; task 3.4 remains for this group.
 
-## Task 3.4 blocked verification
+## Historical task 3.4 blocker
 
 Claim type TASK; status NOT_VERIFIED. The implementer returned BLOCKED, and an
 independent fresh-context Kiro debug investigation returned STOP_FOR_HUMAN with
@@ -169,3 +169,44 @@ uncommitted pending the foundation repair, review and fresh runtime verification
 The packaged failure preceded a comment-only clarification in `cpg/backend.py`
 removing the disproven attribution to legacy frontend autodetection. No executable
 backend behavior changed after that measurement.
+
+## Task 3.4 final verification
+
+Status VERIFIED, claim type TASK; independent review APPROVED after two concrete
+error-path repairs. The final suite passed 1,151 tests with nine skipped in
+100.55 seconds. Ruff, formatting (230 files), mypy (109 source files), compileall,
+diff and all three regression gates passed.
+
+The approved foundation repair is committed as `51fabef`. Physically filtered
+frontend inputs retain original source paths and bytes; the existing layout facts
+mark JavaScript/TypeScript test conventions unshipped. One existing ranker decision
+and region/deadline budget cover supported partitions. Browser-path source remains
+available, while runtime origin stays unspecified without declarations. Vendor,
+cross-language, TypeScript-property and C-bounds limits remain explicit.
+
+Exact source names now travel in taint, dominance and configuration batch censuses.
+PHP aggregation preserves them only when every shard provides named evidence.
+Synthetic `<unknown>` nodes cannot mask missing real files. Unreadable unknown
+extensions produce `source_unreadable` coverage instead of raising PermissionError.
+Legacy callbacks that omit census entirely retain their existing compatibility
+behavior; production queries emit the named census, and unsupported/partial
+sharded evidence cannot establish a complete negative.
+
+`benchmarks/measurements/2026-09-13-partition-scope-smoke.json` records the final
+image and ten matching source/query/fact hashes. The real mixed graph retains
+seven JavaScript files (296 bytes), including tests and browser source, plus two
+PHP files (97 bytes), with vendor nodes absent. One question is selected and three
+deferred; unsupported Go is reported. No false empty/unavailable census remains.
+PHP and JavaScript change-context controls both complete with unchanged-helper
+witnesses, preserving dirty checkout/index state and cleaning snapshot scratch.
+
+A real Python graph supplies named census from all three production query kinds.
+Those unchanged Scala sources were verified before the separate PHP transport and
+discovery repairs; the artifact records that narrower provenance explicitly. A
+real filesystem permission probe verifies the repaired error path, using a clearly
+labeled synthetic graph callback.
+
+The mixed smoke took 197.76 seconds under a 900-second lab allowance; PHP/JavaScript
+context checks took 117.3/95.9 seconds. These manually declared controls do not
+establish automatic scope discovery, production hook latency or alert quality.
+Group 4 now owns novelty comparison, actionability admission and compact output.

.kiro/specs/pre-push-safety-net/research.md

@@ -6,7 +6,7 @@
 - User direction: a scalable pre-push safety net as AI accelerates development; WordPress is one case.
 - Clarifications: advisory must be actionable, not SAST noise; the ranker is the scope mechanism;
   third-party vendor code must remain outside targets and graphs.
-- Requirements, design and tasks are now approved, and groups 1 and 2 plus tasks 3.1–3.3 are verified complete; task 3.4 is blocked on JavaScript frontend test retention; see
+- Requirements, design and tasks are now approved, and groups 1–3 are verified complete after the approved frontend-retention repair; see
   `tasks.md` for verified completion. The discovery log below preserves the earlier drafting state
   and measurements. No hook latency claim is made.
 
@@ -167,3 +167,14 @@ Direct frontend invocation reproduced the omission. Two unconditional upstream
 test filters require an owning foundation repair; see
 `../contributor-scan/frontend-retention-prerequisite.md`. The runtime failure is
 not overridden by passing mocks, static checks or a populated server graph.
+
+### Group 3 completed after frontend and coverage repairs
+
+The approved contributor-scan 2.18 adaptation preserves test directories and
+filename suffixes through both pinned JavaScript frontend filters. Task 3.4 then
+passed independent review with exact graph-path census, including PHP aggregation.
+Review also repaired unreadable discovery inputs throwing rather than reporting
+incomplete coverage. A synthetic graph file must never compensate for an omitted
+real source, and a wrapper must not discard evidence supplied by a frontend.
+See `implementation-group-3.md` for final evidence. Eleven of 27 tasks are complete;
+change attribution, actionability admission and compact output are next.

.kiro/specs/pre-push-safety-net/spec.json

@@ -1,7 +1,7 @@
 {
   "feature_name": "pre-push-safety-net",
   "created_at": "2026-09-12T15:35:00Z",
-  "updated_at": "2026-09-13T09:17:01.788332+00:00",
+  "updated_at": "2026-09-13T09:50:01.821464+00:00",
   "language": "en",
   "phase": "implementation",
   "approvals": {
@@ -19,7 +19,7 @@
     }
   },
   "ready_for_implementation": true,
-  "implementation_approval_note": "Groups 1 and 2 verified complete. Maintainer authorized continuing group 3; remaining approved tasks execute in dependency order with independent review and fresh verification.",
+  "implementation_approval_note": "Groups 1\u20133 verified complete. Contributor-scan frontend-retention prerequisite 2.18 approved and verified. Remaining approved tasks require independent review and fresh verification.",
   "depends_on": [
     "contributor-scan"
   ],
@@ -77,23 +77,17 @@
       "2.3",
       "3.1",
       "3.2",
-      "3.3"
+      "3.3",
+      "3.4"
     ],
-    "remaining_tasks": 17,
+    "remaining_tasks": 16,
     "verification_record": "implementation-group-3.md",
     "verification_records": [
       "implementation-group-1.md",
       "implementation-group-2.md",
       "implementation-group-3.md"
     ],
-    "next_group": 3,
+    "next_group": 4,
-    "blocked_tasks": [
+    "blocked_tasks": []
-      {
-        "task": "3.4",
-        "reason": "Pinned JavaScript frontend silently excludes first-party tests; missing contributor-scan prerequisite.",
-        "record": "../contributor-scan/frontend-retention-prerequisite.md",
-        "next_action": "implement_approved_contributor_scan_2.18"
-      }
-    ]
   }
 }

.kiro/specs/pre-push-safety-net/tasks.md

@@ -1,7 +1,7 @@
 # Implementation Plan: pre-push-safety-net
 
-Status: task plan approved by the maintainer's "approve." on 2026-09-12. Groups 1 and 2 (tasks 1.1–2.3)
+Status: task plan approved by the maintainer's "approve." on 2026-09-12. Groups 1–3 (tasks 1.1–3.4)
-are complete after independent review and fresh verification. Group 3 has 3.1–3.3 verified; 3.4 is blocked on frontend test retention. Groups 4–8 remain pending;
+are complete after independent review and fresh verification (11/27 subtasks). Groups 4–8 remain pending;
 the hook and capability admission are not yet implemented. See `implementation-group-1.md` and `implementation-group-2.md`.
 
 ## Execution contract
@@ -78,7 +78,7 @@ the hook and capability admission are not yet implemented. See `implementation-g
   - _Requirements: 1.4, 2.1, 2.4, 3.4_
   - _Verified 2026-09-12:_ independent review APPROVED; fresh final suite 1,069 passed, 9 skipped; focused 148 passed. Global Ruff/format, touched mypy and regression gates passed. PHP/JavaScript fixtures retain removed spans and unchanged-operation anchors across declaration renames; uncertain moves, repeated lines, binary and unavailable inputs remain unresolved. Forty independent real-Git comparisons checked exact spans/anchors. Network-disabled image smoke preserved committed PHP/JavaScript bytes and dirty non-HEAD state, with cleanup verified; see `implementation-group-2.md`. Lexical anchors do not establish semantic impact or novelty.
 
-- [ ] 3. Integrate bounded execution and ranker scope with the existing engine
+- [x] 3. Integrate bounded execution and ranker scope with the existing engine
 - [x] 3.1 (P) Enforce one remaining-time budget in the backend
   - Thread the absolute deadline through readability probes, frontend builds, overlays, census, query batches, fallbacks and retries; no phase gets a fresh full budget.
   - Bound process-group termination and lease/scratch cleanup by the cancellation allowance, replacing any longer fixed cleanup waits on the push path.
@@ -103,13 +103,14 @@ the hook and capability admission are not yet implemented. See `implementation-g
   - _Depends: 2.3, 3.2_
   - _Requirements: 2.1, 2.2, 2.4, 8.1, 8.2_
   - _Verified 2026-09-13:_ independent review APPROVED after an unknown-call boundary repair; fresh suite 1,128 passed, nine skipped; global Ruff/format/mypy passed. Real PHP/JavaScript Git guard-removal snapshots reached unchanged helper-file witnesses with local dirty state preserved. Production negative regression reproduced then fixed unknown(req.body) being pruned as tier zero: modeled arguments no longer summarize unknown consumers. Existing ranking and detector semantics unchanged; unsupported projections remain explicit. Evidence: `benchmarks/measurements/2026-09-13-change-context-smoke.json`. Lexical attribution is not semantic novelty or alert admission.
-- [ ] 3.4 Apply vendor exclusion and explicit language partitions
+- [x] 3.4 Apply vendor exclusion and explicit language partitions
-  - _Blocked:_ Joern 4.0.625/ASTGen 3.50.1 omit first-party JavaScript tests despite readable filtered inputs. Independent debug returned STOP_FOR_HUMAN; review the owning `contributor-scan/frontend-retention-prerequisite.md` before resuming. Implementation remains uncommitted and unapproved; runtime evidence: `benchmarks/measurements/2026-09-13-partition-retention-blocker.json`.
+  - _Prerequisite resolved 2026-09-13:_ contributor-scan 2.18 independently approved and verified at `51fabef`; both JavaScript test filters repaired by explicit versioned input policy. Task 3.4 subsequently passed its own independent review.
   - Reuse declared layout exclusions before target enumeration and graph build; retain unshipped first-party tests and separate browser/server or language capability coverage.
   - Build supported frontend partitions explicitly; preserve unknown cross-language and vendor semantics without restoring excluded sources.
   - Done when PHP and Node graphs/targets contain no declared vendor code, first-party callers remain visible, and unsupported partitions appear as coverage limits rather than clean results.
   - _Boundary: Ranker Scope Contract, Cache and Engine Boundary integration_
   - _Requirements: 2.2, 2.3, 7.3, 8.2, 8.4_
+  - _Verified 2026-09-13:_ independent review APPROVED; 1,151 passed, nine skipped; Ruff/format/mypy pass. Real PHP/JavaScript graphs retain tests/browser source, omit vendor nodes and share one ranker limit. Named census survives PHP aggregation; missing source names and unreadable discovery inputs remain explicit. Python exercises all three batch census schemas. Evidence: `benchmarks/measurements/2026-09-13-partition-scope-smoke.json`. No hook capability or latency claim.
 
 - [ ] 4. Implement change attribution and actionable output
 - [ ] 4.1 Compare targeted base/head evidence for defect novelty

.kiro/steering/roadmap.md

@@ -27,7 +27,7 @@ engine are out of scope.
 - Equal-score tie-breaking alone cannot bring PMPro or WP Statistics under 50. The existing
   ranker is useful, but its measured ordering is not enabled by the ordinary CLI caller.
 - Docker/Compose/ops now default to 4.0.625; task 1.1 passed local PHP/JavaScript image smoke and independent review. No host deployment or hook latency claim follows from that runtime evidence.
-- `contributor-scan` has 24/40 checked tasks; historical checkboxes are not rewritten by this review.
+- `contributor-scan` has 25/41 checked tasks, including the approved and verified frontend-retention prerequisite 2.18.
   `flow-aware-ranking` is still an unapproved brief despite implementation history.
 
 ## Existing Spec Updates
@@ -46,15 +46,16 @@ engine are out of scope.
 - [ ] pre-push-safety-net — Immutable pushed tips, existing-ranker integration, compatible artifact
   reuse, total deadline, change attribution, actionability admission and compact hook behavior.
   Dependencies: contributor-scan. Requirements/design approved 2026-09-12; tasks generated
-  (8 groups, 27 executable subtasks), independently reviewed and approved; groups 1 and 2 and tasks 3.1–3.3 are complete (10/27 tasks); 3.4 is blocked.
+  (8 groups, 27 executable subtasks), independently reviewed and approved; groups 1–3 are complete (11/27 tasks), including the resolved frontend-retention prerequisite.
 
 ### Next concrete step
 
-Groups 1 and 2 and tasks 3.1–3.3 have passed independent review: runtime, inputs,
+Groups 1–3 have passed independent review: runtime, inputs, scoring, snapshots,
-scoring, snapshots, shared deadline, exact ranker scope and change-context evidence.
+shared deadline, exact ranker scope, change-context evidence and vendor-free language
-Task 3.4 is blocked: real Joern graphs omit first-party JavaScript tests even though
+partitions. The approved contributor-scan prerequisite 2.18 (`51fabef`) repairs both
-filtered inputs contain them. Review `contributor-scan/frontend-retention-prerequisite.md`,
+JavaScript test filters; real graphs retain first-party tests and exclude vendor code.
-repair that foundation contract, then rerun the partition census and independent review.
+Task 3.4 also repairs unreadable-input handling and preserves named PHP census evidence.
+Group 4 is next: novelty comparison, actionability admission and compact advisory output.
 See `pre-push-safety-net/implementation-group-3.md` for evidence and limits.
 The first complete experiment is 5.2, before persistent caching and hook
 installation work. Keep PMPro as the primary core-optimization example, then freeze the core and

.kiro/steering/safety-net.md

@@ -79,8 +79,10 @@ updates as a shortcut to an unmeasured latency promise.
 The product direction above is user-requested. The amended requirements, design and initial
 evaluation profile in `pre-push-safety-net` were approved for task generation on 2026-09-12
 ("ok, next kiro phase"). The maintainer then approved the task plan and starting group 1
-("approve.") and then group 2 ("ok, next kiro phase, approved"). Groups 1 and 2 are now verified
+("approve.") and then group 2 ("ok, next kiro phase, approved"). Groups 1–3 are now verified
-complete, as are group 3 tasks 3.1–3.3 (10/27 total). Task 3.4 is blocked on a
+complete (11/27 total). The maintainer approved the frontend-retention prerequisite;
-confirmed frontend test-retention gap; review the proposed contributor-scan
+contributor-scan 2.18 and pre-push task 3.4 have both passed independent review and
-frontend-retention prerequisite before resuming it. Implementation is authorized under that plan; task completion still requires
+fresh verification. Group 4 is next: novelty comparison, actionability admission and
-independent review and fresh evidence. Hook performance and capability admission remain unproven.
+compact advisory output. Implementation is authorized under the approved plan;
+task completion still requires independent review and fresh evidence. Hook performance
+and capability admission remain unproven.

benchmarks/measurements/2026-09-08-module-audit.json

@@ -1,10 +1,10 @@
 {
   "counts": {
-    "load_bearing": 98,
+    "load_bearing": 99,
     "standalone_capability": 9
   },
-  "modules": 107,
+  "modules": 108,
-  "note": "Refreshed 2026-09-12 after pre-push-safety-net group 2: immutable comparison, materialization and lexical change context are spec-owned; ranker/runner integration remains pending.",
+  "note": "Refreshed after pre-push-safety-net task 3.4: first-party frontend partitions share the authoritative ranker and execution budget.",
   "orphaned": [],
   "rows": [
     {
@@ -129,6 +129,7 @@
         "model.config_value",
         "model.dominance",
         "model.judge",
+        "model.partitions",
         "model.pipeline",
         "model.scan",
         "model.taint"
@@ -149,6 +150,7 @@
         "model.config_value",
         "model.dominance",
         "model.judge",
+        "model.partitions",
         "model.pipeline",
         "model.scan",
         "model.taint"
@@ -338,12 +340,14 @@
       "classification": "load_bearing",
       "importers": [
         "cli",
+        "cpg.backend",
         "model.candidates",
         "model.config_value",
         "model.dominance",
         "model.execution",
         "model.judge",
         "model.layout",
+        "model.partitions",
         "model.pipeline",
         "model.regions",
         "model.report",
@@ -390,6 +394,10 @@
     {
       "classification": "load_bearing",
       "importers": [
+        "cpg.backend",
+        "model.partitions",
+        "model.regions",
+        "model.scan",
         "push.contracts",
         "push.snapshot"
       ],
@@ -450,12 +458,21 @@
     {
       "classification": "load_bearing",
       "importers": [
+        "model.partitions",
         "model.scan",
         "preprocess"
       ],
       "module": "model.layout",
       "reason": "reachable from gate, map_gate, pair_gate or the CLI"
     },
+    {
+      "classification": "load_bearing",
+      "importers": [
+        "model.scan"
+      ],
+      "module": "model.partitions",
+      "reason": "reachable from gate, map_gate, pair_gate or the CLI"
+    },
     {
       "classification": "load_bearing",
       "importers": [
@@ -596,6 +613,7 @@
         "map_gate",
         "mapping",
         "model.candidates",
+        "model.partitions",
         "model.scan",
         "pairs",
         "rank",

ops/README.md

@@ -181,8 +181,19 @@ unsupported Go partition is reported explicitly. One ranker decision and one reg
 limit cover both supported frontends. The shared 900-second budget is a lab allowance;
 this check establishes neither production hook latency nor framework admission.
 
-Current status (2026-09-13): this is a failing regression on Joern 4.0.625. The
+Stock Joern 4.0.625 fails the retention check: the projected test file is readable
-projected JavaScript test file is readable but absent from the resulting graph.
+but absent from the graph. The approved source adaptation fixes both test filters;
-See `benchmarks/measurements/2026-09-13-partition-retention-blocker.json` and the
+see `frontend-retention/README.md` for build pins and inclusion policy, and
-proposed contributor-scan frontend-retention prerequisite. An exit 1 is expected
+`benchmarks/measurements/2026-09-13-frontend-retention-smoke.json` for passing
-until that engine contract is repaired; do not treat it as a passing partition smoke.
+foundation verification. The earlier failing evidence remains in
+`benchmarks/measurements/2026-09-13-partition-retention-blocker.json`.
+
+Setting `OUSAST_SMOKE_INCLUDE_TESTS=0` in the container explicitly disables the
+backend inclusion policy; that negative control must fail the retention assertion.
+The expanded task 3.4 smoke also verifies first-party browser-path retention and
+unshipped test-path records. Runtime origin remains unspecified without declarations.
+
+`ops/smoke_partition_census.py`, using the same container invocation, builds a small
+readable Python graph and exercises the production taint, dominance and configuration
+batch queries. Each must return the actual source filename in its census. This
+checks query compilation and census transport, not security-property coverage.

ops/smoke_context.py

@@ -95,6 +95,10 @@ def main() -> None:
                 enriched = result.change_context
                 assert enriched is not None and enriched.relationships, "changed context did not reach any question"
                 assert result.scope is not None and result.scope.selected
+                invalid_census = {"cpg_empty", "files_unparsed", "partition_file_census_unavailable", "partition_file_census_incomplete"}
+                assert not invalid_census.intersection(result.scope.unresolved_boundaries), result.scope
+                assert not any(d.get("reason") in invalid_census for d in result.degradations), result.degradations
+                assert result.question_outcomes and all(outcome.status == "completed" for outcome in result.question_outcomes), result
                 selected = {question.identity for question in result.scope.selected}
                 assert any(relation.target in selected for relation in enriched.relationships)
                 assert result.findings, "controlled unchanged sink has no witness"

ops/smoke_partition_census.py

@@ -0,0 +1,54 @@
+"""Compile all production batch censuses on a readable real Python graph."""
+
+from __future__ import annotations
+
+import hashlib
+import json
+import tempfile
+import time
+from pathlib import Path
+
+from openultrasast.cpg.backend import JoernBackend
+from openultrasast.model.contracts import ExecutionBudget
+
+
+def main() -> None:
+    budget = ExecutionBudget(time.monotonic() + 600, 2.0)
+    results = {}
+    with tempfile.TemporaryDirectory(prefix="ousast-census-smoke-") as temporary:
+        root = Path(temporary)
+        source = root / "probe.py"
+        source.write_text("def handler(value):\n    return value\n")
+        data = source.read_bytes()
+        assert data, "source input unreadable"
+        backend = JoernBackend(heap_mb=1024, execution_budget=budget)
+        graph = backend.build(root, language="python")
+        assert graph is not None and graph.run_batch is not None, "real Python graph unavailable"
+        try:
+            for kind in ("taint", "dominance", "config"):
+                answer = graph.run_batch(kind, {"probe": {"file": "probe.py", "function": "handler"}})
+                assert isinstance(answer, dict) and "probe" in answer, f"{kind} batch unanswered"
+                census = answer.get("__census__")
+                assert isinstance(census, list) and census, f"{kind} census missing"
+                names = census[0].get("file_names")
+                assert isinstance(names, list) and any(str(p).endswith("probe.py") for p in names), census
+                assert int(census[0]["methods"]) > 0, census
+                results[kind] = census
+        finally:
+            graph.cleanup()
+    print(
+        json.dumps(
+            {
+                "purpose": "Real Python batch schema and query compilation; no security property admission claim",
+                "source_bytes": len(data),
+                "source_sha256": hashlib.sha256(data).hexdigest(),
+                "shared_budget_seconds": 600,
+                "censuses": results,
+            },
+            indent=2,
+        )
+    )
+
+
+if __name__ == "__main__":
+    main()

ops/smoke_partitions.py

@@ -39,6 +39,7 @@ def main() -> None:
         sources = {
             "src/api.php": "<?php\nfunction php_handler($req) { eval($req); }\n",
             "server/api.js": "function js_handler(req) { eval(req.query.code); }\n",
+            "browser/client.js": "function browser_marker() { return 42; }\n",
             "tests/probe.php": "<?php\nfunction php_test_marker() { return 42; }\n",
             "tests/probe.js": "function js_test_marker() { return 42; }\n",
             "vendor/library.php": "<?php\nfunction vendor_php_marker($req) { eval($req); }\n",
@@ -91,6 +92,8 @@ def main() -> None:
                         assert any(str(filename).endswith(path) for filename in files), witness
                         assert method in methods, witness
                     assert "php_handler" not in methods, witness
+                    assert any(str(path).endswith("browser/client.js") for path in files), witness
+                    assert "browser_marker" in methods, witness
                 else:
                     raise AssertionError(f"unexpected frontend: {language}")
                 records.append({"frontend": language, "input_root": str(root), "graph_witness": witness})
@@ -116,12 +119,20 @@ def main() -> None:
         )
         assert {record["frontend"] for record in records} == {"php", "javascript"}, records
         assert result.scope is not None
+        invalid_census = {"cpg_empty", "files_unparsed", "partition_file_census_unavailable", "partition_file_census_incomplete"}
+        assert not invalid_census.intersection(result.scope.unresolved_boundaries), result.scope
+        assert not any(d.get("reason") in invalid_census for d in result.degradations), result.degradations
         assert [q.identity.function for q in result.scope.selected] == ["php_handler"], result.scope
         assert {q.identity.function for q in result.scope.deferred} == {"js_handler", "php_test_marker", "js_test_marker"}, result.scope
         assert len(result.question_outcomes) == 1, result.question_outcomes
         assert result.question_outcomes[0].identity == result.scope.selected[0].identity
         assert all("vendor" not in q.identity.path and "node_modules" not in q.identity.path for q in result.scope.selected), result.scope
         assert any(p.language == "go" and p.status == "unsupported" for p in result.partitions), result.partitions
+        javascript = next(p for p in result.partitions if p.language == "javascript")
+        assert set(javascript.unshipped_paths) == {"tests/probe.js", *suffix_tests}, javascript
+        assert javascript.runtime == "unspecified", "runtime role needs declared evidence"
+        php = next(p for p in result.partitions if p.language == "php")
+        assert php.unshipped_paths == ("tests/probe.php",), php
         assert all(not Path(record["input_root"]).exists() for record in records), "partition scratch leaked"
         report = {
             "purpose": "Real mixed PHP/JavaScript graph census and one global ranker budget; no admission or latency claim",

src/openultrasast/cli.py

@@ -4,7 +4,7 @@ import argparse
 import json
 import tempfile
 from collections.abc import Mapping, Sequence
-from dataclasses import dataclass, replace
+from dataclasses import asdict, dataclass, replace
 from pathlib import Path
 from time import perf_counter
 from typing import TYPE_CHECKING, Any
@@ -903,6 +903,7 @@ def _model_payload(
         "scope": scope.to_payload() if scope is not None else None,
         "question_outcomes": [q.to_payload() for q in outcomes],
         "family_coverage": [c.to_payload() for c in getattr(result, "family_coverage", ())],
+        "partitions": [asdict(p) for p in getattr(result, "partitions", ())],
         "by_rung": dict(getattr(result, "by_rung", {}) or {}),
         "regions_scanned": getattr(result, "regions_scanned", 0),
         "regions_unjudged": getattr(result, "regions_unjudged", 0),

src/openultrasast/cpg/backend.py

@@ -697,9 +697,12 @@ class JoernBackend:
         methods = 0
         files = 0
         answered = 0
+        file_names: set[str] = set()
+        named_census = True
         for shard in shards:
             one = self.query_batch(shard, query, requests)
             if one is None:
+                named_census = False
                 if self.execution_budget is not None:
                     self._note("shard_query_incomplete")
                     return None
@@ -709,11 +712,21 @@ class JoernBackend:
             if census and isinstance(census[0], Mapping):
                 methods += int(str(census[0].get("methods", "0")) or 0)
                 files += int(str(census[0].get("files", "0")) or 0)
+                names = census[0].get("file_names")
+                if isinstance(names, list) and all(isinstance(name, str) for name in names):
+                    file_names.update(names)
+                else:
+                    named_census = False
+            else:
+                named_census = False
             for rid, rows in one.items():
                 merged.setdefault(rid, []).extend(rows)
         if not answered:
             return None
-        merged["__census__"] = [{"methods": str(methods), "files": str(files), "shards": str(len(shards))}]
+        total: dict[str, object] = {"methods": str(methods), "files": str(files), "shards": str(len(shards))}
+        if named_census:
+            total["file_names"] = sorted(file_names)
+        merged["__census__"] = [total]
         return merged
 
     def _apply_overlays(self, cpg_path: Path, scratch: Path) -> bool:

src/openultrasast/cpg/queries/config.sc

@@ -78,7 +78,7 @@
     // per-file warnings -- so "no rows" and "no graph" are indistinguishable to the driver without this.
     // It rides the batch rather than costing its own invocation, because JVM startup is what this whole
     // batching design exists to avoid.
-    val census = ujson.Arr(ujson.Obj("methods" -> cpg.method.size.toString, "files" -> cpg.file.size.toString))
+    val census = ujson.Arr(ujson.Obj("methods" -> cpg.method.size.toString, "files" -> cpg.file.size.toString, "file_names" -> ujson.Arr.from(cpg.file.name.l)))
     println(ujson.write(ujson.Obj.from(answers.toSeq :+ ("__census__" -> census))))
   } else {
     println(ujson.write(ujson.Arr(rowsFor(settings, function, file): _*)))

src/openultrasast/cpg/queries/dominance.sc

@@ -190,7 +190,7 @@
     // per-file warnings -- so "no rows" and "no graph" are indistinguishable to the driver without this.
     // It rides the batch rather than costing its own invocation, because JVM startup is what this whole
     // batching design exists to avoid.
-    val census = ujson.Arr(ujson.Obj("methods" -> cpg.method.size.toString, "files" -> cpg.file.size.toString))
+    val census = ujson.Arr(ujson.Obj("methods" -> cpg.method.size.toString, "files" -> cpg.file.size.toString, "file_names" -> ujson.Arr.from(cpg.file.name.l)))
     println(ujson.write(ujson.Obj.from(answers.toSeq :+ ("__census__" -> census))))
   } else {
     println(ujson.write(ujson.Arr(rowsFor(operations, dischargers, function, file, operationRequires, dischargersByKind): _*)))

src/openultrasast/cpg/queries/taint.sc

@@ -852,7 +852,7 @@
     // per-file warnings -- so "no rows" and "no graph" are indistinguishable to the driver without this.
     // It rides the batch rather than costing its own invocation, because JVM startup is what this whole
     // batching design exists to avoid.
-    val census = ujson.Arr(ujson.Obj("methods" -> cpg.method.size.toString, "files" -> cpg.file.size.toString))
+    val census = ujson.Arr(ujson.Obj("methods" -> cpg.method.size.toString, "files" -> cpg.file.size.toString, "file_names" -> ujson.Arr.from(cpg.file.name.l)))
     println(ujson.write(ujson.Obj.from(answers.toSeq :+ ("__census__" -> census))))
   } else {
     println(

src/openultrasast/model/layout.py

@@ -26,7 +26,12 @@ def _matches(relative: str, patterns: Iterable[str]) -> bool:
     parts = relative.split("/")
     for pattern in patterns:
         if pattern.endswith("/"):
-            if pattern.rstrip("/") in parts[:-1] or pattern.rstrip("/") in parts:
+            directory = pattern.rstrip("/")
+            if any(
+                fnmatch.fnmatch("/".join(parts[start:end]), directory)
+                for start in range(len(parts))
+                for end in range(start + 1, len(parts) + 1)
+            ):
                 return True
         elif fnmatch.fnmatch(relative, pattern) or fnmatch.fnmatch(parts[-1], pattern):
             return True

src/openultrasast/model/partitions.py

@@ -0,0 +1,279 @@
+"""Physical first-party frontend inputs and routing; selection remains in the scan ranker."""
+
+from __future__ import annotations
+
+import os
+import stat
+import tempfile
+import time
+from collections.abc import Callable, Mapping
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Any
+
+from ..cpg.backend import _DeadlineExpired, _remove_owned_tree
+from ..preprocess import LANGUAGE_BY_EXTENSION
+from .contracts import ExecutionBudget
+from .layout import is_test_path, is_vendored, layout_facts
+
+FRONTENDS = {
+    "php": "php",
+    "javascript": "javascript",
+    "typescript": "typescript",
+    "python": "python",
+    "java": "java",
+    "c": "c",
+    "cpp": "c_cpp",
+}
+
+
+@dataclass(frozen=True)
+class PartitionCoverage:
+    language: str
+    frontend: str | None
+    paths: tuple[str, ...]
+    source_bytes: int
+    status: str
+    boundaries: tuple[str, ...] = ()
+    runtime: str = "unspecified"
+    excluded_paths: tuple[str, ...] = ()
+    unshipped_paths: tuple[str, ...] = ()
+
+
+class PartitionGraph:
+    """Routes existing requests without reordering or allocating another scan budget."""
+
+    def __init__(self, budget: ExecutionBudget | None = None) -> None:
+        self.budget = budget
+        self.diagnostics: list[str] = []
+        self.graphs: dict[str, Any] = {}
+        self.roots: dict[str, Path] = {}
+        self.path_languages: dict[str, str] = {}
+        self.partitions: tuple[PartitionCoverage, ...] = ()
+        self.boundaries: tuple[str, ...] = ()
+        self.unparsed: tuple[str, ...] = ()
+        self.cpg_path = Path("partitioned-cpg")
+
+    def cleanup(self) -> None:
+        try:
+            for graph in self.graphs.values():
+                cleanup = getattr(graph, "cleanup", None)
+                if callable(cleanup):
+                    cleanup()
+        finally:
+            for root in self.roots.values():
+                deadline = self.budget.deadline_monotonic + self.budget.cancellation_allowance_seconds if self.budget else float("inf")
+                try:
+                    _remove_owned_tree(root, deadline)
+                except (OSError, _DeadlineExpired):
+                    self.diagnostics.append("scratch_cleanup_incomplete")
+
+    def execution_diagnostics(self) -> tuple[str, ...]:
+        reasons = list(self.diagnostics)
+        for graph in self.graphs.values():
+            read = getattr(graph, "execution_diagnostics", None)
+            if callable(read):
+                reasons.extend(read())
+        return tuple(reasons)
+
+    def _normalize(self, value: Any) -> Any:
+        if isinstance(value, str):
+            for root in self.roots.values():
+                value = value.replace(str(root) + "/", "")
+            return value
+        if isinstance(value, list):
+            return [self._normalize(v) for v in value]
+        if isinstance(value, dict):
+            return {k: self._normalize(v) for k, v in value.items()}
+        return value
+
+    def run(self, kind: str, params: Mapping[str, object]) -> list[object] | None:
+        return self.run_batch(kind, {"one": params}).get("one")
+
+    def run_batch(self, kind: str, requests: Mapping[str, Mapping[str, object]]) -> dict[str, list[object]]:
+        answers: dict[str, list[object]] = {}
+        languages = dict.fromkeys(self.path_languages.get(str(params.get("file", "")), "") for params in requests.values())
+        for language in languages:
+            graph = self.graphs.get(language)
+            if graph is None:
+                continue
+            if self.budget is not None and time.monotonic() >= self.budget.deadline_monotonic:
+                break
+            selected = {rid: params for rid, params in requests.items() if self.path_languages.get(str(params.get("file", ""))) == language}
+            if not selected:
+                continue
+            batch = getattr(graph, "run_batch", None)
+            raw = batch(kind, selected) if callable(batch) else {rid: graph.run(kind, params) for rid, params in selected.items()}
+            if not isinstance(raw, Mapping):
+                continue
+            for rid, rows in raw.items():
+                if rid in selected and isinstance(rows, list):
+                    answers[rid] = self._normalize(rows)
+                elif rid == "__census__" and isinstance(rows, list):
+                    expected = set(next(p.paths for p in self.partitions if p.language == language))
+                    normalized = self._normalize(rows)
+                    names = [r.get("file_names") for r in normalized if isinstance(r, Mapping)]
+                    named = bool(names) and all(isinstance(n, list) and all(isinstance(p, str) for p in n) for n in names)
+                    observed = {p.removeprefix("./") for n in names if isinstance(n, list) for p in n if isinstance(p, str)}
+                    if not named or not expected <= observed:
+                        self.diagnostics.append("partition_file_census_incomplete" if named else "partition_file_census_unavailable")
+                        rows = [
+                            {
+                                "methods": 0,
+                                "files": 0,
+                                "shards": max((int(str(r.get("shards", 1))) for r in rows if isinstance(r, Mapping)), default=1),
+                            }
+                        ]
+                    # Preserve any individual empty or sharded graph's degradation.
+                    previous = answers.get(rid, [])
+                    if not previous or any(
+                        isinstance(r, Mapping) and (int(str(r.get("methods", 0))) == 0 or int(str(r.get("shards", 1))) > 1) for r in rows
+                    ):
+                        answers[rid] = rows
+        return answers
+
+
+def build_partitions(
+    root: Path, build: Callable[[Path, str], Any], budget: ExecutionBudget | None, declared: Mapping[str, tuple[str, ...]] | None = None
+) -> PartitionGraph:
+    graph = PartitionGraph(budget)
+    paths: dict[str, list[Path]] = {}
+    boundaries: set[str] = set()
+    excluded: set[str] = set()
+    facts = layout_facts()
+    root_fd = os.open(root, os.O_RDONLY | os.O_DIRECTORY | os.O_NOFOLLOW)
+    try:
+        for directory, dirs, files, directory_fd in os.fwalk(
+            ".", follow_symlinks=False, dir_fd=root_fd, onerror=lambda error: boundaries.add("source_unreadable")
+        ):
+            if budget is not None and time.monotonic() >= budget.deadline_monotonic:
+                boundaries.add("deadline_exhausted")
+                break
+            base = root / directory
+            for name in list(dirs):
+                if budget is not None and time.monotonic() >= budget.deadline_monotonic:
+                    boundaries.add("deadline_exhausted")
+                    dirs.clear()
+                    break
+                path = base / name
+                relative = path.relative_to(root).as_posix()
+                if name in {".git", ".hg", ".svn", ".openultrasast", ".venv", "venv", "__pycache__"}:
+                    excluded.add(relative + "/")
+                    dirs.remove(name)
+                    continue
+                if path.is_symlink() or is_vendored(relative + "/", facts):
+                    excluded.add(relative + "/")
+                    boundaries.add("symlink_context_unresolved" if path.is_symlink() else "vendor_semantics_unresolved")
+                    dirs.remove(name)
+            for name in files:
+                if budget is not None and time.monotonic() >= budget.deadline_monotonic:
+                    boundaries.add("deadline_exhausted")
+                    break
+                path = base / name
+                relative = path.relative_to(root).as_posix()
+                if is_vendored(relative, facts):
+                    excluded.add(relative)
+                    boundaries.add("vendor_semantics_unresolved")
+                    continue
+                if path.is_symlink() or not path.is_file():
+                    excluded.add(relative)
+                    boundaries.add("symlink_context_unresolved")
+                    continue
+                language = LANGUAGE_BY_EXTENSION.get(path.suffix.lower(), "unknown")
+                if language == "unknown":
+                    try:
+                        fd = os.open(name, os.O_RDONLY | os.O_NOFOLLOW | os.O_NONBLOCK, dir_fd=directory_fd)
+                        with os.fdopen(fd, "rb") as handle:
+                            first = handle.readline(4096).lower()
+                    except OSError:
+                        boundaries.add("source_unreadable")
+                        continue
+                    if first.startswith(b"#!"):
+                        language = "python" if b"python" in first else "javascript" if b"node" in first else "unknown"
+                languages = (language,) if language != "unknown" else (declared or {}).get(relative, ())
+                for item in languages:
+                    paths.setdefault(item, []).append(path)
+                if len(languages) == 1:
+                    graph.path_languages[relative] = languages[0]
+                elif languages:
+                    boundaries.add("ambiguous_frontend_path")
+        if len(paths) > 1:
+            boundaries.add("cross_partition_semantics_unresolved")
+        records: list[PartitionCoverage] = []
+        for language, sources in sorted(paths.items()):
+            relative_paths = tuple(sorted(p.relative_to(root).as_posix() for p in sources))
+            frontend = FRONTENDS.get(language)
+            if frontend is None:
+                records.append(PartitionCoverage(language, None, relative_paths, 0, "unsupported", ("frontend_unsupported",)))
+                boundaries.add("frontend_unsupported")
+                continue
+            scratch = Path(tempfile.mkdtemp(prefix="ousast-input-"))
+            graph.roots[language] = scratch
+            size = 0
+            complete = True
+            for source in sources:
+                if budget is not None and time.monotonic() >= budget.deadline_monotonic:
+                    complete = False
+                    break
+                target = scratch / source.relative_to(root)
+                target.parent.mkdir(parents=True, exist_ok=True)
+                # A no-follow open prevents file symlink substitution between census and read.
+                try:
+                    fd = _open_source(root_fd, source.relative_to(root).parts)
+                except OSError:
+                    boundaries.add("source_unreadable")
+                    complete = False
+                    break
+                with os.fdopen(fd, "rb") as handle, target.open("wb") as output:
+                    while chunk := handle.read(1024 * 1024):
+                        if budget is not None and time.monotonic() >= budget.deadline_monotonic:
+                            complete = False
+                            break
+                        output.write(chunk)
+                        size += len(chunk)
+                if not complete:
+                    break
+            built = build(scratch, frontend) if complete else None
+            if built is not None:
+                graph.graphs[language] = built
+                graph.unparsed += tuple(graph._normalize(list(getattr(built, "unparsed", ()))))
+            record_limits: tuple[str, ...] = ("runtime_unspecified",) if language in {"javascript", "typescript"} else ()
+            if language == "typescript":
+                record_limits += ("typescript_property_support_unvalidated",)
+                boundaries.add("typescript_property_support_unvalidated")
+            if language in {"c", "cpp"}:
+                record_limits += ("c_bounds_arithmetic_unmodeled",)
+            records.append(
+                PartitionCoverage(language, frontend, relative_paths, size, "built" if built is not None else "build_failed", record_limits)
+            )
+        from dataclasses import replace
+
+        graph.partitions = tuple(
+            replace(
+                record, excluded_paths=tuple(sorted(excluded)), unshipped_paths=tuple(p for p in record.paths if is_test_path(p, facts))
+            )
+            for record in records
+        )
+        graph.boundaries = tuple(sorted(boundaries))
+        return graph
+    except BaseException:
+        graph.cleanup()
+        raise
+    finally:
+        os.close(root_fd)
+
+
+def _open_source(root_fd: int, parts: tuple[str, ...]) -> int:
+    descriptor = os.dup(root_fd)
+    try:
+        for component in parts[:-1]:
+            child = os.open(component, os.O_RDONLY | os.O_DIRECTORY | os.O_NOFOLLOW, dir_fd=descriptor)
+            os.close(descriptor)
+            descriptor = child
+        result = os.open(parts[-1], os.O_RDONLY | os.O_NOFOLLOW | os.O_NONBLOCK, dir_fd=descriptor)
+        if not stat.S_ISREG(os.fstat(result).st_mode):
+            os.close(result)
+            raise OSError("source is not a regular file")
+        return result
+    finally:
+        os.close(descriptor)

src/openultrasast/model/scan.py

@@ -43,7 +43,8 @@ from .contracts import (
 from .dominance import request_params as dominance_params
 from .evidence import TIER_EXCLUDE, Evidence, evidence_from_rows
 from .ladder import Rung
-from .layout import with_layout
+from .layout import is_vendored, layout_facts, with_layout
+from .partitions import PartitionCoverage, build_partitions
 from .pipeline import ModelFinding, scan_region
 from .regions import MODULE_SCOPE, ScanRegion, affected_context
 from .specs import ConfigSpec, DominanceSpec, TaintSpec, config_specs, dominance_specs, taint_specs
@@ -126,6 +127,7 @@ class ModelScanResult:
     # could have been found was skipped. Reported so the request count is legible.
     requests_pruned: int = 0
 
+    partitions: tuple[PartitionCoverage, ...] = ()
     change_context: ChangeContext | None = None
     scope: ScopeDecision | None = None
     question_outcomes: tuple[QuestionOutcome, ...] = ()
@@ -158,6 +160,8 @@ def scan_repository(
     if ranking_mode is not None:
         limits = replace(limits, order_by_evidence=ranking_mode == "evidence")
     mode = "evidence" if limits.order_by_evidence else "static"
+    layout = layout_facts()
+    regions = tuple(r for r in regions if not is_vendored(r.path, layout))
     # Reject duplicate canonical questions before any engine work or resource ownership.
     identities = [_identity(unit, r, f) for r in regions for f in r.families]
     if len(set(identities)) != len(identities):
@@ -223,13 +227,29 @@ def _scan_repository_impl(
     for region in regions:
         counts[region.language] = counts.get(region.language, 0) + 1
     dominant = max(counts, key=lambda name: (counts[name], name)) if counts else ""
-    cpg = _build(backend, root, dominant, execution_budget=execution_budget)
+    declarations: dict[str, set[str]] = {}
+    for region in regions:
+        declarations.setdefault(region.path, set()).add(region.language)
+    if root.is_dir():
+        cpg = build_partitions(
+            root,
+            lambda source, language: _build(backend, source, language, execution_budget=execution_budget),
+            execution_budget,
+            {path: tuple(sorted(languages)) for path, languages in declarations.items()},
+        )
+        degradations.extend({"stage": "model", "reason": reason} for reason in cpg.boundaries)
+    else:
+        cpg = _build(backend, root, dominant, execution_budget=execution_budget)
     build_seconds = round(time.monotonic() - build_started, 2)
-    if cpg is None:
+    if cpg is None or (hasattr(cpg, "graphs") and not cpg.graphs):
+        if cpg is not None:
+            owned.append(cpg)
         scope, _ = _scope_work(
             regions, {}, limits, unit, ranking_mode, population_complete, failure="cpg_build_failed", change_context=change_context
         )
+        scope = replace(scope, unresolved_boundaries=tuple(dict.fromkeys((*scope.unresolved_boundaries, *getattr(cpg, "boundaries", ())))))
         return ModelScanResult(
+            partitions=getattr(cpg, "partitions", ()),
             change_context=change_context,
             scope=scope,
             family_coverage=_family_coverage(scope, ()),
@@ -322,7 +342,22 @@ def _scan_repository_impl(
     # Stable question IDs are the IDs sent to every execution request below.
     scope_limits = (
         replace(limits, prune_tier0=False)
-        if any(d.get("reason") in {"files_unparsed", "cpg_empty", "cpg_sharded"} for d in degradations)
+        if any(
+            d.get("reason")
+            in {
+                "files_unparsed",
+                "cpg_empty",
+                "cpg_sharded",
+                "cross_partition_semantics_unresolved",
+                "vendor_semantics_unresolved",
+                "symlink_context_unresolved",
+                "frontend_unsupported",
+                "source_unreadable",
+                "ambiguous_frontend_path",
+                "typescript_property_support_unvalidated",
+            }
+            for d in degradations
+        )
         else limits
     )
     scope, work = _scope_work(
@@ -457,16 +492,32 @@ def _scan_repository_impl(
     if unasked:
         degradations.append({"stage": "model", "reason": "budget_exhausted", "regions_unasked": len(unasked)})
 
-    graph_gaps = tuple(str(d["reason"]) for d in degradations if d.get("reason") in {"files_unparsed", "cpg_empty", "cpg_sharded"})
+    graph_gaps = tuple(
+        str(d["reason"])
+        for d in degradations
+        if d.get("reason")
+        in {
+            "files_unparsed",
+            "cpg_empty",
+            "cpg_sharded",
+            "cross_partition_semantics_unresolved",
+            "vendor_semantics_unresolved",
+            "symlink_context_unresolved",
+            "frontend_unsupported",
+            "source_unreadable",
+            "ambiguous_frontend_path",
+            "typescript_property_support_unvalidated",
+        }
+    )
     for question in scope.selected:
         rid = question.identity.question_id
         if rid not in outcomes:
-            answered = rid in rows_by_id
+            has_answer = rid in rows_by_id
             outcomes[rid] = QuestionOutcome(
                 question.identity,
-                "not_arbitrated" if answered else "unanswered",
+                "not_arbitrated" if has_answer else "unanswered",
                 "deadline_exhausted" if not _within_deadline(execution_budget) else "query_unanswered",
-                _rows_json(rows_by_id[rid]) if answered else None,
+                _rows_json(rows_by_id[rid]) if has_answer else None,
             )
         if question.identity in context_gaps and outcomes[rid].status == "completed":
             outcomes[rid] = replace(outcomes[rid], status="unresolved", reason="change_context_incomplete")
@@ -480,6 +531,7 @@ def _scan_repository_impl(
     # The graph has answered everything it is going to. Its scratch tree holds the CPG, the request files and
     # joern's own working copy -- tens of megabytes per scan -- and nothing else reclaims it.
     return ModelScanResult(
+        partitions=getattr(cpg, "partitions", ()),
         findings=findings,
         by_rung=_tally(findings),
         regions_scanned=scanned,

src/openultrasast/preprocess.py

@@ -97,7 +97,13 @@ def enumerate_source_files(root: Path) -> list[Path]:
     patterns = _load_ignore_patterns(root)
     paths: list[Path] = []
     for path in root.rglob("*"):
-        if not path.is_file() or _has_ignored_dir(root, path) or _is_ignored(root, path, patterns) or _is_vendored(root, path):
+        if (
+            path.is_symlink()
+            or not path.is_file()
+            or _has_ignored_dir(root, path)
+            or _is_ignored(root, path, patterns)
+            or _is_vendored(root, path)
+        ):
             continue
         if detect_language(path) != "unknown":
             paths.append(path)

src/openultrasast/ruleset/semantic/javascript.toml

@@ -92,3 +92,18 @@ calls = ["path.normalize", "path.resolve"]
 # CWE-79. The standard encoders: the DOM/global URI encoders and the common escaping helpers.
 id = "html.escape"
 calls = ["encodeURIComponent", "encodeURI", "escapeHtml", "escape", "sanitizeHtml", "DOMPurify.sanitize"]
+
+
+[[layout]]
+# First-party test conventions are retained as graph context, ordered as unshipped.
+# These declarations cover JS module/JSX and TS/TSX filenames; they do not assert
+# TypeScript property support or a browser/server runtime classification.
+id = "javascript_first_party_tests"
+vendored = ["node_modules/"]
+tests = [
+  "tests/", "test/", "spec/", "__tests__/", "__mocks__/", "e2e/",
+  "*.test.js", "*.test.mjs", "*.test.cjs", "*.test.jsx", "*.test.ts", "*.test.mts", "*.test.cts", "*.test.tsx",
+  "*.spec.js", "*.spec.mjs", "*.spec.cjs", "*.spec.jsx", "*.spec.ts", "*.spec.mts", "*.spec.cts", "*.spec.tsx",
+  "*.mock.js", "*.mock.mjs", "*.mock.cjs", "*.mock.jsx", "*.mock.ts", "*.mock.mts", "*.mock.cts", "*.mock.tsx",
+  "*.e2e.js", "*.e2e.mjs", "*.e2e.cjs", "*.e2e.jsx", "*.e2e.ts", "*.e2e.mts", "*.e2e.cts", "*.e2e.tsx",
+]

tests/test_cpg_backend.py

@@ -956,6 +956,36 @@ def test_a_source_file_the_size_of_a_data_table_is_excluded_and_named(tmp_path:
     assert [Path(name).name for name in result.unparsed] == ["profiles.php"], "and it is reported, not silently dropped"
 
 
+@pytest.mark.parametrize("shard_names", [("app.php",), ("app.php", "helper.php")])
+def test_shard_aggregation_preserves_complete_named_census(monkeypatch, shard_names):
+    from openultrasast.cpg.backend import JoernBackend
+
+    def batch(self, path, query, requests):
+        return {"__census__": [{"methods": "2", "files": "2", "file_names": ["<unknown>", path.name]}]}
+
+    monkeypatch.setattr(JoernBackend, "query_batch", batch)
+    result = JoernBackend().query_batch_across([Path(name) for name in shard_names], "taint", {})
+    assert result is not None
+    assert set(result["__census__"][0]["file_names"]) == {"<unknown>", *shard_names}
+    assert result["__census__"][0]["shards"] == str(len(shard_names))
+
+
+@pytest.mark.parametrize("missing", ["names", "census", "answer"])
+def test_shard_aggregation_does_not_invent_complete_named_census(monkeypatch, missing):
+    from openultrasast.cpg.backend import JoernBackend
+
+    def batch(self, path, query, requests):
+        if path.name == "missing":
+            return None if missing == "answer" else {} if missing == "census" else {"__census__": [{"methods": "2", "files": "2"}]}
+        return {"__census__": [{"methods": "2", "files": "2", "file_names": ["app.php", "<unknown>"]}]}
+
+    monkeypatch.setattr(JoernBackend, "query_batch", batch)
+    result = JoernBackend().query_batch_across([Path("app.php"), Path("missing")], "taint", {})
+    assert result is not None
+    assert "file_names" not in result["__census__"][0]
+    assert result["__census__"][0]["shards"] == "2"
+
+
 def test_a_clean_build_whose_census_cannot_be_taken_is_a_failed_build(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
     """`cpg query census failed: timeout` was logged as a warning during a build whose every later query
     then timed out too. A graph that cannot answer the census inside the query timeout cannot answer a taint

tests/test_cpg_budget.py

@@ -261,6 +261,7 @@ def test_cleanup_does_not_follow_substituted_directory_links(tmp_path: Path) ->
 
 
 def test_driver_propagates_budget_and_rejects_legacy_without_launch(tmp_path: Path) -> None:
+    (tmp_path / "input.py").write_bytes(b"print(1)\n")
     from openultrasast.cpg.backend import CpgResult
     from openultrasast.model.scan import scan_repository
 
@@ -296,6 +297,7 @@ def test_driver_propagates_budget_and_rejects_legacy_without_launch(tmp_path: Pa
 
 
 def test_budgeted_driver_does_not_retry_internal_typeerror(tmp_path: Path) -> None:
+    (tmp_path / "input.py").write_bytes(b"print(1)\n")
     from openultrasast.model.scan import scan_repository
 
     class Backend:
@@ -337,6 +339,7 @@ def test_cancelled_graph_session_does_not_poison_next_build(tmp_path: Path, monk
 
 
 def test_driver_cleans_graph_on_unexpected_exception(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
+    (tmp_path / "input.py").write_bytes(b"print(1)\n")
     from openultrasast.cpg.backend import CpgResult
     from openultrasast.model import scan
 

tests/test_layout_facts.py

@@ -65,11 +65,13 @@ def test_test_regions_are_not_product_and_order_last() -> None:
     )
 
 
-def test_the_scan_hands_the_vendored_trees_to_the_build(tmp_path: Path) -> None:
+def test_the_scan_excludes_vendored_trees_before_the_build(tmp_path: Path) -> None:
     from openultrasast.model.regions import ScanRegion
     from openultrasast.model.scan import ScanBudget, scan_repository
 
     (tmp_path / "vendor").mkdir()
+    (tmp_path / "a.php").write_bytes(b"<?php echo 1;")
+    (tmp_path / "vendor/b.php").write_bytes(b"<?php vendor_secret();")
     seen: dict[str, object] = {}
 
     class _Backend:
@@ -80,8 +82,10 @@ def test_the_scan_hands_the_vendored_trees_to_the_build(tmp_path: Path) -> None:
             from openultrasast.cpg.backend import CpgResult
 
             seen["exclude"] = tuple(exclude)
+            seen["files"] = tuple(p.relative_to(root).as_posix() for p in root.rglob("*") if p.is_file())
             return CpgResult(cpg_path=Path("cpg.bin"), run=lambda q, p: [], run_batch=lambda q, r: {rid: [] for rid in r})
 
     region = ScanRegion(path="a.php", function="f", language="php", families=("injection",), rank=0.5, source="x")
     scan_repository(tmp_path, [region], backend=_Backend(), budget=ScanBudget(max_model_calls=0))
-    assert seen["exclude"] == ("vendor",)
+    assert seen["exclude"] == ()
+    assert seen["files"] == ("a.php",)

tests/test_model_partitions.py

@@ -0,0 +1,337 @@
+from pathlib import Path
+
+from openultrasast.cpg.backend import CpgResult
+from openultrasast.model.regions import ScanRegion
+from openultrasast.model.scan import ScanBudget, scan_repository
+
+
+def test_polyglot_graph_inputs_and_single_scope(tmp_path: Path) -> None:
+    sources = {
+        "api.php": "<?php function f($x) { echo $x; }",
+        "api.js": "function f(x) { return x; }",
+        "vendor/bad.php": "<?php echo 1;",
+        "node_modules/bad.js": "bad()",
+        "tests/test.js": "test()",
+        "other.rs": "fn main() {}",
+    }
+    for name, content in sources.items():
+        path = tmp_path / name
+        path.parent.mkdir(parents=True, exist_ok=True)
+        path.write_text(content)
+    seen = []
+
+    class Backend:
+        def build(self, root, *, language="", **kwargs):
+            files = tuple(sorted(p.relative_to(root).as_posix() for p in root.rglob("*") if p.is_file()))
+            assert all(p.read_bytes() for p in root.rglob("*") if p.is_file())
+            seen.append((language, files))
+            return CpgResult(Path("cpg.bin"), lambda q, p: [], run_batch=lambda q, requests: {rid: [] for rid in requests})
+
+    regions = [
+        ScanRegion(path=p, function="f", language=language, families=("injection",), rank=r, source="entry_point")
+        for p, language, r in [
+            ("api.php", "php", 1.0),
+            ("api.js", "javascript", 0.9),
+            ("vendor/bad.php", "php", 2.0),
+            ("tests/test.js", "javascript", 3.0),
+        ]
+    ]
+    result = scan_repository(
+        tmp_path, regions, backend=Backend(), budget=ScanBudget(max_regions=1, tiering=False), population_complete=True
+    )
+    assert seen == [("javascript", ("api.js", "tests/test.js")), ("php", ("api.php",))]
+    assert result.scope.selected[0].identity.path == "api.php"
+    assert all(q.identity.path != "vendor/bad.php" for q in (*result.scope.selected, *result.scope.deferred))
+    assert any(p.language == "rust" and p.status == "unsupported" for p in result.partitions)
+    assert "cross_partition_semantics_unresolved" in result.scope.unresolved_boundaries
+
+
+def test_file_patterns_symlinks_and_witness_paths(tmp_path, monkeypatch):
+    import dataclasses
+
+    from openultrasast.model import layout, partitions
+    from openultrasast.preprocess import preprocess_repository
+    from openultrasast.semantic.facts import LayoutFact
+
+    facts = (LayoutFact(id="declared", language="javascript", vendored=("*.bundle.js", "bundled/lib/"), tests=("tests/",)),)
+    monkeypatch.setattr(layout, "layout_facts", lambda *args: facts)
+    monkeypatch.setattr(partitions, "layout_facts", lambda *args: facts)
+    outside = tmp_path.parent / "external.js"
+    outside.write_text("external_secret()")
+    (tmp_path / "escape.js").symlink_to(outside)
+    (tmp_path / "bad.bundle.js").write_text("vendor_secret()")
+    (tmp_path / "app.js").write_text("function f(x) { return x; }")
+    (tmp_path / "bundled/lib").mkdir(parents=True)
+    (tmp_path / "bundled/lib/hidden.js").write_text("vendor_secret()")
+    _, targets = preprocess_repository(tmp_path)
+    assert [t.path for t in targets] == ["app.js"]
+    roots = []
+
+    def build(root, language):
+        roots.append(root)
+        assert [p.relative_to(root).as_posix() for p in root.rglob("*") if p.is_file()] == ["app.js"]
+        return CpgResult(
+            Path("graph"),
+            lambda q, p: [],
+            run_batch=lambda q, reqs: {rid: [{"file": str(root / "app.js"), "site": str(root / "app.js") + ":1"}] for rid in reqs},
+        )
+
+    graph = partitions.build_partitions(tmp_path, build, None)
+    try:
+        assert graph.partitions[0].source_bytes == len((tmp_path / "app.js").read_bytes())
+        assert graph.run("taint", {"file": "app.js"}) == [{"file": "app.js", "site": "app.js:1"}]
+        assert {"vendor_semantics_unresolved", "symlink_context_unresolved"} <= set(graph.boundaries)
+        assert dataclasses.asdict(graph.partitions[0])["runtime"] == "unspecified"
+    finally:
+        graph.cleanup()
+    assert all(not root.exists() for root in roots)
+
+
+def test_shared_deadline_stops_later_frontend_and_no_false_clean(tmp_path):
+    import time
+
+    from openultrasast.model.contracts import ExecutionBudget
+    from openultrasast.model.partitions import build_partitions
+
+    for name in ("a.js", "b.php"):
+        (tmp_path / name).write_text("readable")
+    seen = []
+    budget = ExecutionBudget(time.monotonic() + 0.1, 0.1)
+
+    def build(root, language):
+        seen.append(language)
+        time.sleep(0.12)
+        return CpgResult(Path("graph"), lambda q, p: [])
+
+    graph = build_partitions(tmp_path, build, budget)
+    try:
+        assert seen == ["javascript"]
+        assert [p.status for p in graph.partitions] == ["built", "build_failed"]
+        assert graph.run("taint", {"file": "b.php"}) is None
+    finally:
+        graph.cleanup()
+
+
+def test_evidence_ranker_selects_once_across_languages(tmp_path):
+    from test_model_scope import vector
+
+    for name in ("a.php", "b.js"):
+        (tmp_path / name).write_text("readable source")
+    executed = []
+
+    class Backend:
+        def build(self, root, *, language="", **kwargs):
+            def batch(kind, requests):
+                if any(p.get("evidenceOnly") for p in requests.values()):
+                    return {rid: vector(language == "javascript") for rid in requests}
+                executed.extend(p["file"] for p in requests.values())
+                return {rid: [] for rid in requests}
+
+            return CpgResult(Path("graph"), lambda q, p: [], run_batch=batch)
+
+    regions = [
+        ScanRegion(path=path, function="f", language=language, families=("injection",), rank=rank, source="entry_point")
+        for path, language, rank in [("a.php", "php", 1.0), ("b.js", "javascript", 0.5)]
+    ]
+    result = scan_repository(tmp_path, regions, backend=Backend(), ranking_mode="evidence", budget=ScanBudget(max_regions=1))
+    assert len(result.scope.selected) == 1
+    assert result.scope.selected[0].identity.path == "b.js"
+    assert executed == ["b.js"]
+    assert result.question_outcomes[0].status == "unresolved"
+
+
+def test_unsupported_only_population_is_named(tmp_path):
+    (tmp_path / "main.go").write_text("package main")
+
+    class Backend:
+        def build(self, *args, **kwargs):
+            raise AssertionError("unsupported frontend must not build")
+
+    result = scan_repository(tmp_path, [], backend=Backend(), population_complete=True)
+    assert result.partitions[0].language == "go"
+    assert result.partitions[0].status == "unsupported"
+    assert "frontend_unsupported" in result.scope.unresolved_boundaries
+
+
+def test_partial_partition_census_never_completes_questions(tmp_path):
+    for name in ("a.py", "b.py"):
+        (tmp_path / name).write_text("print('readable')")
+
+    class Backend:
+        def build(self, root, **kwargs):
+            return CpgResult(
+                Path("graph"),
+                lambda q, p: [],
+                run_batch=lambda q, reqs: {**{rid: [] for rid in reqs}, "__census__": [{"methods": 2, "files": 1}]},
+            )
+
+    region = ScanRegion(path="a.py", function="f", language="python", families=("injection",), rank=1.0, source="entry_point")
+    result = scan_repository(tmp_path, [region], backend=Backend(), budget=ScanBudget(tiering=False))
+    assert result.question_outcomes[0].status == "unresolved"
+    assert "cpg_empty" in result.scope.unresolved_boundaries
+
+
+def test_projection_refuses_substituted_ancestor(tmp_path):
+    import os
+
+    import pytest
+
+    from openultrasast.model.partitions import _open_source
+
+    source = tmp_path / "source"
+    source.mkdir()
+    outside = tmp_path / "outside"
+    outside.mkdir()
+    (outside / "secret.js").write_bytes(b"must not enter the graph")
+    (source / "nested").symlink_to(outside, target_is_directory=True)
+    descriptor = os.open(source, os.O_RDONLY | os.O_DIRECTORY)
+    try:
+        with pytest.raises(OSError):
+            _open_source(descriptor, ("nested", "secret.js"))
+    finally:
+        os.close(descriptor)
+
+
+def test_partition_records_reach_cli(tmp_path):
+    from openultrasast.cli import _model_payload
+
+    (tmp_path / "main.go").write_bytes(b"package main")
+    result = scan_repository(tmp_path, [], backend=object())
+    payload = _model_payload(result)
+    assert payload["partitions"][0]["language"] == "go"
+    assert payload["partitions"][0]["status"] == "unsupported"
+
+
+def test_stale_region_cannot_bypass_vendor_graph_exclusion(tmp_path):
+    (tmp_path / "api.php").write_bytes(b"<?php echo 1;")
+    (tmp_path / "vendor").mkdir()
+    (tmp_path / "vendor/bad.php").write_bytes(b"<?php vendor_secret();")
+    seen = []
+
+    class Backend:
+        def build(self, root, **kwargs):
+            seen.extend(p.relative_to(root).as_posix() for p in root.rglob("*") if p.is_file())
+            return CpgResult(Path("graph"), lambda q, p: [])
+
+    region = ScanRegion(path="missing.php", function="f", language="php", families=("injection",), rank=1.0, source="entry_point")
+    result = scan_repository(tmp_path, [region], backend=Backend())
+    assert seen == ["api.php"]
+    assert result.question_outcomes[0].status == "unanswered"
+
+
+def test_unreadable_unknown_extension_reports_incomplete_without_crashing(tmp_path, monkeypatch):
+    from openultrasast.model import partitions
+
+    (tmp_path / "app.js").write_bytes(b"function handler(value) { return value; }\n")
+    (tmp_path / "README.md").write_bytes(b"unreadable declaration candidate\n")
+    original_open = partitions.os.open
+
+    def unreadable_open(path, flags, *args, **kwargs):
+        if str(path) == "README.md":
+            raise PermissionError("controlled unreadable discovery input")
+        return original_open(path, flags, *args, **kwargs)
+
+    monkeypatch.setattr(partitions.os, "open", unreadable_open)
+
+    class Backend:
+        def build(self, root, **kwargs):
+            assert (root / "app.js").read_bytes()
+            return CpgResult(Path("graph"), lambda query, params: [])
+
+    region = ScanRegion("app.js", "handler", "javascript", ("injection",), 1.0, "entry_point")
+    result = scan_repository(tmp_path, [region], backend=Backend(), budget=ScanBudget(tiering=False))
+    assert result.partitions[0].status == "built"
+    assert "source_unreadable" in result.scope.unresolved_boundaries
+    assert result.question_outcomes[0].status == "unresolved"
+
+
+def test_declared_javascript_tests_remain_inputs_but_unshipped(tmp_path):
+    from openultrasast.model.layout import with_layout
+    from openultrasast.model.partitions import build_partitions
+
+    names = ("unit.test.js", "unit.spec.mjs", "unit.mock.cjs", "unit.e2e.jsx", "unit.test.ts", "unit.spec.tsx", "__tests__/unit.js")
+    for name in (*names, "app.js"):
+        source = tmp_path / name
+        source.parent.mkdir(parents=True, exist_ok=True)
+        source.write_text("function readable() {}")
+    graph = build_partitions(tmp_path, lambda root, language: CpgResult(Path("graph"), lambda q, p: []), None)
+    try:
+        assert set(names) <= {p for record in graph.partitions for p in record.paths}
+        assert set(names) == {p for record in graph.partitions for p in record.unshipped_paths}
+        regions = [
+            ScanRegion(path=p, function="readable", language="javascript", families=("injection",), rank=1.0, source="entry_point")
+            for p in (*names, "app.js")
+        ]
+        assert {r.path for r in with_layout(regions) if not r.shipped} == set(names)
+    finally:
+        graph.cleanup()
+
+
+def test_synthetic_census_node_cannot_mask_missing_source(tmp_path):
+    from openultrasast.model.partitions import build_partitions
+
+    for name in ("a.php", "b.php"):
+        (tmp_path / name).write_text("<?php echo 1;")
+
+    def build(root, language):
+        return CpgResult(
+            Path("graph"),
+            lambda q, p: [],
+            run_batch=lambda q, reqs: {
+                **{rid: [] for rid in reqs},
+                "__census__": [{"methods": 3, "files": 2, "file_names": ["<unknown>", str(root / "a.php")]}],
+            },
+        )
+
+    graph = build_partitions(tmp_path, build, None)
+    try:
+        result = graph.run_batch("taint", {"q": {"file": "a.php"}})
+        assert result["__census__"][0]["files"] == 0
+        assert "partition_file_census_incomplete" in graph.execution_diagnostics()
+    finally:
+        graph.cleanup()
+
+
+def test_named_census_accepts_exact_paths_for_every_query_kind(tmp_path):
+    from openultrasast.model.partitions import build_partitions
+
+    (tmp_path / "app.php").write_text("<?php echo 1;")
+
+    def build(root, language):
+        return CpgResult(
+            Path("graph"),
+            lambda q, p: [],
+            run_batch=lambda q, reqs: {
+                **{rid: [] for rid in reqs},
+                "__census__": [{"methods": 2, "files": 2, "file_names": ["<unknown>", str(root / "app.php")]}],
+            },
+        )
+
+    graph = build_partitions(tmp_path, build, None)
+    try:
+        for kind in ("taint", "dominance", "config"):
+            assert graph.run_batch(kind, {"q": {"file": "app.php"}})["__census__"][0]["files"] == 2
+        assert not graph.execution_diagnostics()
+    finally:
+        graph.cleanup()
+
+
+def test_unnamed_sharded_census_remains_explicitly_incomplete(tmp_path):
+    from openultrasast.model.partitions import build_partitions
+
+    (tmp_path / "app.php").write_text("<?php echo 1;")
+    graph = build_partitions(
+        tmp_path,
+        lambda root, language: CpgResult(
+            Path("graph"),
+            lambda q, p: [],
+            run_batch=lambda q, reqs: {**{rid: [] for rid in reqs}, "__census__": [{"methods": 20, "files": 20, "shards": 2}]},
+        ),
+        None,
+    )
+    try:
+        result = graph.run_batch("taint", {"q": {"file": "app.php"}})
+        assert result["__census__"] == [{"methods": 0, "files": 0, "shards": 2}]
+        assert "partition_file_census_unavailable" in graph.execution_diagnostics()
+    finally:
+        graph.cleanup()

tests/test_model_scan.py

@@ -394,7 +394,7 @@ def test_a_query_the_engine_could_not_answer_is_recorded_not_ignored() -> None:
     assert result.findings == ()
 
 
-def test_a_region_cap_that_hides_most_of_a_repository_is_reported() -> None:
+def test_a_region_cap_that_hides_most_of_a_repository_is_reported(tmp_path: Path) -> None:
     """Paid Memberships Pro yields 4,463 regions from 637 files. A default budget of 500 examines 11% of it,
     and a report that does not say so invites its silence to be read as a clean bill of health for the rest.
     """
@@ -414,14 +414,16 @@ def test_a_region_cap_that_hides_most_of_a_repository_is_reported() -> None:
         ScanRegion(path=f"f{n}.py", function="handler", language="python", families=("injection",), rank=0.5, source="entry_point")
         for n in range(30)
     ]
-    result = scan_repository(Path("."), regions, backend=_Backend(), budget=ScanBudget(max_model_calls=0, max_regions=10))
+    for region in regions:
+        (tmp_path / region.path).write_bytes(b"print(1)\n")
+    result = scan_repository(tmp_path, regions, backend=_Backend(), budget=ScanBudget(max_model_calls=0, max_regions=10))
 
     truncated = [d for d in result.degradations if d.get("reason") == "regions_truncated"]
     assert truncated, "the cap hid two thirds of the regions and said nothing"
     assert truncated[0]["examined"] == 10 and truncated[0]["total"] == 30
 
 
-def test_tier_zero_pairs_are_not_asked_and_nothing_else_is_skipped() -> None:
+def test_tier_zero_pairs_are_not_asked_and_nothing_else_is_skipped(tmp_path: Path) -> None:
     """Phase 1 of flow-aware-ranking. A pair with no sink of its family in reach cannot yield a flow, so its
     taint request is never issued -- and every other pair's request still is."""
     from openultrasast.model.regions import ScanRegion
@@ -466,7 +468,8 @@ def test_tier_zero_pairs_are_not_asked_and_nothing_else_is_skipped() -> None:
     region = ScanRegion(
         path="a.py", function="run", language="python", families=("injection", "path", "deserialization"), rank=0.9, source="entry_point"
     )
-    result = scan_repository(Path("."), [region], backend=_Backend(), budget=ScanBudget(max_model_calls=0))
+    (tmp_path / "a.py").write_bytes(b"print(1)\n")
+    result = scan_repository(tmp_path, [region], backend=_Backend(), budget=ScanBudget(max_model_calls=0))
 
     taint = next(i for i in issued if i["kind"] == "taint")
     assert taint["count"] == 1, "only the pair with a sink in reach is asked"
@@ -474,7 +477,7 @@ def test_tier_zero_pairs_are_not_asked_and_nothing_else_is_skipped() -> None:
     assert not any(d.get("reason") == "query_failed" for d in result.degradations)
 
 
-def test_the_budget_is_spent_by_evidence_when_asked() -> None:
+def test_the_budget_is_spent_by_evidence_when_asked(tmp_path: Path) -> None:
     """Phase 2 of flow-aware-ranking. With `order_by_evidence`, the evidence pass runs over every region and
     the region cap is cut from the (tier, score) order, so a low-ranked region with an open sink and a
     local source is examined before a top-ranked region with no sink in reach. Off, the static order holds."""
@@ -523,15 +526,17 @@ def test_the_budget_is_spent_by_evidence_when_asked() -> None:
     top = ScanRegion(path="top.py", function="handle", language="python", families=("injection",), rank=1.0, source="entry_point")
     low = ScanRegion(path="low.py", function="run", language="python", families=("injection",), rank=0.3, source="entry_point")
 
+    (tmp_path / "top.py").write_bytes(b"print(1)\n")
+    (tmp_path / "low.py").write_bytes(b"print(2)\n")
     static = _Backend()
-    scan_repository(Path("."), [top, low], backend=static, budget=ScanBudget(max_model_calls=0, max_regions=1))
+    scan_repository(tmp_path, [top, low], backend=static, budget=ScanBudget(max_model_calls=0, max_regions=1))
     assert static.evidence_sizes == [1] and static.taint_files == [], (
         "static order: the budget holds `top`, whose pair is tier 0 and pruned"
     )
 
     ordered = _Backend()
     result = scan_repository(
-        Path("."), [top, low], backend=ordered, budget=ScanBudget(max_model_calls=0, max_regions=1, order_by_evidence=True)
+        tmp_path, [top, low], backend=ordered, budget=ScanBudget(max_model_calls=0, max_regions=1, order_by_evidence=True)
     )
     assert ordered.evidence_sizes == [2], "the evidence pass covers every region, not only the budget"
     assert ordered.taint_files == [["low.py"]], "and the budget is spent on the region the evidence points at"

tests/test_model_scope.py

@@ -23,7 +23,7 @@ class Backend:
 
     def build(self, root, **kwargs):
         # Prove these tests provide real readable source to the driver boundary.
-        assert any(p.read_bytes() for p in root.iterdir() if p.is_file())
+        assert any(p.read_bytes() for p in root.rglob("*") if p.is_file())
         if self.fail:
             return None
         return SimpleNamespace(cpg_path=Path("test.cpg"), run_batch=self.batch, cleanup=lambda: None)
@@ -162,7 +162,7 @@ def test_later_timeout_retains_raw_answer_without_claiming_arbitration(tmp_path)
         def batch(self, kind, requests):
             rows = super().batch(kind, requests)
             if kind == "taint" and not any(p.get("evidenceOnly") for p in requests.values()):
-                time.sleep(0.08)
+                time.sleep(0.3)
                 return {rid: [{"kind": "retained", "value": 42}] for rid in requests}
             return rows
 
@@ -171,7 +171,7 @@ def test_later_timeout_retains_raw_answer_without_claiming_arbitration(tmp_path)
         [region("a.py", families=("injection", "access_control"))],
         backend=Slow(),
         ranking_mode="static",
-        execution_budget=ExecutionBudget(time.monotonic() + 0.05, 0.2),
+        execution_budget=ExecutionBudget(time.monotonic() + 0.2, 0.2),
     )
     outcomes = {q.identity.family: q for q in result.question_outcomes}
     assert outcomes["injection"].status == "not_arbitrated"