feat(model): first repository measurement, and the four things it found

contributor-scan 2.2. VAmPI at a pinned commit, scanned whole rather than
as excerpts, with benchmarks/repos/measure.py recording build time, query
time, arbitration time, peak memory, cost, unjudged regions, and whether
each known vulnerability was found and at which rung.

The corpus could not have found any of this. Every defect below needs a
repository to become visible, and three of them were invisible behind a
passing test suite.

**The batch path was unreachable.** query_batch was implemented, tested and
committed while CpgResult exposed only `run`, so the driver's capability
check found nothing and fell back to one JVM per region-family: the first
attempt at this measurement ran eighteen minutes on 520 lines, issuing
`--param function=` one at a time. Every test called query_batch directly,
so none of them noticed. CpgResult now carries run_batch and a test asserts
the capability is reachable through the object the driver is handed.
The earlier 275s->106s improvement attributed to batching came from region
deduplication alone.

**The budget stopped free work.** It bounds model calls, but was enforced
by breaking the region loop -- spending a paid limit to stop unpaid CPG
arbitration. 6 of 22 regions were examined and VAmPI's BOLA, which the
graph entails for nothing, was never reached. It now withholds the judge
only. 22 of 22 regions arbitrated, 17 counted as regions_unasked (silent
because unaffordable is not silent because decided), and the BOLA is
entailed by the model layer.

**A region with no function asked about the whole repository.** Sending
function="" matched every hit in the tree and the driver attributed that
answer to whoever asked: one `host='0.0.0.0'` in app.py became eight
identical entailed findings in eight files that do not contain it. Queries
now carry the region's file. Entailed findings 14 -> 7, query time
70.9s -> 40.9s. The single-request form stays unscoped so the committed
pair measurements remain reproducible.

**The JVM had no heap bound.** Joern claimed a quarter of physical RAM and
runs were killed under memory pressure at different points, so the numbers
described the host rather than the engine. -J-Xmx2048m by default,
OPENULTRASAST_CPG_HEAP_MB to raise it.

Also: arbitrated findings carried `function:line` where the report layer
read `path:line`, so every entailed finding reached a contributor with a
function name where its file belonged; and the witness line was parsed by
splitting on a comma, which produced sites like
`config.py:17): permissive literal '0.0.0.0':?`.

Reading: GO on one CPG per repository -- 6.2s to build, 177s end to end,
869MB, $0.0067, nothing infeasible. NO-GO on region-scoped taint as the
only shape: VAmPI's SQL injection runs from a handler parameter into a
function in another module, and a per-function region sees neither end.
That is the ordinary shape of a real injection bug, so it is recorded as
task 2.4 along with three smaller defects this run exposed.

Gates byte-identical.

Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01T7XD7x5KonMdmGRkp63EQG

tests/test_cpg_backend.py

@@ -162,3 +162,97 @@ def test_a_batch_result_that_omits_a_request_returns_nothing_for_it() -> None:
     backend = JoernBackend(runner=lambda c, **k: _Partial())
     rows = backend.query_batch(Path("/tmp/c.bin"), "taint", {"r1": {}, "r2": {}})
     assert rows.get("r1") == [] and "r2" not in rows
+
+
+def test_a_built_cpg_offers_the_batch_path_to_its_driver(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
+    """The capability must be reachable through the object the driver is handed, not merely present.
+
+    ``query_batch`` was implemented, tested and committed while ``build`` returned a result carrying only
+    ``run``. The driver discovers batching by looking for ``run_batch`` on that result, found nothing, and
+    fell back to one JVM per region -- silently, because every test called ``query_batch`` directly. A
+    repository scan is what made it visible: ~14s per region-family, one ``--param function=`` invocation at
+    a time.
+    """
+    from openultrasast.cpg.backend import BEGIN, END, JoernBackend
+
+    scripts: list[list[str]] = []
+
+    def runner(command, **kwargs):  # type: ignore[no-untyped-def]
+        class _Done:
+            returncode = 0
+            stderr = ""
+            stdout = f'{BEGIN}\n{{"r1": [{{"sink": "os.system(x)"}}]}}\n{END}\n'
+
+        if "--output" in command:  # the parse step; make the artifact it promises
+            Path(command[command.index("--output") + 1]).write_text("cpg")
+        else:
+            scripts.append(list(command))
+        return _Done()
+
+    monkeypatch.setenv("OPENULTRASAST_JOERN_PROBE", "on")
+    result = JoernBackend(runner=runner).build(tmp_path)
+
+    assert result is not None
+    assert result.run_batch is not None, "a backend that can batch must expose it, or the driver cannot find it"
+    rows = result.run_batch("taint", {"r1": {"sources": ("request.args",), "sinks": ("os.system",)}})
+    assert rows["r1"], "the batch path must reach the engine, not just exist"
+    assert len(scripts) == 1, "one invocation for the batch"
+    assert not any(arg.startswith("function=") for arg in scripts[0]), "a batch carries requests, not one function"
+
+
+def test_the_engine_runs_under_a_bounded_heap(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
+    """A JVM with no -Xmx claims a quarter of physical RAM.
+
+    That is not a tuning preference. The first repository measurements were killed under memory pressure at
+    different points, so their numbers described the host rather than the engine, and on a contributor's
+    laptop an unbounded heap is the difference between a scan running in the background and a scan the
+    machine notices.
+    """
+    from openultrasast.cpg.backend import JoernBackend
+
+    commands: list[list[str]] = []
+
+    def runner(command, **kwargs):  # type: ignore[no-untyped-def]
+        class _Done:
+            returncode = 0
+            stderr = ""
+            stdout = ""
+
+        commands.append(list(command))
+        if "--output" in command:
+            Path(command[command.index("--output") + 1]).write_text("cpg")
+        return _Done()
+
+    monkeypatch.setenv("OPENULTRASAST_JOERN_PROBE", "on")
+    monkeypatch.delenv("OPENULTRASAST_CPG_HEAP_MB", raising=False)
+    result = JoernBackend(runner=runner).build(tmp_path)
+    assert result is not None
+    result.run("taint", {})
+
+    assert commands, "nothing ran"
+    for command in commands:
+        assert any(arg.startswith("-J-Xmx") for arg in command), f"unbounded heap in {command[0]}"
+
+
+def test_the_operator_can_raise_the_heap(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
+    """A bound that cannot be lifted is a bound that gets removed. A big repository may need a big heap."""
+    from openultrasast.cpg.backend import JoernBackend
+
+    commands: list[list[str]] = []
+
+    def runner(command, **kwargs):  # type: ignore[no-untyped-def]
+        class _Done:
+            returncode = 0
+            stderr = ""
+            stdout = ""
+
+        commands.append(list(command))
+        if "--output" in command:
+            Path(command[command.index("--output") + 1]).write_text("cpg")
+        return _Done()
+
+    monkeypatch.setenv("OPENULTRASAST_JOERN_PROBE", "on")
+    monkeypatch.setenv("OPENULTRASAST_CPG_HEAP_MB", "6144")
+    JoernBackend(runner=runner).build(tmp_path)
+
+    assert "-J-Xmx6144m" in commands[0]