cgraphx 2.0.2 → 2.0.6

package/dist/core/code/engine/core/ingestion/cfg/emit.js

@@ -0,0 +1,646 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.bindingKey = exports.hasEmitSafeFacts = exports.isEmitSafeCfg = exports.basicBlockId = exports.DEFAULT_PDG_MAX_REACHING_DEF_BLOCK_REVISITS = exports.DEFAULT_PDG_MAX_REACHING_DEF_FACTS_PER_FUNCTION = exports.REACHING_DEF_FACTS_PER_EDGE_CAP = exports.POST_DOMINATE_DEBUG_ENV = exports.DEFAULT_PDG_MAX_CDG_MATERIALIZATION_PER_FUNCTION = exports.DEFAULT_PDG_MAX_CDG_EDGES_PER_FUNCTION = exports.DEFAULT_PDG_MAX_REACHING_DEF_EDGES_PER_FUNCTION = exports.DEFAULT_MAX_CFG_EDGES_PER_FUNCTION = exports.CALLEE_ID_SEP = exports.CALLEES_TRUNCATED_SENTINEL = void 0;
+exports.calleesOfBlock = calleesOfBlock;
+exports.calleeIdsOfBlock = calleeIdsOfBlock;
+exports.emitFileCfgs = emitFileCfgs;
+exports.emitFileReachingDefs = emitFileReachingDefs;
+exports.emitFileCdg = emitFileCdg;
+const utils_js_1 = require("../../../lib/utils.js");
+const reaching_defs_js_1 = require("./reaching-defs.js");
+const control_dependence_js_1 = require("./control-dependence.js");
+const post_dominators_js_1 = require("./post-dominators.js");
+const synthetic_escape_js_1 = require("./synthetic-escape.js");
+const call_site_harvest_js_1 = require("./visitors/call-site-harvest.js");
+const callee_id_sink_js_1 = require("../scope-resolution/graph-bridge/callee-id-sink.js");
+const reaching_def_reason_codec_js_1 = require("./reaching-def-reason-codec.js");
+const callee_cell_format_js_1 = require("./callee-cell-format.js");
+/**
+ * Cell-format constants live in the LEAF module `callee-cell-format.ts` and are
+ * re-exported here so every existing importer keeps working. The consumer side
+ * (`mcp/local/pdg-impact.ts`) imports them from the leaf directly: importing any
+ * binding from THIS module evaluates it, and with it the whole analyze-only CFG
+ * closure — 8 modules on every MCP server start to read two strings (#2802
+ * review). Producer and consumer still resolve to one definition, so the drift
+ * these shared constants exist to prevent stays impossible.
+ */
+var callee_cell_format_js_2 = require("./callee-cell-format.js");
+Object.defineProperty(exports, "CALLEES_TRUNCATED_SENTINEL", { enumerable: true, get: function () { return callee_cell_format_js_2.CALLEES_TRUNCATED_SENTINEL; } });
+Object.defineProperty(exports, "CALLEE_ID_SEP", { enumerable: true, get: function () { return callee_cell_format_js_2.CALLEE_ID_SEP; } });
+/**
+ * Default per-function CFG edge cap. A pathological generated function could
+ * otherwise emit an unbounded edge set; the cap bounds graph growth and is
+ * overridable via `--pdg` options. `0` (in options) means no cap (unlimited
+ * — see the `cap` mapping in {@link emitFileCfgs}); `undefined` means this
+ * default.
+ */
+exports.DEFAULT_MAX_CFG_EDGES_PER_FUNCTION = 5000;
+/**
+ * Default per-function REACHING_DEF edge cap (#2082 M2 KTD9). 4000 mirrors
+ * Joern's per-method `maxNumberOfDefinitions` — the closest production prior
+ * art — but truncates-and-warns instead of silently skipping the function.
+ * Counts (defBlock, useBlock, binding) DEDUPED edges, not statement-level
+ * facts. `0` ⇒ unlimited; `undefined` ⇒ this default.
+ */
+exports.DEFAULT_PDG_MAX_REACHING_DEF_EDGES_PER_FUNCTION = 4000;
+/**
+ * Default per-function CDG edge cap (#2085 M5). CDG edge count is bounded by
+ * (blocks × control-nesting-depth) — comparable to the CFG edge count — so it
+ * reuses the CFG default of 5000. Counts DEDUPED (controller, dependent, label)
+ * edges (the pure {@link computeControlDependence} already dedups). `0` ⇒
+ * unlimited; `undefined` ⇒ this default. Folded into the `RepoMeta.pdg` stamp
+ * (U5) so introducing CDG forces a full writeback for pre-CDG `--pdg` indexes.
+ */
+exports.DEFAULT_PDG_MAX_CDG_EDGES_PER_FUNCTION = 5000;
+/**
+ * Heap-safety ceiling on {@link computeControlDependence}'s pre-dedup
+ * materialization (#2188 review). The walk is O(edges × post-dom depth), and its
+ * `out` IS the deduped-edge quantity the per-function cap trims — so, UNLIKE
+ * REACHING_DEF's facts ceiling, this is deliberately NOT derived from the
+ * runtime edge cap (doing so would pre-truncate the very set the cap reports on,
+ * losing the exact dropped count). A fixed, generous multiple of the default
+ * edge cap: far above any real function — a catastrophe backstop only. When hit,
+ * the per-function cap reporting plus the `truncated` flag keep it observable
+ * (never a silent drop).
+ */
+exports.DEFAULT_PDG_MAX_CDG_MATERIALIZATION_PER_FUNCTION = 8 * exports.DEFAULT_PDG_MAX_CDG_EDGES_PER_FUNCTION;
+/**
+ * Env flag that additionally emits diagnostic `POST_DOMINATE` edges
+ * (block → its immediate post-dominator) alongside CDG (#2085 M5 KTD8). Off in
+ * every normal `--pdg` run — these are for inspecting the post-dom tree, not a
+ * queryable product surface. Accepts `1`/`true` (case-insensitive).
+ */
+exports.POST_DOMINATE_DEBUG_ENV = 'CGRAPH_PDG_EMIT_POST_DOMINATE';
+/**
+ * Fact-materialization headroom over the edge cap (#2082 M2 U3/F3): facts are
+ * O(defs×uses) BY SPEC in merge-heavy code, and the edge cap alone bounds the
+ * GRAPH, not the per-function memory spike of materializing facts before
+ * dedup. {@link emitFileReachingDefs} hands `edgeCap × this` to
+ * `computeReachingDefs` as `maxFacts` (unlimited when the edge cap is 0) —
+ * single source of truth; the DEFAULT constant below is derived, never the
+ * mechanism.
+ */
+exports.REACHING_DEF_FACTS_PER_EDGE_CAP = 4;
+/** Derived emit-path fact limit at the default edge cap (bench/doc anchor). */
+exports.DEFAULT_PDG_MAX_REACHING_DEF_FACTS_PER_FUNCTION = exports.REACHING_DEF_FACTS_PER_EDGE_CAP * exports.DEFAULT_PDG_MAX_REACHING_DEF_EDGES_PER_FUNCTION;
+/**
+ * Fixpoint-iteration budget for {@link computeReachingDefs}, as a multiple of
+ * the function's block count ({@link emitFileReachingDefs} passes
+ * `blocks.length × this` as `maxBlockVisits`). Iterative reaching-defs on a
+ * reducible CFG converges in O(loop-nesting-depth) passes, so a worklist
+ * re-visits each block a small multiple of times for real code; this budget
+ * tolerates a nesting depth far beyond any hand-written function (real code is
+ * ≤ ~15 deep) while truncating the pathological deep nest that otherwise drives
+ * the solver to O(blocks²) — measured at seconds + GB on a machine-generated
+ * 2000-line all-loops function whose fact count stays linear (so `maxFacts`
+ * never fires). Truncation degrades to a sound empty REACHING_DEF for that one
+ * function (status `truncated`), never wrong facts.
+ *
+ * As of #2201 this ceiling is an adversarial-only backstop that effectively
+ * never fires on real code: the production solver auto-selects the SSA-sparse
+ * path for the looping functions that would breach it, and the SSA path has no
+ * fixpoint iteration (it answers reaching queries from the def-use graph in one
+ * pass) so it computes the full facts where the dense worklist would have
+ * truncated. The budget is still consulted on the dense fallback path (small /
+ * loop-free functions, and throw-edge / unreachable-block functions the SSA path
+ * does not model). WTO / loop-aware iteration ordering was benchmarked and
+ * rejected (0% faster — the cost was dense-set propagation, not visitation
+ * order); SSA-sparse was the real fix. See reaching-defs.ts.
+ */
+exports.DEFAULT_PDG_MAX_REACHING_DEF_BLOCK_REVISITS = 64;
+/**
+ * The single BasicBlock id template (module doc). Exported for the M3 taint
+ * emit path (taint/emit.ts), whose TAINTED/SANITIZES edges must address the
+ * SAME persisted block nodes — a re-derived copy of this template would
+ * silently dangle the moment either drifted.
+ */
+const basicBlockId = (filePath, functionStartLine, functionStartColumn, blockIndex) => `BasicBlock:${filePath}:${functionStartLine}:${functionStartColumn}:${blockIndex}`;
+exports.basicBlockId = basicBlockId;
+/**
+ * Whether an untrusted `cfgSideChannel` element is safe to feed to
+ * {@link emitFileCfgs}. Deliberately NOT full FunctionCfg validation — it
+ * checks exactly the fields whose corruption is SILENT given emit's
+ * mechanics: {@link basicBlockId} string-templates every id-anchor value
+ * (filePath, function start line/column, block index, edge endpoints) and
+ * the graph's addNode/addRelationship are no-throw Map inserts. Unchecked,
+ * a missing anchor field cross-wires same-`undefined`-id blocks across
+ * functions (addNode is first-writer-wins), and an edge endpoint that
+ * matches no block index becomes a dangling `BasicBlock:…:<n>` edge that
+ * detonates much later at DB bulk-load instead of throwing here — so
+ * endpoints are checked for MEMBERSHIP in the block-index set, not just
+ * integer-ness. Lives in this module so the guard evolves with the id
+ * templating it defends (#2099 F4; M2 fields that join the id path must
+ * join this check).
+ */
+const isEmitSafeCfg = (cfg) => {
+    if (typeof cfg?.filePath !== 'string' ||
+        !Number.isInteger(cfg.functionStartLine) ||
+        !Number.isInteger(cfg.functionStartColumn) ||
+        !Array.isArray(cfg.blocks) ||
+        !Array.isArray(cfg.edges)) {
+        return false;
+    }
+    // Contiguity (index === position), not just integer-ness: every consumer —
+    // this module's id templating AND the reaching-defs solver's
+    // position-indexed adjacency arrays — assumes blocks[i].index === i. A
+    // membership-only check would admit a compacted channel ({index:0},{index:5})
+    // whose edge 0→5 passes membership but indexes past the arrays downstream.
+    for (let i = 0; i < cfg.blocks.length; i++) {
+        if (cfg.blocks[i]?.index !== i)
+            return false;
+    }
+    const n = cfg.blocks.length;
+    // entry/exit must land on real blocks — the solver feeds entryIndex straight
+    // into its RPO walk, where an out-of-range index throws and (worse than this
+    // one element) costs the whole FILE's REACHING_DEF pass (tri-review P3).
+    if (!Number.isInteger(cfg.entryIndex) ||
+        cfg.entryIndex < 0 ||
+        cfg.entryIndex >= n ||
+        !Number.isInteger(cfg.exitIndex) ||
+        cfg.exitIndex < 0 ||
+        cfg.exitIndex >= n) {
+        return false;
+    }
+    return cfg.edges.every((e) => Number.isInteger(e?.from) &&
+        Number.isInteger(e?.to) &&
+        e.from >= 0 &&
+        e.from < n &&
+        e.to >= 0 &&
+        e.to < n);
+};
+exports.isEmitSafeCfg = isEmitSafeCfg;
+/**
+ * Whether a structurally-valid CFG's M2 statement facts are safe to feed to
+ * the reaching-defs solver + REACHING_DEF id templating (#2082 U1/U4): the
+ * binding table's name/declLine/declColumn template into edge ids, and
+ * statement def/use indices must stay IN RANGE of the table (an escaping
+ * index would fabricate `undefined`-keyed ids). Deliberately SEPARATE from
+ * {@link isEmitSafeCfg}: malformed facts must cost only the function's
+ * REACHING_DEF projection — degrading to M1 behavior (CFG emitted, no facts)
+ * — never the BasicBlock/CFG layer itself.
+ */
+const hasEmitSafeFacts = (cfg) => {
+    const bindings = cfg.bindings;
+    if (bindings === undefined) {
+        // Pre-M2 channel — statements must be absent too.
+        return cfg.blocks.every((b) => b.statements === undefined);
+    }
+    if (!Array.isArray(bindings))
+        return false;
+    for (const b of bindings) {
+        if (typeof b?.name !== 'string' ||
+            !Number.isInteger(b.declLine) ||
+            !Number.isInteger(b.declColumn)) {
+            return false;
+        }
+    }
+    const bindingCount = bindings.length;
+    const inRange = (i) => Number.isInteger(i) && i >= 0 && i < bindingCount;
+    for (const b of cfg.blocks) {
+        const stmts = b.statements;
+        if (stmts === undefined)
+            continue;
+        if (!Array.isArray(stmts))
+            return false;
+        for (const s of stmts) {
+            if (!Number.isInteger(s?.line) || !Array.isArray(s.defs) || !Array.isArray(s.uses)) {
+                return false;
+            }
+            if (!s.defs.every(inRange) || !s.uses.every(inRange))
+                return false;
+            if (s.mayDefs !== undefined) {
+                if (!Array.isArray(s.mayDefs) || !s.mayDefs.every(inRange))
+                    return false;
+            }
+        }
+    }
+    return true;
+};
+exports.hasEmitSafeFacts = hasEmitSafeFacts;
+/**
+ * Emit BasicBlock nodes + CFG edges for every function CFG in `cfgs`.
+ *
+ * `maxEdgesPerFunction` caps edges per function. On overflow we stop emitting
+ * that function's remaining edges and call `onWarn` naming the dropped count —
+ * no silent truncation (KTD6/R6). Block nodes are always fully emitted (their
+ * count is bounded by the function's statement count); only edges are capped.
+ */
+/**
+ * Space-joined, sorted, de-duplicated leaf callee names invoked directly in a
+ * block (`call`/`new` sites; the leaf of a dotted path — `child_process.exec` ⇒
+ * `exec`). This is the persisted substrate for statement-precise inter-procedural
+ * impact: a callee reached from a function is "proven" to be impacted by a
+ * changed statement iff its name appears in the callees of a block in that
+ * statement's dependence slice. `sites` is harvested only for TS/JS under `--pdg`
+ * (and absent on synthetic ENTRY/EXIT), so the field is empty elsewhere and the
+ * bridge degrades to the prior (callgraph-equal) behavior. Space-joined because
+ * leaf names are identifiers (no spaces) and the field is itself one CSV cell.
+ */
+function calleesOfBlock(block) {
+    const names = new Set();
+    for (const stmt of block.statements ?? []) {
+        // A statement whose recorded sites reached the per-statement cap may have
+        // dropped over-cap callees (the harvester stops at the cap). Flag the block
+        // callee-unknown so the impact bridge keeps it callgraph-equal rather than
+        // under-proving an absent-but-real callee.
+        if ((stmt.sites?.length ?? 0) >= call_site_harvest_js_1.DEFAULT_PDG_MAX_SITES_PER_STATEMENT) {
+            names.add(callee_cell_format_js_1.CALLEES_TRUNCATED_SENTINEL);
+        }
+        for (const site of stmt.sites ?? []) {
+            if (site.kind === 'member-read')
+                continue;
+            const callee = site.callee;
+            if (!callee)
+                continue;
+            const leaf = callee.slice(callee.lastIndexOf('.') + 1);
+            if (leaf)
+                names.add(leaf);
+        }
+    }
+    return [...names].sort().join(' ');
+}
+/**
+ * Tab-joined ({@link CALLEE_ID_SEP}), sorted, de-duplicated RESOLVED callee symbol ids invoked
+ * directly in a block — the SOUND parallel to {@link calleesOfBlock}'s leaf
+ * names (#2227 follow-up plan U3, KTD1/KTD2/KTD7). Each block site's call-site
+ * anchor `at` (U1) is joined by EXACT position to the per-file resolved-id map
+ * `fileMap` (U2's `(line,col) → Set<calleeId>`), so a callee reached from a
+ * function is proven impacted by a changed statement iff its resolved id — not
+ * just its leaf NAME — appears in a slice block's `calleeIds`. This eliminates
+ * the same-leaf-name collision (false-proven) and import-alias (false-unproven)
+ * the name predicate suffers on overloading languages.
+ *
+ * The site partitioning is inherited verbatim from {@link calleesOfBlock}: the
+ * SAME `member-read`-skip and the SAME per-statement site cap (R7) — a capped
+ * statement adds {@link CALLEES_TRUNCATED_SENTINEL} so the bridge keeps the
+ * block callee-unknown for ids too (callgraph-equal rather than under-proving).
+ * Because `at` is the SAME anchor the CALLS resolution keyed `atRange` on
+ * (KTD7), the join lands on exactly the sites the name harvest partitioned,
+ * including the nested-function exclusion (so a single-line inline closure's
+ * inner call never leaks its id into the outer block).
+ *
+ * `fileMap` is the resolved-id map for THIS file (`calleeIdAccumulator.get(
+ * filePath)` in run.ts). Absent (pdg off, or a file with no captured CALLS) ⇒
+ * `''` — the bridge then degrades to the leaf-name fallback (R3). A site whose
+ * `at` is absent (pre-U1 channel) or whose position is not in the map
+ * contributes no id (graceful, never throws).
+ */
+function calleeIdsOfBlock(block, fileMap) {
+    if (fileMap === undefined)
+        return '';
+    const ids = new Set();
+    for (const stmt of block.statements ?? []) {
+        // Mirror calleesOfBlock's cap signal: an over-cap statement dropped sites,
+        // so the id list is INCOMPLETE — flag the block callee-unknown (R7).
+        if ((stmt.sites?.length ?? 0) >= call_site_harvest_js_1.DEFAULT_PDG_MAX_SITES_PER_STATEMENT) {
+            ids.add(callee_cell_format_js_1.CALLEES_TRUNCATED_SENTINEL);
+        }
+        for (const site of stmt.sites ?? []) {
+            if (site.kind === 'member-read')
+                continue;
+            const at = site.at;
+            if (!at)
+                continue;
+            const resolved = fileMap.get((0, callee_id_sink_js_1.calleeIdPosKey)(at[0], at[1]));
+            if (resolved === undefined)
+                continue;
+            for (const id of resolved)
+                ids.add(id);
+        }
+    }
+    return [...ids].sort().join(callee_cell_format_js_1.CALLEE_ID_SEP);
+}
+function emitFileCfgs(graph, cfgs, maxEdgesPerFunction = exports.DEFAULT_MAX_CFG_EDGES_PER_FUNCTION, onWarn, calleeIdMap) {
+    const result = { blocks: 0, edges: 0, droppedEdges: 0, cappedFunctions: 0 };
+    const cap = maxEdgesPerFunction > 0 ? maxEdgesPerFunction : Infinity;
+    for (const cfg of cfgs) {
+        const { filePath, functionStartLine, functionStartColumn } = cfg;
+        for (const b of cfg.blocks) {
+            graph.addNode({
+                id: (0, exports.basicBlockId)(filePath, functionStartLine, functionStartColumn, b.index),
+                label: 'BasicBlock',
+                properties: {
+                    name: '', // BasicBlock has no name column; identified by id + span
+                    filePath,
+                    startLine: b.startLine,
+                    endLine: b.endLine,
+                    text: b.text,
+                    // Space-joined leaf callee names invoked in this block — the
+                    // statement-precise inter-procedural reach substrate. Harvested from
+                    // the per-statement `sites` (already on the side channel); dropping
+                    // them here is what made the impact-mode bridge labeling degenerate.
+                    callees: calleesOfBlock(b),
+                    // Space-joined RESOLVED callee symbol ids — the SOUND parallel to
+                    // `callees`, joined from the U2 map by each site's exact `at`
+                    // position (#2227 follow-up U3). Absent map (pdg off / no captures)
+                    // ⇒ `''`, and the bridge falls back to the leaf-name match (R3).
+                    calleeIds: calleeIdsOfBlock(b, calleeIdMap),
+                },
+            });
+            result.blocks++;
+        }
+        let emittedForFn = 0;
+        for (const e of cfg.edges) {
+            if (emittedForFn >= cap) {
+                const dropped = cfg.edges.length - emittedForFn;
+                result.droppedEdges += dropped;
+                result.cappedFunctions++;
+                onWarn?.(`[cfg] ${filePath}:${functionStartLine}: per-function CFG edge cap ` +
+                    `(${maxEdgesPerFunction}) reached — dropped ${dropped} of ${cfg.edges.length} edges`);
+                break;
+            }
+            const sourceId = (0, exports.basicBlockId)(filePath, functionStartLine, functionStartColumn, e.from);
+            const targetId = (0, exports.basicBlockId)(filePath, functionStartLine, functionStartColumn, e.to);
+            graph.addRelationship({
+                id: (0, utils_js_1.generateId)('CFG', `${sourceId}->${targetId}:${e.kind}`),
+                type: 'CFG',
+                sourceId,
+                targetId,
+                confidence: 1.0,
+                reason: e.kind, // CfgEdgeKind (seq/cond-true/loop-back/…) — queryable
+            });
+            result.edges++;
+            emittedForFn++;
+        }
+    }
+    return result;
+}
+/**
+ * Stable identity for a binding inside edge ids (#2082 M2 KTD3/KTD9):
+ * `name:declLine:declCol` for declared bindings, `name@module` for synthetic
+ * ones. Distinct same-name bindings never share a key; identifier characters
+ * cannot contain the id separators. Exported for the M3 taint emit path —
+ * TAINTED/SANITIZES ids key bindings with the same discipline.
+ */
+const bindingKey = (b) => b.synthetic ? `${b.name}@module` : `${b.name}:${b.declLine}:${b.declColumn}`;
+exports.bindingKey = bindingKey;
+/**
+ * Compute reaching definitions per function and persist the bounded
+ * REACHING_DEF projection (#2082 M2 U4).
+ *
+ * Facts are DEDUPED to (defBlock, useBlock, binding) before budgeting — the
+ * persisted columns (`from,to,type,confidence,reason,step`; relationship ids
+ * are in-memory-only, the CodeRelation table has no id column) cannot
+ * distinguish finer rows, so statement-indexed ids would only manufacture
+ * byte-identical duplicate rows that burn budget. Statement granularity lives
+ * in the in-memory {@link computeReachingDefs} result, which the M3 taint
+ * engine recomputes on demand — the budget here governs only this projection
+ * and can never drop a taint fact.
+ *
+ * R7 (no silent truncation) covers BOTH layers: the per-function edge cap AND
+ * the solver's fact-materialization limit (which can fire without the edge
+ * cap ever being reached, since dedup is many-to-one) each produce one
+ * unconditional `onWarn`. The edge-cap warn names the top bindings by fact
+ * count — overflow is almost always one variable, which is exactly the datum
+ * M3 tuning wants.
+ */
+function emitFileReachingDefs(graph, cfgs, maxEdgesPerFunction = exports.DEFAULT_PDG_MAX_REACHING_DEF_EDGES_PER_FUNCTION, onWarn, 
+// U12: a per-file memoized solver lets the RD-emit / harvest / taint passes
+// share the SAME per-function fixpoint (this caller is its own cache bucket —
+// it passes maxBlockVisits, the harvest/taint callers do not). Defaults to the
+// plain solver so existing callers are unaffected.
+solve = reaching_defs_js_1.computeReachingDefs) {
+    const result = {
+        edges: 0,
+        droppedEdges: 0,
+        cappedFunctions: 0,
+        truncatedFunctions: 0,
+        malformedFactFunctions: 0,
+        facts: 0,
+    };
+    const cap = maxEdgesPerFunction > 0 ? maxEdgesPerFunction : Infinity;
+    const maxFacts = Number.isFinite(cap) ? cap * exports.REACHING_DEF_FACTS_PER_EDGE_CAP : 0; // 0 ⇒ unlimited
+    for (const cfg of cfgs) {
+        // Graceful degradation: malformed M2 facts cost only this function's
+        // REACHING_DEF projection — its BasicBlock/CFG layer was already emitted.
+        if (!(0, exports.hasEmitSafeFacts)(cfg)) {
+            result.malformedFactFunctions++;
+            onWarn?.(`[reaching-defs] ${cfg.filePath}:${cfg.functionStartLine}: malformed ` +
+                `statement facts (bad binding table or out-of-range fact indices) — ` +
+                `REACHING_DEF skipped for this function; its CFG is unaffected`);
+            continue;
+        }
+        const r = solve(cfg, {
+            maxFacts,
+            maxBlockVisits: cfg.blocks.length * exports.DEFAULT_PDG_MAX_REACHING_DEF_BLOCK_REVISITS,
+        });
+        if (r.status === 'no-facts')
+            continue;
+        result.facts += r.facts.length;
+        const { filePath, functionStartLine, functionStartColumn } = cfg;
+        if (r.status === 'truncated') {
+            result.truncatedFunctions++;
+            onWarn?.(`[reaching-defs] ${filePath}:${functionStartLine}: fact materialization ` +
+                `limit (${maxFacts}) reached — facts beyond it were not computed; ` +
+                `the persisted REACHING_DEF projection for this function is sparse`);
+        }
+        else if (r.status === 'overflow') {
+            result.truncatedFunctions++;
+            onWarn?.(`[reaching-defs] ${filePath}:${functionStartLine}: a basic block exceeds ` +
+                `the def-key stride (≥2^21 coalesced statements — minified/generated ` +
+                `code) — REACHING_DEF skipped for this function (computing any facts ` +
+                `would risk wrong-block aliasing); its CFG is unaffected`);
+            continue;
+        }
+        // Dedup to (defBlock, useBlock, binding) — facts arrive sorted, so the
+        // deduped order (and therefore cap truncation) is deterministic. ONE edge per
+        // group (the edge COUNT is unchanged — substrate/bench safe), but the FU-B-2
+        // annotation AGGREGATES the FULL ordered list of (defLine, useLine) pairs for
+        // that group into the persisted `reason`. The first fact of a group (facts
+        // sort by def block, def stmt, use block, use stmt, binding) keeps the group's
+        // emit position; every subsequent fact of the SAME (block-pair, binding)
+        // appends its line pair to that group's list. Carrying the full list (not just
+        // the first pair) is what makes a SAME-BINDING reassignment chain recoverable:
+        // `acc = f(acc); acc = g(acc)` coalesces into one self-block whose
+        // `acc@N->acc@N+1` and `acc@N+1->acc@N+2` steps share the one group — a
+        // first-pair-only annotation could chain N->N+1 but never reach N+2. Dedup of
+        // exact-duplicate pairs within a group keeps the list compact (a `x = x + 1`
+        // self-fact never re-adds the same pair).
+        const groupIndex = new Map();
+        const deduped = [];
+        for (const f of r.facts) {
+            const key = `${f.def.blockIndex}:${f.use.blockIndex}:${f.bindingIdx}`;
+            const at = groupIndex.get(key);
+            const pair = { defLine: f.def.line, useLine: f.use.line };
+            if (at === undefined) {
+                groupIndex.set(key, deduped.length);
+                deduped.push({
+                    defBlock: f.def.blockIndex,
+                    useBlock: f.use.blockIndex,
+                    bindingIdx: f.bindingIdx,
+                    pairs: [pair],
+                });
+                continue;
+            }
+            const list = deduped[at].pairs;
+            // Skip an exact-duplicate (defLine, useLine) — a self-referential statement
+            // (`x = x + 1`) emits the same line pair more than once; the list only needs
+            // each distinct step once for the projection walk.
+            if (!list.some((p) => p.defLine === pair.defLine && p.useLine === pair.useLine)) {
+                list.push(pair);
+            }
+        }
+        let emittedForFn = 0;
+        for (const edge of deduped) {
+            if (emittedForFn >= cap) {
+                const dropped = deduped.length - emittedForFn;
+                result.droppedEdges += dropped;
+                result.cappedFunctions++;
+                // Tallied lazily — cap overflow is the rare path; the common uncapped
+                // case must not pay a per-fact counting pass.
+                const factsPerBinding = new Map();
+                for (const f of r.facts) {
+                    factsPerBinding.set(f.bindingIdx, (factsPerBinding.get(f.bindingIdx) ?? 0) + 1);
+                }
+                const top = [...factsPerBinding.entries()]
+                    .sort((a, b) => b[1] - a[1] || a[0] - b[0])
+                    .slice(0, 2)
+                    .map(([idx, count]) => `${r.bindings[idx]?.name ?? `#${idx}`}(${count} facts)`)
+                    .join(', ');
+                onWarn?.(`[reaching-defs] ${filePath}:${functionStartLine}: per-function ` +
+                    `REACHING_DEF edge cap (${maxEdgesPerFunction}) reached — dropped ` +
+                    `${dropped} of ${deduped.length} edges; top bindings: ${top}`);
+                break;
+            }
+            const binding = r.bindings[edge.bindingIdx];
+            const sourceId = (0, exports.basicBlockId)(filePath, functionStartLine, functionStartColumn, edge.defBlock);
+            const targetId = (0, exports.basicBlockId)(filePath, functionStartLine, functionStartColumn, edge.useBlock);
+            graph.addRelationship({
+                // Single function anchor — the two block ids share it, so templating
+                // it once halves the id size (ids are in-memory-only but ~4000 of
+                // them per capped function is real transient heap).
+                id: (0, utils_js_1.generateId)('REACHING_DEF', `${filePath}:${functionStartLine}:${functionStartColumn}:` +
+                    `${edge.defBlock}->${edge.useBlock}:${(0, exports.bindingKey)(binding)}`),
+                type: 'REACHING_DEF',
+                sourceId,
+                targetId,
+                confidence: 1.0,
+                // FU-B-2: the source-level binding name (M0/S1 verdict — name FIRST so
+                // `pdg_query` flows stays queryable) PLUS a compact versioned annotation
+                // carrying the FULL ordered list of def/use source LINE pairs for this
+                // (block-pair, binding) group. For a self-edge (defBlock === useBlock)
+                // this captures the intra-block def@L→use@L' chain — including a
+                // SAME-BINDING reassignment chain (`acc@24->acc@25->acc@26`) — that the
+                // block-granular projection lost; the statement projection (pdg-impact.ts)
+                // walks the list forward to fixpoint to recover the coalesced block's
+                // interior statements.
+                reason: (0, reaching_def_reason_codec_js_1.encodeReachingDefReasonPairs)(binding.name, edge.pairs),
+            });
+            result.edges++;
+            emittedForFn++;
+        }
+    }
+    return result;
+}
+/** Whether the POST_DOMINATE debug env flag is enabled (`1`/`true`). */
+const postDominateDebugEnabled = () => {
+    const v = process.env[exports.POST_DOMINATE_DEBUG_ENV];
+    return v === '1' || v?.toLowerCase() === 'true';
+};
+/**
+ * Compute control dependence per function and persist the bounded CDG
+ * projection (#2085 M5 U4). Mirrors {@link emitFileReachingDefs}: the pure
+ * {@link computeControlDependence} already dedups to (controller, dependent,
+ * label), so the per-function cap applies to deduped edges and overflow logs
+ * one unconditional `onWarn` naming the dropped count — no silent truncation
+ * (R6/R7). The branch label ('T'|'F') rides the `reason` column (KTD3),
+ * mirroring how CFG stores its edge kind.
+ *
+ * When {@link POST_DOMINATE_DEBUG_ENV} is set, also emits diagnostic
+ * `POST_DOMINATE` edges (block → its immediate post-dominator). These are NOT
+ * capped or counted against the CDG budget — they exist only for inspecting the
+ * post-dom tree and never appear in a normal run.
+ */
+function emitFileCdg(graph, cfgs, maxEdgesPerFunction = exports.DEFAULT_PDG_MAX_CDG_EDGES_PER_FUNCTION, onWarn) {
+    const result = {
+        edges: 0,
+        droppedEdges: 0,
+        cappedFunctions: 0,
+        postDominateEdges: 0,
+        skippedUnsoundFunctions: 0,
+    };
+    const cap = maxEdgesPerFunction > 0 ? maxEdgesPerFunction : Infinity;
+    const emitPostDom = postDominateDebugEnabled();
+    for (const cfg of cfgs) {
+        const { filePath, functionStartLine, functionStartColumn } = cfg;
+        // Synthetic-escape pass (#2197 U1): restore EXIT reverse-reachability for a
+        // genuine exit-unreachable CYCLE (an unconditional `goto`-cycle / infinite
+        // loop) so the post-dom / CDG pass runs instead of being withheld. A no-op
+        // (returns `cfg` unchanged) for terminating functions and properly-escaped
+        // loops — those stay byte-identical. The synthetic edges are ANALYSIS-ONLY:
+        // they live on the returned shallow clone, never on the persisted `cfg`, so
+        // CFG / REACHING_DEF and the byte-identical-off golden are unaffected. Both
+        // the gate below AND the post-dom / CDG passes must see the augmented view
+        // (KTD7 — the Ferrante walk re-reads `cfg.edges`).
+        const view = (0, synthetic_escape_js_1.augmentForPostDom)(cfg);
+        // Sound post-dominance requires EXIT reachable from every entry-reachable
+        // block (#2188 review). The synthetic-escape pass recovers genuine cycles;
+        // anything STILL unreachable after it is a residual non-cycle anomaly (a
+        // dangling/dead-end block, a branch-less trapping spin, or a construction
+        // error) — NOT something we bridge (that would mask the bug). Skip CDG for
+        // it and surface the skip. CFG and REACHING_DEF (emitted elsewhere,
+        // independent of post-dominance) are kept.
+        if (!(0, post_dominators_js_1.isExitReachableFromAllBlocks)(view)) {
+            result.skippedUnsoundFunctions++;
+            onWarn?.(`[cdg] ${filePath}:${functionStartLine}: EXIT not reachable from all ` +
+                `blocks — CDG skipped for this function (CFG/REACHING_DEF unaffected)`);
+            continue;
+        }
+        // Compute the post-dom tree once and feed it to the control-dependence
+        // pass (avoids recomputing it) and to the optional POST_DOMINATE emit. The
+        // CDG edges reference BLOCK INDICES, which are identical in `view` and `cfg`
+        // (the augmentation only appends edges), so persisting them keyed off the
+        // original block ids is correct.
+        const tree = (0, post_dominators_js_1.computePostDominators)(view);
+        // Bound the pre-dedup materialization (heap parity with REACHING_DEF). The
+        // fixed ceiling is a catastrophe backstop; the per-function edge cap below
+        // remains the reporting authority. A ceiling hit is surfaced, not silent.
+        const { edges: cdgEdges, truncated } = (0, control_dependence_js_1.computeControlDependence)(view, tree, exports.DEFAULT_PDG_MAX_CDG_MATERIALIZATION_PER_FUNCTION);
+        if (truncated) {
+            onWarn?.(`[cdg] ${filePath}:${functionStartLine}: control-dependence materialization ` +
+                `ceiling (${exports.DEFAULT_PDG_MAX_CDG_MATERIALIZATION_PER_FUNCTION}) reached — ` +
+                `edge counts for this function are a floor`);
+        }
+        let emittedForFn = 0;
+        for (const edge of cdgEdges) {
+            if (emittedForFn >= cap) {
+                const dropped = cdgEdges.length - emittedForFn;
+                result.droppedEdges += dropped;
+                result.cappedFunctions++;
+                onWarn?.(`[cdg] ${filePath}:${functionStartLine}: per-function CDG edge cap ` +
+                    `(${maxEdgesPerFunction}) reached — dropped ${dropped} of ${cdgEdges.length} edges`);
+                break;
+            }
+            const sourceId = (0, exports.basicBlockId)(filePath, functionStartLine, functionStartColumn, edge.controllerBlock);
+            const targetId = (0, exports.basicBlockId)(filePath, functionStartLine, functionStartColumn, edge.dependentBlock);
+            graph.addRelationship({
+                id: (0, utils_js_1.generateId)('CDG', `${filePath}:${functionStartLine}:${functionStartColumn}:` +
+                    `${edge.controllerBlock}->${edge.dependentBlock}:${edge.label}`),
+                type: 'CDG',
+                sourceId,
+                targetId,
+                confidence: 1.0,
+                reason: edge.label, // 'T' | 'F' — queryable, mirrors CFG's kind-in-reason
+            });
+            result.edges++;
+            emittedForFn++;
+        }
+        if (emitPostDom) {
+            for (let b = 0; b < tree.ipdom.length; b++) {
+                const ip = tree.ipdom[b];
+                if (ip === post_dominators_js_1.NO_IPDOM)
+                    continue;
+                graph.addRelationship({
+                    id: (0, utils_js_1.generateId)('POST_DOMINATE', `${filePath}:${functionStartLine}:${functionStartColumn}:${b}->${ip}`),
+                    type: 'POST_DOMINATE',
+                    sourceId: (0, exports.basicBlockId)(filePath, functionStartLine, functionStartColumn, b),
+                    targetId: (0, exports.basicBlockId)(filePath, functionStartLine, functionStartColumn, ip),
+                    confidence: 1.0,
+                    reason: '',
+                });
+                result.postDominateEdges++;
+            }
+        }
+    }
+    return result;
+}