circle-ir 3.180.0 → 3.182.0

package/dist/core/circle-ir-core.cjs

@@ -10066,6 +10066,10 @@ function processGoBlock(node, defs, uses, scopeStack, counters) {
     } else if (child.type === "for_statement") {
       const rangeClause = findChildByTypeGo(child, "range_clause");
       if (rangeClause) {
+        const right = rangeClause.childForFieldName("right");
+        if (right) {
+          extractGoUses(right, uses, scopeStack);
+        }
         const left = rangeClause.childForFieldName("left");
         if (left) {
           extractGoLhsDefs(left, defs, scopeStack, child.startPosition.row + 1);
@@ -10073,6 +10077,7 @@ function processGoBlock(node, defs, uses, scopeStack, counters) {
       }
     } else if (child.type === "call_expression") {
       extractGoUses(child, uses, scopeStack);
+      recordGoOpaqueCodecDestDef(child, defs, scopeStack);
     } else if (child.type === "return_statement") {
       for (let i2 = 0; i2 < child.childCount; i2++) {
         const expr = child.child(i2);
@@ -10165,6 +10170,63 @@ function findChildByTypeGo(node, type) {
   }
   return null;
 }
+var GO_OPAQUE_CODEC_METHODS = /* @__PURE__ */ new Set([
+  "Unmarshal",
+  // json/xml/yaml/toml/gob
+  "Decode",
+  // json.NewDecoder(r).Decode(&dest), gob.Decoder.Decode
+  "NewDecoder",
+  // wrapper; handled via chained Decode above
+  "UnmarshalYAML",
+  "UnmarshalJSON",
+  "UnmarshalText",
+  "UnmarshalBinary"
+]);
+function recordGoOpaqueCodecDestDef(call, defs, scopeStack) {
+  const fn = call.childForFieldName("function");
+  if (!fn || fn.type !== "selector_expression") return;
+  const fieldNode = fn.childForFieldName("field");
+  if (!fieldNode) return;
+  const method = getNodeText(fieldNode);
+  if (!GO_OPAQUE_CODEC_METHODS.has(method)) return;
+  const argsNode = call.childForFieldName("arguments");
+  if (!argsNode) return;
+  const args2 = [];
+  for (let i2 = 0; i2 < argsNode.childCount; i2++) {
+    const c = argsNode.child(i2);
+    if (!c) continue;
+    if (c.type === "," || c.type === "(" || c.type === ")") continue;
+    args2.push(c);
+  }
+  if (args2.length === 0) return;
+  const destArg = args2.length >= 2 ? args2[1] : args2[0];
+  const destName = extractGoAddressableVarName(destArg);
+  if (!destName || destName === "_") return;
+  const line = call.startPosition.row + 1;
+  const def = {
+    id: defs.length + 1,
+    variable: destName,
+    kind: "local",
+    line
+  };
+  defs.push(def);
+  currentScope(scopeStack).set(destName, def.id);
+}
+function extractGoAddressableVarName(node) {
+  if (node.type === "unary_expression") {
+    const operand = node.childForFieldName("operand") ?? node.child(1) ?? null;
+    if (operand) return extractGoAddressableVarName(operand);
+    return null;
+  }
+  if (node.type === "identifier") {
+    return getNodeText(node);
+  }
+  if (node.type === "selector_expression") {
+    const field = node.childForFieldName("field");
+    if (field) return getNodeText(field);
+  }
+  return null;
+}
 
 // src/analysis/config-loader.ts
 function parseConfig(content) {
@@ -10677,6 +10739,35 @@ var DEFAULT_SOURCES = [
   { method: "get", class: "Jedis", type: "db_input", severity: "medium", return_tainted: true, languages: ["java"] },
   { method: "hget", class: "Jedis", type: "db_input", severity: "medium", return_tainted: true, languages: ["java"] },
   { method: "mget", class: "Jedis", type: "db_input", severity: "medium", return_tainted: true, languages: ["java"] },
+  // --- Serverless transport channels (cognium-dev #213 first slice) ---
+  //
+  // AWS Lambda / API Gateway invocation-event properties. The Lambda
+  // handler signature `(event, context) => …` (JS/TS) or
+  // `def handler(event, context)` (Python) receives an `event` object
+  // whose properties carry the untrusted HTTP-request-shaped payload:
+  //
+  //   event.body                              — request body (string)
+  //   event.queryStringParameters             — `?a=b` params
+  //   event.multiValueQueryStringParameters   — `?a=1&a=2` params
+  //   event.pathParameters                    — path template captures
+  //   event.headers                           — request headers
+  //   event.multiValueHeaders                 — repeated headers
+  //   event.requestContext                    — API Gateway request context
+  //
+  // Vercel Serverless Functions and Cloudflare Workers both use `req`
+  // / `request` receivers that are already covered by the Express-
+  // style patterns above (line ~398). This block specifically covers
+  // the `event`-shaped API Gateway convention that those don't reach.
+  { property: "body", object: "event", type: "http_body", severity: "high", property_tainted: true, languages: ["javascript", "typescript"] },
+  { property: "queryStringParameters", object: "event", type: "http_query", severity: "high", property_tainted: true, languages: ["javascript", "typescript"] },
+  { property: "multiValueQueryStringParameters", object: "event", type: "http_query", severity: "high", property_tainted: true, languages: ["javascript", "typescript"] },
+  { property: "pathParameters", object: "event", type: "http_path", severity: "high", property_tainted: true, languages: ["javascript", "typescript"] },
+  { property: "headers", object: "event", type: "http_header", severity: "high", property_tainted: true, languages: ["javascript", "typescript"] },
+  { property: "multiValueHeaders", object: "event", type: "http_header", severity: "high", property_tainted: true, languages: ["javascript", "typescript"] },
+  { property: "body", object: "event", type: "http_body", severity: "high", property_tainted: true, languages: ["python"] },
+  { property: "queryStringParameters", object: "event", type: "http_query", severity: "high", property_tainted: true, languages: ["python"] },
+  { property: "pathParameters", object: "event", type: "http_path", severity: "high", property_tainted: true, languages: ["python"] },
+  { property: "headers", object: "event", type: "http_header", severity: "high", property_tainted: true, languages: ["python"] },
   // --- JWT claims (unverified decode — PyJWT / jose / jsonwebtoken / auth0 java-jwt / golang-jwt) ---
   // A JWT's payload is *always* attacker-authored. Even after verification
   // the *contents* of the claims (username, role, custom fields) are not
@@ -14958,15 +15049,21 @@ function propagateTaint(graphOrDfg, callsOrSources, sourcesOrSinks, sinksOrSanit
     const callsAtSink = callsByLine.get(sink.line) ?? [];
     for (const call of callsAtSink) {
       for (const arg of call.arguments) {
-        if (arg.variable) {
-          if (sink.argPositions && sink.argPositions.length > 0) {
-            if (!sink.argPositions.includes(arg.position)) {
-              continue;
-            }
+        if (sink.argPositions && sink.argPositions.length > 0) {
+          if (!sink.argPositions.includes(arg.position)) {
+            continue;
           }
-          for (const use of usesAtSink) {
-            if (use.variable === arg.variable && use.def_id !== null) {
+        }
+        const candidateUses = arg.variable ? usesAtSink.filter((u) => u.variable === arg.variable) : usesAtSink;
+        {
+          for (const use of candidateUses) {
+            if (use.def_id !== null) {
               if (allTaintedDefIds.has(use.def_id)) {
+                if (!arg.variable) {
+                  if (typeof arg.expression !== "string" || arg.expression.length === 0) continue;
+                  const re = new RegExp(`(?:^|[^A-Za-z0-9_$])${use.variable.replace(/[.*+?^${}()|[\]\\]/g, "\\$&")}(?:[^A-Za-z0-9_$]|$)`);
+                  if (!re.test(arg.expression)) continue;
+                }
                 const taintInfo = taintByDefId.get(use.def_id);
                 if (taintInfo) {
                   const isSanitized = checkSanitized(
@@ -15004,7 +15101,15 @@ function propagateTaint(graphOrDfg, callsOrSources, sourcesOrSinks, sinksOrSanit
       }
     }
   }
-  return { taintedVars, flows, reachableSinks };
+  const seen = /* @__PURE__ */ new Set();
+  const deduped = [];
+  for (const f of flows) {
+    const key = `${f.source.line}|${f.sink.line}|${f.sink.type}|${f.path.map((p) => p.variable).join(">")}`;
+    if (seen.has(key)) continue;
+    seen.add(key);
+    deduped.push(f);
+  }
+  return { taintedVars, flows: deduped, reachableSinks };
 }
 function findInitialTaint(sources, callsByLine, defsByLine) {
   const tainted = [];

package/dist/core/circle-ir-core.js

@@ -10000,6 +10000,10 @@ function processGoBlock(node, defs, uses, scopeStack, counters) {
     } else if (child.type === "for_statement") {
       const rangeClause = findChildByTypeGo(child, "range_clause");
       if (rangeClause) {
+        const right = rangeClause.childForFieldName("right");
+        if (right) {
+          extractGoUses(right, uses, scopeStack);
+        }
         const left = rangeClause.childForFieldName("left");
         if (left) {
           extractGoLhsDefs(left, defs, scopeStack, child.startPosition.row + 1);
@@ -10007,6 +10011,7 @@ function processGoBlock(node, defs, uses, scopeStack, counters) {
       }
     } else if (child.type === "call_expression") {
       extractGoUses(child, uses, scopeStack);
+      recordGoOpaqueCodecDestDef(child, defs, scopeStack);
     } else if (child.type === "return_statement") {
       for (let i2 = 0; i2 < child.childCount; i2++) {
         const expr = child.child(i2);
@@ -10099,6 +10104,63 @@ function findChildByTypeGo(node, type) {
   }
   return null;
 }
+var GO_OPAQUE_CODEC_METHODS = /* @__PURE__ */ new Set([
+  "Unmarshal",
+  // json/xml/yaml/toml/gob
+  "Decode",
+  // json.NewDecoder(r).Decode(&dest), gob.Decoder.Decode
+  "NewDecoder",
+  // wrapper; handled via chained Decode above
+  "UnmarshalYAML",
+  "UnmarshalJSON",
+  "UnmarshalText",
+  "UnmarshalBinary"
+]);
+function recordGoOpaqueCodecDestDef(call, defs, scopeStack) {
+  const fn = call.childForFieldName("function");
+  if (!fn || fn.type !== "selector_expression") return;
+  const fieldNode = fn.childForFieldName("field");
+  if (!fieldNode) return;
+  const method = getNodeText(fieldNode);
+  if (!GO_OPAQUE_CODEC_METHODS.has(method)) return;
+  const argsNode = call.childForFieldName("arguments");
+  if (!argsNode) return;
+  const args2 = [];
+  for (let i2 = 0; i2 < argsNode.childCount; i2++) {
+    const c = argsNode.child(i2);
+    if (!c) continue;
+    if (c.type === "," || c.type === "(" || c.type === ")") continue;
+    args2.push(c);
+  }
+  if (args2.length === 0) return;
+  const destArg = args2.length >= 2 ? args2[1] : args2[0];
+  const destName = extractGoAddressableVarName(destArg);
+  if (!destName || destName === "_") return;
+  const line = call.startPosition.row + 1;
+  const def = {
+    id: defs.length + 1,
+    variable: destName,
+    kind: "local",
+    line
+  };
+  defs.push(def);
+  currentScope(scopeStack).set(destName, def.id);
+}
+function extractGoAddressableVarName(node) {
+  if (node.type === "unary_expression") {
+    const operand = node.childForFieldName("operand") ?? node.child(1) ?? null;
+    if (operand) return extractGoAddressableVarName(operand);
+    return null;
+  }
+  if (node.type === "identifier") {
+    return getNodeText(node);
+  }
+  if (node.type === "selector_expression") {
+    const field = node.childForFieldName("field");
+    if (field) return getNodeText(field);
+  }
+  return null;
+}
 
 // src/analysis/config-loader.ts
 function parseConfig(content) {
@@ -10611,6 +10673,35 @@ var DEFAULT_SOURCES = [
   { method: "get", class: "Jedis", type: "db_input", severity: "medium", return_tainted: true, languages: ["java"] },
   { method: "hget", class: "Jedis", type: "db_input", severity: "medium", return_tainted: true, languages: ["java"] },
   { method: "mget", class: "Jedis", type: "db_input", severity: "medium", return_tainted: true, languages: ["java"] },
+  // --- Serverless transport channels (cognium-dev #213 first slice) ---
+  //
+  // AWS Lambda / API Gateway invocation-event properties. The Lambda
+  // handler signature `(event, context) => …` (JS/TS) or
+  // `def handler(event, context)` (Python) receives an `event` object
+  // whose properties carry the untrusted HTTP-request-shaped payload:
+  //
+  //   event.body                              — request body (string)
+  //   event.queryStringParameters             — `?a=b` params
+  //   event.multiValueQueryStringParameters   — `?a=1&a=2` params
+  //   event.pathParameters                    — path template captures
+  //   event.headers                           — request headers
+  //   event.multiValueHeaders                 — repeated headers
+  //   event.requestContext                    — API Gateway request context
+  //
+  // Vercel Serverless Functions and Cloudflare Workers both use `req`
+  // / `request` receivers that are already covered by the Express-
+  // style patterns above (line ~398). This block specifically covers
+  // the `event`-shaped API Gateway convention that those don't reach.
+  { property: "body", object: "event", type: "http_body", severity: "high", property_tainted: true, languages: ["javascript", "typescript"] },
+  { property: "queryStringParameters", object: "event", type: "http_query", severity: "high", property_tainted: true, languages: ["javascript", "typescript"] },
+  { property: "multiValueQueryStringParameters", object: "event", type: "http_query", severity: "high", property_tainted: true, languages: ["javascript", "typescript"] },
+  { property: "pathParameters", object: "event", type: "http_path", severity: "high", property_tainted: true, languages: ["javascript", "typescript"] },
+  { property: "headers", object: "event", type: "http_header", severity: "high", property_tainted: true, languages: ["javascript", "typescript"] },
+  { property: "multiValueHeaders", object: "event", type: "http_header", severity: "high", property_tainted: true, languages: ["javascript", "typescript"] },
+  { property: "body", object: "event", type: "http_body", severity: "high", property_tainted: true, languages: ["python"] },
+  { property: "queryStringParameters", object: "event", type: "http_query", severity: "high", property_tainted: true, languages: ["python"] },
+  { property: "pathParameters", object: "event", type: "http_path", severity: "high", property_tainted: true, languages: ["python"] },
+  { property: "headers", object: "event", type: "http_header", severity: "high", property_tainted: true, languages: ["python"] },
   // --- JWT claims (unverified decode — PyJWT / jose / jsonwebtoken / auth0 java-jwt / golang-jwt) ---
   // A JWT's payload is *always* attacker-authored. Even after verification
   // the *contents* of the claims (username, role, custom fields) are not
@@ -14892,15 +14983,21 @@ function propagateTaint(graphOrDfg, callsOrSources, sourcesOrSinks, sinksOrSanit
     const callsAtSink = callsByLine.get(sink.line) ?? [];
     for (const call of callsAtSink) {
       for (const arg of call.arguments) {
-        if (arg.variable) {
-          if (sink.argPositions && sink.argPositions.length > 0) {
-            if (!sink.argPositions.includes(arg.position)) {
-              continue;
-            }
+        if (sink.argPositions && sink.argPositions.length > 0) {
+          if (!sink.argPositions.includes(arg.position)) {
+            continue;
           }
-          for (const use of usesAtSink) {
-            if (use.variable === arg.variable && use.def_id !== null) {
+        }
+        const candidateUses = arg.variable ? usesAtSink.filter((u) => u.variable === arg.variable) : usesAtSink;
+        {
+          for (const use of candidateUses) {
+            if (use.def_id !== null) {
               if (allTaintedDefIds.has(use.def_id)) {
+                if (!arg.variable) {
+                  if (typeof arg.expression !== "string" || arg.expression.length === 0) continue;
+                  const re = new RegExp(`(?:^|[^A-Za-z0-9_$])${use.variable.replace(/[.*+?^${}()|[\]\\]/g, "\\$&")}(?:[^A-Za-z0-9_$]|$)`);
+                  if (!re.test(arg.expression)) continue;
+                }
                 const taintInfo = taintByDefId.get(use.def_id);
                 if (taintInfo) {
                   const isSanitized = checkSanitized(
@@ -14938,7 +15035,15 @@ function propagateTaint(graphOrDfg, callsOrSources, sourcesOrSinks, sinksOrSanit
       }
     }
   }
-  return { taintedVars, flows, reachableSinks };
+  const seen = /* @__PURE__ */ new Set();
+  const deduped = [];
+  for (const f of flows) {
+    const key = `${f.source.line}|${f.sink.line}|${f.sink.type}|${f.path.map((p) => p.variable).join(">")}`;
+    if (seen.has(key)) continue;
+    seen.add(key);
+    deduped.push(f);
+  }
+  return { taintedVars, flows: deduped, reachableSinks };
 }
 function findInitialTaint(sources, callsByLine, defsByLine) {
   const tainted = [];

package/dist/core/extractors/dfg.js

@@ -1385,6 +1385,14 @@ function processGoBlock(node, defs, uses, scopeStack, counters) {
             // range clause: for k, v := range expr
             const rangeClause = findChildByTypeGo(child, 'range_clause');
             if (rangeClause) {
+                const right = rangeClause.childForFieldName('right');
+                // Record uses of the range source on the same line as the loop-var
+                // defs, so `computeChains` links the source def → loop-var def, which
+                // is what carries taint into the loop body. Emit before defs so RHS
+                // uses see the outer binding, not the new loop var. (Issue #243.)
+                if (right) {
+                    extractGoUses(right, uses, scopeStack);
+                }
                 const left = rangeClause.childForFieldName('left');
                 if (left) {
                     extractGoLhsDefs(left, defs, scopeStack, child.startPosition.row + 1);
@@ -1394,6 +1402,14 @@ function processGoBlock(node, defs, uses, scopeStack, counters) {
         else if (child.type === 'call_expression') {
             // Extract uses from call arguments
             extractGoUses(child, uses, scopeStack);
+            // Opaque-codec destination re-def (cognium-dev #243).
+            //
+            // Calls like `json.Unmarshal(bytes, &dest)`, `xml.Unmarshal(...)`,
+            // `gob.NewDecoder(r).Decode(&dest)`, `yaml.Unmarshal(...)` populate
+            // `dest` via reflection. Model that as a re-definition of `dest` on
+            // this line so `computeChains` links the source-arg use to a fresh
+            // `dest` def and taint propagates through the codec.
+            recordGoOpaqueCodecDestDef(child, defs, scopeStack);
         }
         else if (child.type === 'return_statement') {
             // Extract uses from return expressions
@@ -1514,4 +1530,87 @@ function findChildByTypeGo(node, type) {
     }
     return null;
 }
+/**
+ * cognium-dev #243 — opaque codec destination re-defs for Go.
+ *
+ * Package/method pairs whose second (or only) arg is a destination that the
+ * call populates via reflection. Modelling them as re-defs of the dest lets
+ * `computeChains` link source-arg → dest and taint propagates through the
+ * codec.
+ */
+const GO_OPAQUE_CODEC_METHODS = new Set([
+    'Unmarshal', // json/xml/yaml/toml/gob
+    'Decode', // json.NewDecoder(r).Decode(&dest), gob.Decoder.Decode
+    'NewDecoder', // wrapper; handled via chained Decode above
+    'UnmarshalYAML',
+    'UnmarshalJSON',
+    'UnmarshalText',
+    'UnmarshalBinary',
+]);
+function recordGoOpaqueCodecDestDef(call, defs, scopeStack) {
+    // Only handle direct selector calls: `pkg.Method(...)` or `receiver.Decode(...)`.
+    const fn = call.childForFieldName('function');
+    if (!fn || fn.type !== 'selector_expression')
+        return;
+    const fieldNode = fn.childForFieldName('field');
+    if (!fieldNode)
+        return;
+    const method = getNodeText(fieldNode);
+    if (!GO_OPAQUE_CODEC_METHODS.has(method))
+        return;
+    const argsNode = call.childForFieldName('arguments');
+    if (!argsNode)
+        return;
+    // Positional args (skip commas / parens).
+    const args = [];
+    for (let i = 0; i < argsNode.childCount; i++) {
+        const c = argsNode.child(i);
+        if (!c)
+            continue;
+        if (c.type === ',' || c.type === '(' || c.type === ')')
+            continue;
+        args.push(c);
+    }
+    if (args.length === 0)
+        return;
+    // Destination convention:
+    //   - Unmarshal(bytes, &dest) → args[1]
+    //   - Decode(&dest)           → args[0]
+    const destArg = args.length >= 2 ? args[1] : args[0];
+    const destName = extractGoAddressableVarName(destArg);
+    if (!destName || destName === '_')
+        return;
+    const line = call.startPosition.row + 1;
+    const def = {
+        id: defs.length + 1,
+        variable: destName,
+        kind: 'local',
+        line,
+    };
+    defs.push(def);
+    currentScope(scopeStack).set(destName, def.id);
+}
+/**
+ * Return the addressed variable name for an opaque-codec destination arg.
+ * Handles `&x`, bare `x`, and `&pkg.Y`-style receivers (returns `Y` — the
+ * codec still populates that concrete addressable location).
+ */
+function extractGoAddressableVarName(node) {
+    if (node.type === 'unary_expression') {
+        // &x — operand carries the identifier
+        const operand = node.childForFieldName('operand') ?? node.child(1) ?? null;
+        if (operand)
+            return extractGoAddressableVarName(operand);
+        return null;
+    }
+    if (node.type === 'identifier') {
+        return getNodeText(node);
+    }
+    if (node.type === 'selector_expression') {
+        const field = node.childForFieldName('field');
+        if (field)
+            return getNodeText(field);
+    }
+    return null;
+}
 //# sourceMappingURL=dfg.js.map