Merge pull request #1183 from castler/fix-rule-7-0-2-compiler-generated

Fix RULE-7-0-2 false positives: UnknownType and reference-to-bool dereference

Author: MichaelRFairhurst

change_notes/2026-08-25-fix-fp-rule-7-0-2-unknown-and-reference-bool.md

@@ -0,0 +1,13 @@
+- `RULE-7-0-2` - `NoImplicitBoolConversion.ql`:
+   - Fixed false positives where a conversion's source expression type could
+     not be resolved to a concrete type (`UnknownType`), which occurs only in
+     template-dependent contexts that the extractor cannot resolve, e.g. a
+     `constexpr bool` variable template whose initializer is itself another
+     dependent variable template such as `std::conjunction_v<...>`.
+   - Fixed false positives on reference-dereference conversions (`bool&`/
+     `bool&&` to `bool`), which occur e.g. via the compiler-synthesized
+     `std::get<N>(...)` call used to implement structured binding
+     decomposition (`auto [a, b] = some_pair_or_tuple_expr;` where `b` is
+     `bool`). Dereferencing a reference to `bool` does not change the type or
+     representation of the value, so this is not a conversion to `bool` in
+     the sense intended by the rule.

cpp/misra/src/rules/RULE-7-0-2/NoImplicitBoolConversion.ql

@@ -56,6 +56,30 @@ where
       e = conv and
       conv.getType().getUnspecifiedType() instanceof BoolType and
       not conv.getExpr().getType().getUnspecifiedType() instanceof BoolType and
+      // Exclude conversions whose source expression type could not be resolved to a concrete
+      // type (`UnknownType`). This shape occurs only in template-dependent contexts that the
+      // extractor has not (and, for uninstantiated templates, cannot) resolve to an actual type
+      // - e.g. a `constexpr bool` variable template / `noexcept(...)` specifier built from
+      // another variable template such as `std::conjunction_v<...>`. There is no actual,
+      // concrete conversion to `bool` that a developer wrote or that the compiler ultimately
+      // performs; flagging it produces a false positive because the "conversion from 'unknown'"
+      // is an artifact of incomplete extraction of template-dependent/library code, not a
+      // genuine violation.
+      not conv.getExpr().getType() instanceof UnknownType and
+      // Exclude reference-dereference conversions (`T& -> T` / `T&& -> T`) whose referenced type
+      // is itself `bool` once its reference-ness is stripped, e.g. the compiler-synthesized
+      // `std::get<N>(...)` call used to implement structured binding decomposition
+      // (`auto [a, b] = some_pair_or_tuple_expr;` where `b` is `bool`). `getUnspecifiedType()`
+      // resolves typedefs/specifiers but does not strip reference qualification, so a `bool&`/
+      // `bool&&` is not itself recognised as already being `bool` by the check above. But
+      // dereferencing a reference to `bool` does not represent a conversion from a different
+      // type to `bool` - the referenced value is already a `bool` - so this is not a violation.
+      not conv.getExpr()
+          .getType()
+          .getUnspecifiedType()
+          .(ReferenceType)
+          .getBaseType()
+          .getUnspecifiedType() instanceof BoolType and
       // Exception 2: Contextual conversion from pointer
       not (
         isPointerType(conv.getExpr().getType()) and

cpp/misra/test/rules/RULE-7-0-2/test.cpp

@@ -213,4 +213,59 @@ void test_member_function_pointer_conversion() {
   bool l3 = l1;              // NON_COMPLIANT
   bool l4 = l2;              // NON_COMPLIANT
   bool l5 = (l1 != nullptr); // COMPLIANT
+}
+
+// Regression test for a false positive where the compiler-synthesized
+// `get<N>(...)` call used to implement structured binding decomposition (`auto
+// [a, b] = ...;`) was incorrectly flagged as a "conversion to bool", even
+// though the decomposed member is already `bool` - dereferencing a
+// `bool&`/`bool&&` reference is not a conversion from another type to `bool`.
+struct BoolPair {
+  std::int32_t first;
+  bool second;
+};
+
+template <std::size_t I> auto get(const BoolPair &p) {
+  if constexpr (I == 0) {
+    return p.first;
+  } else {
+    return p.second;
+  }
+}
+
+namespace std {
+template <class T> struct tuple_size;
+template <std::size_t I, class T> struct tuple_element;
+
+template <> struct tuple_size<BoolPair> {
+  static constexpr std::size_t value = 2;
+};
+template <> struct tuple_element<0, BoolPair> { using type = std::int32_t; };
+template <> struct tuple_element<1, BoolPair> { using type = bool; };
+} // namespace std
+
+BoolPair make_bool_pair();
+
+void test_structured_binding_bool_decomposition() {
+  auto [l1, l2] =
+      make_bool_pair(); // COMPLIANT - structured binding decomposition, not a
+                        // real conversion to bool
+  if (l2) {             // COMPLIANT
+  }
+  bool l3 = l2; // COMPLIANT - l2 is already bool
+}
+
+// Regression test for a false positive where a `constexpr bool` variable
+// template, whose initializer is itself a dependent expression built from
+// another variable template, was incorrectly flagged as a "conversion from
+// 'unknown' to bool". The extractor cannot resolve a concrete type for a
+// dependent, uninstantiated template expression, so it should not be treated as
+// an actual conversion.
+template <typename T> constexpr bool is_something_v = true;
+
+template <typename T>
+constexpr bool derived_from_template_v = is_something_v<T>; // COMPLIANT
+
+bool test_dependent_variable_template_instantiation() {
+  return derived_from_template_v<std::int32_t>; // COMPLIANT
 }