Merge pull request #20862 from MathiasVP/union-content-field-content-common-base-class

C++: Create a common base class for 'FieldContent' and 'UnionContent'

Author: MathiasVP

cpp/ql/lib/change-notes/2025-11-19-content.md

@@ -0,0 +1,4 @@
+---
+category: minorAnalysis
+---
+* The class `DataFlow::FieldContent` now covers both `union` and `struct`/`class` types. A new predicate `FieldContent.getAField` has been added to access the union members associated with the `FieldContent`. The old `FieldContent` has been renamed to `NonUnionFieldContent`.

cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/DataFlowPrivate.qll

@@ -861,6 +861,10 @@ predicate jumpStep(Node n1, Node n2) {
     n2.(FlowSummaryNode).getSummaryNode())
 }
 
+bindingset[c]
+pragma[inline_late]
+private int getIndirectionIndexLate(Content c) { result = c.getIndirectionIndex() }
+
 /**
  * Holds if data can flow from `node1` to `node2` via an assignment to `f`.
  * Thus, `node2` references an object with a field `f` that contains the
@@ -873,23 +877,17 @@ predicate jumpStep(Node n1, Node n2) {
 predicate storeStepImpl(Node node1, Content c, Node node2, boolean certain) {
   exists(
     PostFieldUpdateNode postFieldUpdate, int indirectionIndex1, int numberOfLoads,
-    StoreInstruction store
+    StoreInstruction store, FieldContent fc
   |
     postFieldUpdate = node2 and
-    nodeHasInstruction(node1, store, pragma[only_bind_into](indirectionIndex1)) and
+    fc = c and
+    nodeHasInstruction(node1, pragma[only_bind_into](store),
+      pragma[only_bind_into](indirectionIndex1)) and
     postFieldUpdate.getIndirectionIndex() = 1 and
     numberOfLoadsFromOperand(postFieldUpdate.getFieldAddress(),
-      store.getDestinationAddressOperand(), numberOfLoads, certain)
-  |
-    exists(FieldContent fc | fc = c |
-      fc.getField() = postFieldUpdate.getUpdatedField() and
-      fc.getIndirectionIndex() = 1 + indirectionIndex1 + numberOfLoads
-    )
-    or
-    exists(UnionContent uc | uc = c |
-      uc.getAField() = postFieldUpdate.getUpdatedField() and
-      uc.getIndirectionIndex() = 1 + indirectionIndex1 + numberOfLoads
-    )
+      store.getDestinationAddressOperand(), numberOfLoads, certain) and
+    fc.getAField() = postFieldUpdate.getUpdatedField() and
+    getIndirectionIndexLate(fc) = 1 + indirectionIndex1 + numberOfLoads
   )
   or
   // models-as-data summarized flow
@@ -965,22 +963,17 @@ predicate nodeHasInstruction(Node node, Instruction instr, int indirectionIndex)
  * `node2`.
  */
 predicate readStep(Node node1, ContentSet c, Node node2) {
-  exists(FieldAddress fa1, Operand operand, int numberOfLoads, int indirectionIndex2 |
+  exists(
+    FieldAddress fa1, Operand operand, int numberOfLoads, int indirectionIndex2, FieldContent fc
+  |
+    fc = c and
     nodeHasOperand(node2, operand, indirectionIndex2) and
     // The `1` here matches the `node2.getIndirectionIndex() = 1` conjunct
     // in `storeStep`.
     nodeHasOperand(node1, fa1.getObjectAddressOperand(), 1) and
-    numberOfLoadsFromOperand(fa1, operand, numberOfLoads, _)
-  |
-    exists(FieldContent fc | fc = c |
-      fc.getField() = fa1.getField() and
-      fc.getIndirectionIndex() = indirectionIndex2 + numberOfLoads
-    )
-    or
-    exists(UnionContent uc | uc = c |
-      uc.getAField() = fa1.getField() and
-      uc.getIndirectionIndex() = indirectionIndex2 + numberOfLoads
-    )
+    numberOfLoadsFromOperand(fa1, operand, numberOfLoads, _) and
+    fc.getAField() = fa1.getField() and
+    getIndirectionIndexLate(fc) = indirectionIndex2 + numberOfLoads
   )
   or
   // models-as-data summarized flow
@@ -1574,7 +1567,7 @@ pragma[inline]
 ContentApprox getContentApprox(Content c) {
   exists(string prefix, Field f |
     prefix = result.(FieldApproxContent).getPrefix() and
-    f = c.(FieldContent).getField() and
+    f = c.(NonUnionFieldContent).getField() and
     fieldHasApproxName(f, prefix)
   )
   or

cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/DataFlowUtil.qll

@@ -2093,8 +2093,8 @@ private Field getAFieldWithSize(Union u, int bytes) {
 
 cached
 private newtype TContent =
-  TFieldContent(Field f, int indirectionIndex) {
-    // the indirection index for field content starts at 1 (because `TFieldContent` is thought of as
+  TNonUnionContent(Field f, int indirectionIndex) {
+    // the indirection index for field content starts at 1 (because `TNonUnionContent` is thought of as
     // the address of the field, `FieldAddress` in the IR).
     indirectionIndex = [1 .. SsaImpl::getMaxIndirectionsForType(f.getUnspecifiedType())] and
     // Reads and writes of union fields are tracked using `UnionContent`.
@@ -2124,14 +2124,14 @@ private newtype TContent =
  */
 class Content extends TContent {
   /** Gets a textual representation of this element. */
-  abstract string toString();
+  string toString() { none() } // overridden in subclasses
 
   predicate hasLocationInfo(string path, int sl, int sc, int el, int ec) {
     path = "" and sl = 0 and sc = 0 and el = 0 and ec = 0
   }
 
   /** Gets the indirection index of this `Content`. */
-  abstract int getIndirectionIndex();
+  int getIndirectionIndex() { none() } // overridden in subclasses
 
   /**
    * INTERNAL: Do not use.
@@ -2142,7 +2142,7 @@ class Content extends TContent {
    * For example, a write to a field `f` implies that any content of
    * the form `*f` is also cleared.
    */
-  abstract predicate impliesClearOf(Content c);
+  predicate impliesClearOf(Content c) { none() } // overridden in subclasses
 }
 
 /**
@@ -2162,22 +2162,42 @@ private module ContentStars {
 
 private import ContentStars
 
-/** A reference through a non-union instance field. */
+private class TFieldContent = TNonUnionContent or TUnionContent;
+
+/**
+ * A `Content` that references a `Field`. This may be a field of a `struct`,
+ * `class`, or `union`. In the case of a `union` there may be multiple fields
+ * associated with the same `Content`.
+ */
 class FieldContent extends Content, TFieldContent {
+  /** Gets a `Field` of this `Content`. */
+  Field getAField() { none() }
+
+  /**
+   * Gets the field associated with this `Content`, if a unique one exists.
+   */
+  final Field getField() { result = unique( | | this.getAField()) }
+
+  override int getIndirectionIndex() { none() } // overridden in subclasses
+
+  override string toString() { none() } // overridden in subclasses
+
+  override predicate impliesClearOf(Content c) { none() } // overridden in subclasses
+}
+
+/** A reference through a non-union instance field. */
+class NonUnionFieldContent extends FieldContent, TNonUnionContent {
   private Field f;
   private int indirectionIndex;
 
-  FieldContent() { this = TFieldContent(f, indirectionIndex) }
+  NonUnionFieldContent() { this = TNonUnionContent(f, indirectionIndex) }
 
   override string toString() { result = contentStars(this) + f.toString() }
 
-  Field getField() { result = f }
+  override Field getAField() { result = f }
 
   /** Gets the indirection index of this `FieldContent`. */
-  pragma[inline]
-  override int getIndirectionIndex() {
-    pragma[only_bind_into](result) = pragma[only_bind_out](indirectionIndex)
-  }
+  override int getIndirectionIndex() { result = indirectionIndex }
 
   override predicate impliesClearOf(Content c) {
     exists(FieldContent fc |
@@ -2191,7 +2211,7 @@ class FieldContent extends Content, TFieldContent {
 }
 
 /** A reference through an instance field of a union. */
-class UnionContent extends Content, TUnionContent {
+class UnionContent extends FieldContent, TUnionContent {
   private Union u;
   private int indirectionIndex;
   private int bytes;
@@ -2201,16 +2221,13 @@ class UnionContent extends Content, TUnionContent {
   override string toString() { result = contentStars(this) + u.toString() }
 
   /** Gets a field of the underlying union of this `UnionContent`, if any. */
-  Field getAField() { result = u.getAField() and getFieldSize(result) = bytes }
+  override Field getAField() { result = u.getAField() and getFieldSize(result) = bytes }
 
   /** Gets the underlying union of this `UnionContent`. */
   Union getUnion() { result = u }
 
   /** Gets the indirection index of this `UnionContent`. */
-  pragma[inline]
-  override int getIndirectionIndex() {
-    pragma[only_bind_into](result) = pragma[only_bind_out](indirectionIndex)
-  }
+  override int getIndirectionIndex() { result = indirectionIndex }
 
   override predicate impliesClearOf(Content c) {
     exists(UnionContent uc |
@@ -2234,10 +2251,7 @@ class ElementContent extends Content, TElementContent {
 
   ElementContent() { this = TElementContent(indirectionIndex) }
 
-  pragma[inline]
-  override int getIndirectionIndex() {
-    pragma[only_bind_into](result) = pragma[only_bind_out](indirectionIndex)
-  }
+  override int getIndirectionIndex() { result = indirectionIndex }
 
   override predicate impliesClearOf(Content c) { none() }
 

cpp/ql/src/utils/modelgenerator/internal/CaptureModels.qll

@@ -190,7 +190,7 @@ module ModelGeneratorCommonInput implements ModelGeneratorCommonInputSig<Cpp::Lo
   predicate isRelevantType(Type t) { any() }
 
   Type getUnderlyingContentType(DataFlow::ContentSet c) {
-    result = c.(DataFlow::FieldContent).getField().getUnspecifiedType() or
+    result = c.(DataFlow::NonUnionFieldContent).getField().getUnspecifiedType() or
     result = c.(DataFlow::UnionContent).getUnion().getUnspecifiedType()
   }
 
@@ -340,12 +340,7 @@ private module SummaryModelGeneratorInput implements SummaryModelGeneratorInputS
     )
   }
 
-  predicate isField(DataFlow::ContentSet cs) {
-    exists(DataFlow::Content c | cs.isSingleton(c) |
-      c instanceof DataFlow::FieldContent or
-      c instanceof DataFlow::UnionContent
-    )
-  }
+  predicate isField(DataFlow::ContentSet cs) { cs.isSingleton(any(DataFlow::FieldContent fc)) }
 
   predicate isCallback(DataFlow::ContentSet c) { none() }