Merge branch 'main' into bazookamusic/avro-updated

Author: BazookaMusic

cpp/ql/lib/change-notes/2026-06-30-mad-qualified-field-names.md

@@ -0,0 +1,4 @@
+---
+category: deprecated
+---
+* Models-as-data flow summaries now use fully qualified field names (for example, `MyNamespace::MyStruct::myField`) instead of unqualified field names such as `myField`. We recommend updating existing flow summaries to use fully qualified field names. Unqualified field names are still supported, but that support will be removed in a future release.

cpp/ql/lib/change-notes/2026-06-30-variables-as-mad-sources-and-sinks.md

@@ -0,0 +1,4 @@
+---
+category: breaking
+---
+* Removed support for using variables as sources and sinks in models-as-data. Users of this feature should convert such sources and sinks to models defined using the QL language.

cpp/ql/lib/semmle/code/cpp/dataflow/ExternalFlow.qll

@@ -931,31 +931,6 @@ private Element interpretElement0(
     signature = "" and
     elementSpec(namespace, type, subtypes, name, signature, _)
   )
-  or
-  // Member variables
-  elementSpec(namespace, type, subtypes, name, signature, _) and
-  signature = "" and
-  exists(Class namedClass, Class classWithMember, MemberVariable member |
-    member.getName() = name and
-    member = classWithMember.getAMember() and
-    namedClass.hasQualifiedName(namespace, type) and
-    result = member
-  |
-    // field declared in the named type or a subtype of it (or an extension of any)
-    subtypes = true and
-    classWithMember = namedClass.getADerivedClass*()
-    or
-    // field declared directly in the named type (or an extension of it)
-    subtypes = false and
-    classWithMember = namedClass
-  )
-  or
-  // Global or namespace variables
-  elementSpec(namespace, type, subtypes, name, signature, _) and
-  signature = "" and
-  type = "" and
-  subtypes = false and
-  result = any(GlobalOrNamespaceVariable v | v.hasQualifiedName(namespace, name))
 }
 
 cached

cpp/ql/lib/semmle/code/cpp/dataflow/internal/FlowSummaryImpl.qll

@@ -6,6 +6,7 @@ private import cpp as Cpp
 private import codeql.dataflow.internal.FlowSummaryImpl
 private import codeql.dataflow.internal.AccessPathSyntax as AccessPath
 private import semmle.code.cpp.ir.dataflow.internal.DataFlowPrivate
+private import semmle.code.cpp.ir.dataflow.internal.DataFlowNodes
 private import semmle.code.cpp.ir.dataflow.internal.DataFlowUtil
 private import semmle.code.cpp.ir.dataflow.internal.DataFlowImplSpecific as DataFlowImplSpecific
 private import semmle.code.cpp.dataflow.ExternalFlow
@@ -20,8 +21,22 @@ module Input implements InputSig<Location, DataFlowImplSpecific::CppDataFlow> {
 
   class SinkBase = Void;
 
+  class FlowSummaryCallBase = CallInstruction;
+
   predicate callableFromSource(SummarizedCallableBase c) { exists(c.getBlock()) }
 
+  FlowSummaryCallBase getASourceCall(SummarizedCallableBase sc) {
+    result.getStaticCallTarget() = sc
+  }
+
+  DataFlowCallable getSummarizedCallableAsDataFlowCallable(SummarizedCallableBase c) {
+    result.asSummarizedCallable() = c
+  }
+
+  DataFlowCallable getSourceCallEnclosingCallable(FlowSummaryCallBase call) {
+    result.asSourceCallable() = call.getEnclosingFunction()
+  }
+
   ArgumentPosition callbackSelfParameterPosition() { result = TDirectPosition(-1) }
 
   ReturnKind getStandardReturnValueKind() { result = getReturnValueKind("") }
@@ -30,6 +45,10 @@ module Input implements InputSig<Location, DataFlowImplSpecific::CppDataFlow> {
     arg = repeatStars(result.(NormalReturnKind).getIndirectionIndex())
   }
 
+  ParameterPosition getFlowSummaryParameterPosition(ReturnKind rk) {
+    result = TFlowSummaryPosition(rk)
+  }
+
   string encodeParameterPosition(ParameterPosition pos) { result = pos.toString() }
 
   string encodeArgumentPosition(ArgumentPosition pos) { result = pos.toString() }
@@ -40,12 +59,24 @@ module Input implements InputSig<Location, DataFlowImplSpecific::CppDataFlow> {
     arg = repeatStars(rk.(NormalReturnKind).getIndirectionIndex())
   }
 
+  bindingset[namespace, type, base]
+  private string formatQualifiedName(string namespace, string type, string base) {
+    if namespace = ""
+    then result = type + "::" + base
+    else result = namespace + "::" + type + "::" + base
+  }
+
   string encodeContent(ContentSet cs, string arg) {
-    exists(FieldContent c |
+    exists(FieldContent c, string namespace, string type, string base |
       cs.isSingleton(c) and
       // FieldContent indices have 0 for the address, 1 for content, so we need to subtract one.
       result = "Field" and
-      arg = repeatStars(c.getIndirectionIndex() - 1) + c.getField().getName()
+      c.getField().hasQualifiedName(namespace, type, base)
+    |
+      arg = repeatStars(c.getIndirectionIndex() - 1) + formatQualifiedName(namespace, type, base)
+      or
+      // TODO: This disjunct can be removed once we stop supporting unqualified field names.
+      arg = repeatStars(c.getIndirectionIndex() - 1) + base
     )
     or
     exists(ElementContent ec |
@@ -102,10 +133,22 @@ module Input implements InputSig<Location, DataFlowImplSpecific::CppDataFlow> {
 private import Make<Location, DataFlowImplSpecific::CppDataFlow, Input> as Impl
 
 private module StepsInput implements Impl::Private::StepsInputSig {
+  Impl::Private::SummaryNode getSummaryNode(Node n) {
+    result = n.(FlowSummaryNode).getSummaryNode()
+  }
+
   DataFlowCall getACall(Public::SummarizedCallable sc) {
     result.getStaticCallTarget().getUnderlyingCallable() = sc
   }
 
+  Node getSourceOutNode(Input::FlowSummaryCallBase call, ReturnKind rk) {
+    exists(IndirectReturnOutNode out | result = out |
+      out.getCallInstruction() = call and
+      pragma[only_bind_out](rk.(NormalReturnKind).getIndirectionIndex()) =
+        pragma[only_bind_out](out.getIndirectionIndex())
+    )
+  }
+
   DataFlowCallable getSourceNodeEnclosingCallable(Input::SourceBase source) { none() }
 
   Node getSourceNode(Input::SourceBase source, Impl::Private::SummaryComponentStack s) { none() }
@@ -218,40 +261,11 @@ module SourceSinkInterpretationInput implements
 
   /** Provides additional sink specification logic. */
   bindingset[c]
-  predicate interpretOutput(string c, InterpretNode mid, InterpretNode node) {
-    // Allow variables to be picked as output nodes.
-    exists(Node n, Element ast |
-      n = node.asNode() and
-      ast = mid.asElement()
-    |
-      c = "" and
-      n.asExpr().(VariableAccess).getTarget() = ast
-    )
-  }
+  predicate interpretOutput(string c, InterpretNode mid, InterpretNode node) { none() }
 
   /** Provides additional source specification logic. */
   bindingset[c]
-  predicate interpretInput(string c, InterpretNode mid, InterpretNode node) {
-    exists(Node n, Element ast, VariableAccess e |
-      n = node.asNode() and
-      ast = mid.asElement() and
-      e.getTarget() = ast
-    |
-      // Allow variables to be picked as input nodes.
-      // We could simply do this as `e = n.asExpr()`, but that would not allow
-      // us to pick `x` as a sink in an example such as `x = source()` (but
-      // only subsequent uses of `x`) since the variable access on `x` doesn't
-      // actually load the value of `x`. So instead, we pick the instruction
-      // node corresponding to the generated `StoreInstruction` and use the
-      // expression associated with the destination instruction. This means
-      // that the `x` in `x = source()` can be marked as an input.
-      c = "" and
-      exists(StoreInstruction store |
-        store.getDestinationAddress().getUnconvertedResultExpression() = e and
-        n.asInstruction() = store
-      )
-    )
-  }
+  predicate interpretInput(string c, InterpretNode mid, InterpretNode node) { none() }
 }
 
 module Private {

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

@@ -1534,12 +1534,8 @@ class FlowSummaryNode extends Node, TFlowSummaryNode {
     result = this.getSummaryNode().getSummarizedCallable()
   }
 
-  /**
-   * Gets the enclosing callable. For a `FlowSummaryNode` this is always the
-   * summarized function this node is part of.
-   */
   override DataFlowCallable getEnclosingCallable() {
-    result.asSummarizedCallable() = this.getSummarizedCallable()
+    result = FlowSummaryImpl::Private::getEnclosingCallable(this.getSummaryNode())
   }
 
   override Location getLocationImpl() { result = this.getSummarizedCallable().getLocation() }

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

@@ -561,6 +561,21 @@ class SummaryArgumentNode extends ArgumentNode, FlowSummaryNode {
   }
 }
 
+/** An argument node that re-enters return output as input to a flow summary. */
+private class FlowSummaryArgumentNode extends ArgumentNode, FlowSummaryNode {
+  private CallInstruction callInstruction;
+  private ReturnKind rk;
+
+  FlowSummaryArgumentNode() {
+    this.getSummaryNode() = FlowSummaryImpl::Private::summaryArgumentNode(callInstruction, rk)
+  }
+
+  override predicate argumentOf(DataFlowCall call, ArgumentPosition pos) {
+    call.asCallInstruction() = callInstruction and
+    pos = TFlowSummaryPosition(rk)
+  }
+}
+
 /** A parameter position represented by an integer. */
 class ParameterPosition = Position;
 
@@ -616,6 +631,18 @@ class IndirectionPosition extends Position, TIndirectionPosition {
   final override int getIndirectionIndex() { result = indirectionIndex }
 }
 
+class FlowSummaryPosition extends Position, TFlowSummaryPosition {
+  ReturnKind rk;
+
+  FlowSummaryPosition() { this = TFlowSummaryPosition(rk) }
+
+  override string toString() { result = "write to: " + rk.toString() }
+
+  override int getArgumentIndex() { none() }
+
+  final override int getIndirectionIndex() { result = rk.getIndirectionIndex() }
+}
+
 newtype TPosition =
   TDirectPosition(int argumentIndex) {
     exists(any(CallInstruction c).getArgument(argumentIndex))
@@ -634,7 +661,8 @@ newtype TPosition =
       p = f.getParameter(argumentIndex) and
       indirectionIndex = [1 .. Ssa::getMaxIndirectionsForType(p.getUnspecifiedType()) - 1]
     )
-  }
+  } or
+  TFlowSummaryPosition(ReturnKind rk) { FlowSummaryImpl::Private::relevantFlowSummaryPosition(rk) }
 
 private newtype TReturnKind =
   TNormalReturnKind(int indirectionIndex) {
@@ -1378,6 +1406,8 @@ predicate nodeIsHidden(Node n) {
   n instanceof InitialGlobalValue
   or
   n instanceof SsaSynthNode
+  or
+  n.(FlowSummaryNode).getSummaryNode().isHidden()
 }
 
 predicate neverSkipInPathGraph(Node n) {

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

@@ -158,7 +158,7 @@ private module Cached {
     model = ""
     or
     // models-as-data summarized flow
-    FlowSummaryImpl::Private::Steps::summaryLocalStep(nodeFrom.(FlowSummaryNode).getSummaryNode(),
+    FlowSummaryImpl::Private::Steps::summaryLocalStep(nodeFrom,
       nodeTo.(FlowSummaryNode).getSummaryNode(), true, model)
   }
 

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

@@ -67,7 +67,7 @@ private module Cached {
     model = ""
     or
     // models-as-data summarized flow
-    FlowSummaryImpl::Private::Steps::summaryLocalStep(nodeFrom.(FlowSummaryNode).getSummaryNode(),
+    FlowSummaryImpl::Private::Steps::summaryLocalStep(nodeFrom,
       nodeTo.(FlowSummaryNode).getSummaryNode(), false, model)
     or
     // object->field conflation for content that is a `TaintInheritingContent`.