Merge pull request #517 from dave-bartolomeo/dave/IRFilter

C++: Don't generate IR for functions with bad ASTs

cpp/ql/src/semmle/code/cpp/PrintAST.qll

@@ -59,18 +59,36 @@ private predicate locationSortKeys(Locatable ast, string file, int line,
   )
 }
 
+private Function getEnclosingFunction(Locatable ast) {
+  result = ast.(Expr).getEnclosingFunction() or
+  result = ast.(Stmt).getEnclosingFunction() or
+  result = ast.(Initializer).getExpr().getEnclosingFunction() or
+  result = ast.(Parameter).getFunction() or
+  exists(DeclStmt stmt |
+    stmt.getADeclarationEntry() = ast and
+    result = stmt.getEnclosingFunction()
+  ) or
+  result = ast
+}
+
 /**
  * Most nodes are just a wrapper around `Locatable`, but we do synthesize new
  * nodes for things like parameter lists and constructor init lists.
  */
 private newtype TPrintASTNode =
-  TASTNode(Locatable ast) or
-  TParametersNode(Function func) or
+  TASTNode(Locatable ast) {
+    shouldPrintFunction(getEnclosingFunction(ast))
+  } or
+  TParametersNode(Function func) {
+    shouldPrintFunction(func)
+  } or
   TConstructorInitializersNode(Constructor ctor) {
-    ctor.hasEntryPoint()
+    ctor.hasEntryPoint() and
+    shouldPrintFunction(ctor)
   } or
   TDestructorDestructionsNode(Destructor dtor) {
-    dtor.hasEntryPoint()
+    dtor.hasEntryPoint() and
+    shouldPrintFunction(dtor)
   }
 
 /**

cpp/ql/src/semmle/code/cpp/ir/IRConfiguration.qll

@@ -0,0 +1,19 @@
+import cpp
+
+private newtype TIRConfiguration = MkIRConfiguration()
+
+/**
+ * The query can extend this class to control which functions have IR generated for them.
+ */
+class IRConfiguration extends TIRConfiguration {
+  string toString() {
+    result = "IRConfiguration"
+  }
+
+  /**
+   * Holds if IR should be created for function `func`. By default, holds for all functions.
+   */
+  predicate shouldCreateIRForFunction(Function func) {
+    any()
+  }
+}

cpp/ql/src/semmle/code/cpp/ir/implementation/aliased_ssa/PrintIR.qll

@@ -1,5 +1,40 @@
 private import IR
 import cpp
+import semmle.code.cpp.ir.IRConfiguration
+
+private newtype TPrintIRConfiguration = MkPrintIRConfiguration()
+
+/**
+ * The query can extend this class to control which functions are printed.
+ */
+class PrintIRConfiguration extends TPrintIRConfiguration {
+  string toString() {
+    result = "PrintIRConfiguration"
+  }
+
+  /**
+   * Holds if the IR for `func` should be printed. By default, holds for all
+   * functions.
+   */
+  predicate shouldPrintFunction(Function func) {
+    any()
+  }
+}
+
+private predicate shouldPrintFunction(Function func) {
+  exists(PrintIRConfiguration config |
+    config.shouldPrintFunction(func)
+  )
+}
+
+/**
+ * Override of `IRConfiguration` to only create IR for the functions that are to be dumped.
+ */
+private class FilteredIRConfiguration extends IRConfiguration {
+  override predicate shouldCreateIRForFunction(Function func) {
+    shouldPrintFunction(func)
+  }
+}
 
 private string getAdditionalInstructionProperty(Instruction instr, string key) {
   exists(IRPropertyProvider provider |
@@ -14,9 +49,15 @@ private string getAdditionalBlockProperty(IRBlock block, string key) {
 }
 
 private newtype TPrintableIRNode =
-  TPrintableFunctionIR(FunctionIR funcIR) or
-  TPrintableIRBlock(IRBlock block) or
-  TPrintableInstruction(Instruction instr)
+  TPrintableFunctionIR(FunctionIR funcIR) {
+    shouldPrintFunction(funcIR.getFunction())
+  } or
+  TPrintableIRBlock(IRBlock block) {
+    shouldPrintFunction(block.getFunction())
+  } or
+  TPrintableInstruction(Instruction instr) {
+    shouldPrintFunction(instr.getFunction())
+  }
 
 /**
  * A node to be emitted in the IR graph.

cpp/ql/src/semmle/code/cpp/ir/implementation/raw/PrintIR.qll

@@ -1,5 +1,40 @@
 private import IR
 import cpp
+import semmle.code.cpp.ir.IRConfiguration
+
+private newtype TPrintIRConfiguration = MkPrintIRConfiguration()
+
+/**
+ * The query can extend this class to control which functions are printed.
+ */
+class PrintIRConfiguration extends TPrintIRConfiguration {
+  string toString() {
+    result = "PrintIRConfiguration"
+  }
+
+  /**
+   * Holds if the IR for `func` should be printed. By default, holds for all
+   * functions.
+   */
+  predicate shouldPrintFunction(Function func) {
+    any()
+  }
+}
+
+private predicate shouldPrintFunction(Function func) {
+  exists(PrintIRConfiguration config |
+    config.shouldPrintFunction(func)
+  )
+}
+
+/**
+ * Override of `IRConfiguration` to only create IR for the functions that are to be dumped.
+ */
+private class FilteredIRConfiguration extends IRConfiguration {
+  override predicate shouldCreateIRForFunction(Function func) {
+    shouldPrintFunction(func)
+  }
+}
 
 private string getAdditionalInstructionProperty(Instruction instr, string key) {
   exists(IRPropertyProvider provider |
@@ -14,9 +49,15 @@ private string getAdditionalBlockProperty(IRBlock block, string key) {
 }
 
 private newtype TPrintableIRNode =
-  TPrintableFunctionIR(FunctionIR funcIR) or
-  TPrintableIRBlock(IRBlock block) or
-  TPrintableInstruction(Instruction instr)
+  TPrintableFunctionIR(FunctionIR funcIR) {
+    shouldPrintFunction(funcIR.getFunction())
+  } or
+  TPrintableIRBlock(IRBlock block) {
+    shouldPrintFunction(block.getFunction())
+  } or
+  TPrintableInstruction(Instruction instr) {
+    shouldPrintFunction(instr.getFunction())
+  }
 
 /**
  * A node to be emitted in the IR graph.

cpp/ql/src/semmle/code/cpp/ir/implementation/raw/internal/TranslatedElement.qll

@@ -1,5 +1,7 @@
 import cpp
+import cpp
 import semmle.code.cpp.ir.implementation.raw.IR
+private import semmle.code.cpp.ir.IRConfiguration
 private import semmle.code.cpp.ir.implementation.Opcode
 private import semmle.code.cpp.ir.internal.OperandTag
 private import semmle.code.cpp.ir.internal.TempVariableTag
@@ -83,7 +85,8 @@ private predicate ignoreExprOnly(Expr expr) {
     // Ignore the allocator call, because we always synthesize it. Don't ignore
     // its arguments, though, because we use them as part of the synthesis.
     newExpr.getAllocatorCall() = expr
-  )
+  ) or
+  not translateFunction(expr.getEnclosingFunction())
 }
 
 /**
@@ -94,6 +97,49 @@ private predicate ignoreExpr(Expr expr) {
   ignoreExprAndDescendants(getRealParent*(expr))
 }
 
+/**
+ * Holds if `func` contains an AST that cannot be translated into IR. This is mostly used to work
+ * around extractor bugs. Once the relevant extractor bugs are fixed, this predicate can be removed.
+ */
+private predicate isInvalidFunction(Function func) {
+  exists(Literal literal |
+    // Constructor field inits within a compiler-generated copy constructor have a source expression
+    // that is a `Literal` with no value.
+    literal = func.(Constructor).getAnInitializer().(ConstructorFieldInit).getExpr() and
+    not exists(literal.getValue())
+  ) or
+  exists(ThisExpr thisExpr |
+    // An instantiation of a member function template is not treated as a `MemberFunction` if it has
+    // only non-type template arguments.
+    thisExpr.getEnclosingFunction() = func and
+    not func instanceof MemberFunction
+  ) or
+  exists(Expr expr |
+    // Expression missing a type.
+    expr.getEnclosingFunction() = func and
+    not exists(expr.getType())
+  )
+}
+
+/**
+ * Holds if `func` should be translated to IR.
+ */
+private predicate translateFunction(Function func) {
+  not func.isFromUninstantiatedTemplate(_) and
+  func.hasEntryPoint() and
+  not isInvalidFunction(func) and
+  exists(IRConfiguration config |
+    config.shouldCreateIRForFunction(func)
+  )
+}
+
+/**
+ * Holds if `stmt` should be translated to IR.
+ */
+private predicate translateStmt(Stmt stmt) {
+  translateFunction(stmt.getEnclosingFunction())
+}
+
 /**
  * Holds if `expr` is most naturally evaluated as control flow, rather than as
  * a value.
@@ -236,7 +282,7 @@ newtype TTranslatedElement =
   } or
   // The initialization of a field via a member of an initializer list.
   TTranslatedExplicitFieldInitialization(Expr ast, Field field,
-    Expr expr) {
+      Expr expr) {
     exists(ClassAggregateLiteral initList |
       not ignoreExpr(initList) and
       ast = initList and
@@ -260,14 +306,14 @@ newtype TTranslatedElement =
   } or
   // The initialization of an array element via a member of an initializer list.
   TTranslatedExplicitElementInitialization(
-    ArrayAggregateLiteral initList, int elementIndex) {
+      ArrayAggregateLiteral initList, int elementIndex) {
     not ignoreExpr(initList) and
     exists(initList.getElementExpr(elementIndex))
   } or
   // The value initialization of a range of array elements that were omitted
   // from an initializer list.
   TTranslatedElementValueInitialization(ArrayAggregateLiteral initList,
-    int elementIndex, int elementCount) {
+      int elementIndex, int elementCount) {
     not ignoreExpr(initList) and
     isFirstValueInitializedElementInRange(initList, elementIndex) and
     elementCount = 
@@ -287,28 +333,35 @@ newtype TTranslatedElement =
     not ignoreExpr(destruction)
   } or
   // A statement
-  TTranslatedStmt(Stmt stmt) or
+  TTranslatedStmt(Stmt stmt) {
+    translateStmt(stmt)
+  } or
   // A function
   TTranslatedFunction(Function func) {
-    func.hasEntryPoint() and
-    not func.isFromUninstantiatedTemplate(_)
+    translateFunction(func)
   } or
   // A constructor init list
   TTranslatedConstructorInitList(Function func) {
-    func.hasEntryPoint()
+    translateFunction(func)
   } or
   // A destructor destruction list
   TTranslatedDestructorDestructionList(Function func) {
-    func.hasEntryPoint()
+    translateFunction(func)
   } or
   // A function parameter
   TTranslatedParameter(Parameter param) {
-    param.getFunction().hasEntryPoint() or
-    exists(param.getCatchBlock())
+    exists(Function func |
+      (
+        func = param.getFunction() or
+        func = param.getCatchBlock().getEnclosingFunction()
+      ) and
+      translateFunction(func)
+    )
   } or
   // A local declaration
   TTranslatedDeclarationEntry(DeclarationEntry entry) {
     exists(DeclStmt declStmt |
+      translateStmt(declStmt) and
       declStmt.getADeclarationEntry() = entry
     )
   } or

cpp/ql/src/semmle/code/cpp/ir/implementation/unaliased_ssa/PrintIR.qll

@@ -1,5 +1,40 @@
 private import IR
 import cpp
+import semmle.code.cpp.ir.IRConfiguration
+
+private newtype TPrintIRConfiguration = MkPrintIRConfiguration()
+
+/**
+ * The query can extend this class to control which functions are printed.
+ */
+class PrintIRConfiguration extends TPrintIRConfiguration {
+  string toString() {
+    result = "PrintIRConfiguration"
+  }
+
+  /**
+   * Holds if the IR for `func` should be printed. By default, holds for all
+   * functions.
+   */
+  predicate shouldPrintFunction(Function func) {
+    any()
+  }
+}
+
+private predicate shouldPrintFunction(Function func) {
+  exists(PrintIRConfiguration config |
+    config.shouldPrintFunction(func)
+  )
+}
+
+/**
+ * Override of `IRConfiguration` to only create IR for the functions that are to be dumped.
+ */
+private class FilteredIRConfiguration extends IRConfiguration {
+  override predicate shouldCreateIRForFunction(Function func) {
+    shouldPrintFunction(func)
+  }
+}
 
 private string getAdditionalInstructionProperty(Instruction instr, string key) {
   exists(IRPropertyProvider provider |
@@ -14,9 +49,15 @@ private string getAdditionalBlockProperty(IRBlock block, string key) {
 }
 
 private newtype TPrintableIRNode =
-  TPrintableFunctionIR(FunctionIR funcIR) or
-  TPrintableIRBlock(IRBlock block) or
-  TPrintableInstruction(Instruction instr)
+  TPrintableFunctionIR(FunctionIR funcIR) {
+    shouldPrintFunction(funcIR.getFunction())
+  } or
+  TPrintableIRBlock(IRBlock block) {
+    shouldPrintFunction(block.getFunction())
+  } or
+  TPrintableInstruction(Instruction instr) {
+    shouldPrintFunction(instr.getFunction())
+  }
 
 /**
  * A node to be emitted in the IR graph.