[-Wunsafe-buffer-usage] Add fixits for array to pointer assignment (llvm#81343)

Introducing CArrayToPtrAssignment gadget and implementing fixits for some cases
of array being assigned to pointer.

Key observations:
- const size array can be assigned to std::span and bounds are propagated
- const size array can't be on LHS of assignment
This means array to pointer assignment has no strategy implications.

Fixits are implemented for cases where one of the variables in the assignment is
safe. For assignment of a safe array to unsafe pointer we know that the RHS will
never be transformed since it's safe and can immediately emit the optimal fixit.
Similarly for assignment of unsafe array to safe pointer.
(Obviously this is not and can't be future-proof in regards to what
variables we consider unsafe and that is fine.)

Fixits for assignment from unsafe array to unsafe pointer (from Array to Span
strategy) are not implemented in this patch as that needs to be properly designed
first - we might possibly implement optimal fixits for partially transformed
cases, put both variables in a single fixit group or do something else.

(cherry picked from commit 6fce42f)

Author: jkorous-apple

clang/include/clang/Analysis/Analyses/UnsafeBufferUsageGadgets.def

@@ -45,7 +45,8 @@ FIXABLE_GADGET(UPCAddressofArraySubscript) // '&DRE[any]' in an Unspecified Poin
 FIXABLE_GADGET(UPCStandalonePointer)
 FIXABLE_GADGET(UPCPreIncrement)            // '++Ptr' in an Unspecified Pointer Context
 FIXABLE_GADGET(UUCAddAssign)            // 'Ptr += n' in an Unspecified Untyped Context
-FIXABLE_GADGET(PointerAssignment)
+FIXABLE_GADGET(PtrToPtrAssignment)
+FIXABLE_GADGET(CArrayToPtrAssignment)
 FIXABLE_GADGET(PointerInit)
 
 #undef FIXABLE_GADGET

clang/lib/Analysis/UnsafeBufferUsage.cpp

@@ -7,11 +7,14 @@
 //===----------------------------------------------------------------------===//
 
 #include "clang/Analysis/Analyses/UnsafeBufferUsage.h"
+#include "clang/AST/ASTContext.h"
 #include "clang/AST/Decl.h"
 #include "clang/AST/Expr.h"
 #include "clang/AST/RecursiveASTVisitor.h"
+#include "clang/AST/Stmt.h"
 #include "clang/AST/StmtVisitor.h"
 #include "clang/ASTMatchers/ASTMatchFinder.h"
+#include "clang/ASTMatchers/ASTMatchers.h"
 #include "clang/Basic/CharInfo.h"
 #include "clang/Basic/SourceLocation.h"
 #include "clang/Lex/Lexer.h"
@@ -760,21 +763,22 @@ class PointerInitGadget : public FixableGadget {
 ///  \code
 ///  p = q;
 ///  \endcode
-class PointerAssignmentGadget : public FixableGadget {
+/// where both `p` and `q` are pointers.
+class PtrToPtrAssignmentGadget : public FixableGadget {
 private:
   static constexpr const char *const PointerAssignLHSTag = "ptrLHS";
   static constexpr const char *const PointerAssignRHSTag = "ptrRHS";
   const DeclRefExpr *PtrLHS; // the LHS pointer expression in `PA`
   const DeclRefExpr *PtrRHS; // the RHS pointer expression in `PA`
 
 public:
-  PointerAssignmentGadget(const MatchFinder::MatchResult &Result)
-      : FixableGadget(Kind::PointerAssignment),
+  PtrToPtrAssignmentGadget(const MatchFinder::MatchResult &Result)
+      : FixableGadget(Kind::PtrToPtrAssignment),
         PtrLHS(Result.Nodes.getNodeAs<DeclRefExpr>(PointerAssignLHSTag)),
         PtrRHS(Result.Nodes.getNodeAs<DeclRefExpr>(PointerAssignRHSTag)) {}
 
   static bool classof(const Gadget *G) {
-    return G->getKind() == Kind::PointerAssignment;
+    return G->getKind() == Kind::PtrToPtrAssignment;
   }
 
   static Matcher matcher() {
@@ -809,6 +813,60 @@ class PointerAssignmentGadget : public FixableGadget {
   }
 };
 
+/// An assignment expression of the form:
+///  \code
+///  ptr = array;
+///  \endcode
+/// where `p` is a pointer and `array` is a constant size array.
+class CArrayToPtrAssignmentGadget : public FixableGadget {
+private:
+  static constexpr const char *const PointerAssignLHSTag = "ptrLHS";
+  static constexpr const char *const PointerAssignRHSTag = "ptrRHS";
+  const DeclRefExpr *PtrLHS; // the LHS pointer expression in `PA`
+  const DeclRefExpr *PtrRHS; // the RHS pointer expression in `PA`
+
+public:
+  CArrayToPtrAssignmentGadget(const MatchFinder::MatchResult &Result)
+      : FixableGadget(Kind::CArrayToPtrAssignment),
+        PtrLHS(Result.Nodes.getNodeAs<DeclRefExpr>(PointerAssignLHSTag)),
+        PtrRHS(Result.Nodes.getNodeAs<DeclRefExpr>(PointerAssignRHSTag)) {}
+
+  static bool classof(const Gadget *G) {
+    return G->getKind() == Kind::CArrayToPtrAssignment;
+  }
+
+  static Matcher matcher() {
+    auto PtrAssignExpr = binaryOperator(
+        allOf(hasOperatorName("="),
+              hasRHS(ignoringParenImpCasts(
+                  declRefExpr(hasType(hasCanonicalType(constantArrayType())),
+                              toSupportedVariable())
+                      .bind(PointerAssignRHSTag))),
+              hasLHS(declRefExpr(hasPointerType(), toSupportedVariable())
+                         .bind(PointerAssignLHSTag))));
+
+    return stmt(isInUnspecifiedUntypedContext(PtrAssignExpr));
+  }
+
+  virtual std::optional<FixItList>
+  getFixits(const FixitStrategy &S) const override;
+
+  virtual const Stmt *getBaseStmt() const override {
+    // FIXME: This should be the binary operator, assuming that this method
+    // makes sense at all on a FixableGadget.
+    return PtrLHS;
+  }
+
+  virtual DeclUseList getClaimedVarUseSites() const override {
+    return DeclUseList{PtrLHS, PtrRHS};
+  }
+
+  virtual std::optional<std::pair<const VarDecl *, const VarDecl *>>
+  getStrategyImplications() const override {
+    return {};
+  }
+};
+
 /// A call of a function or method that performs unchecked buffer operations
 /// over one of its pointer parameters.
 class UnsafeBufferUsageAttrGadget : public WarningGadget {
@@ -1428,7 +1486,7 @@ bool clang::internal::anyConflict(const SmallVectorImpl<FixItHint> &FixIts,
 }
 
 std::optional<FixItList>
-PointerAssignmentGadget::getFixits(const FixitStrategy &S) const {
+PtrToPtrAssignmentGadget::getFixits(const FixitStrategy &S) const {
   const auto *LeftVD = cast<VarDecl>(PtrLHS->getDecl());
   const auto *RightVD = cast<VarDecl>(PtrRHS->getDecl());
   switch (S.lookup(LeftVD)) {
@@ -1447,6 +1505,42 @@ PointerAssignmentGadget::getFixits(const FixitStrategy &S) const {
   return std::nullopt;
 }
 
+/// \returns fixit that adds .data() call after \DRE.
+static inline std::optional<FixItList> createDataFixit(const ASTContext &Ctx,
+                                                       const DeclRefExpr *DRE);
+
+std::optional<FixItList>
+CArrayToPtrAssignmentGadget::getFixits(const FixitStrategy &S) const {
+  const auto *LeftVD = cast<VarDecl>(PtrLHS->getDecl());
+  const auto *RightVD = cast<VarDecl>(PtrRHS->getDecl());
+  // TLDR: Implementing fixits for non-Wontfix strategy on both LHS and RHS is
+  // non-trivial.
+  //
+  // CArrayToPtrAssignmentGadget doesn't have strategy implications because
+  // constant size array propagates its bounds. Because of that LHS and RHS are
+  // addressed by two different fixits.
+  //
+  // At the same time FixitStrategy S doesn't reflect what group a fixit belongs
+  // to and can't be generally relied on in multi-variable Fixables!
+  //
+  // E. g. If an instance of this gadget is fixing variable on LHS then the
+  // variable on RHS is fixed by a different fixit and its strategy for LHS
+  // fixit is as if Wontfix.
+  //
+  // The only exception is Wontfix strategy for a given variable as that is
+  // valid for any fixit produced for the given input source code.
+  if (S.lookup(LeftVD) == FixitStrategy::Kind::Span) {
+    if (S.lookup(RightVD) == FixitStrategy::Kind::Wontfix) {
+      return FixItList{};
+    }
+  } else if (S.lookup(LeftVD) == FixitStrategy::Kind::Wontfix) {
+    if (S.lookup(RightVD) == FixitStrategy::Kind::Array) {
+      return createDataFixit(RightVD->getASTContext(), PtrRHS);
+    }
+  }
+  return std::nullopt;
+}
+
 std::optional<FixItList>
 PointerInitGadget::getFixits(const FixitStrategy &S) const {
   const auto *LeftVD = PtrInitLHS;
@@ -1864,6 +1958,19 @@ PointerDereferenceGadget::getFixits(const FixitStrategy &S) const {
   return std::nullopt;
 }
 
+static inline std::optional<FixItList> createDataFixit(const ASTContext &Ctx,
+                                                       const DeclRefExpr *DRE) {
+  const SourceManager &SM = Ctx.getSourceManager();
+  // Inserts the .data() after the DRE
+  std::optional<SourceLocation> EndOfOperand =
+      getPastLoc(DRE, SM, Ctx.getLangOpts());
+
+  if (EndOfOperand)
+    return FixItList{{FixItHint::CreateInsertion(*EndOfOperand, ".data()")}};
+
+  return std::nullopt;
+}
+
 // Generates fix-its replacing an expression of the form UPC(DRE) with
 // `DRE.data()`
 std::optional<FixItList>
@@ -1872,14 +1979,7 @@ UPCStandalonePointerGadget::getFixits(const FixitStrategy &S) const {
   switch (S.lookup(VD)) {
   case FixitStrategy::Kind::Array:
   case FixitStrategy::Kind::Span: {
-    ASTContext &Ctx = VD->getASTContext();
-    SourceManager &SM = Ctx.getSourceManager();
-    // Inserts the .data() after the DRE
-    std::optional<SourceLocation> EndOfOperand =
-        getPastLoc(Node, SM, Ctx.getLangOpts());
-
-    if (EndOfOperand)
-      return FixItList{{FixItHint::CreateInsertion(*EndOfOperand, ".data()")}};
+    return createDataFixit(VD->getASTContext(), Node);
     // FIXME: Points inside a macro expansion.
     break;
   }

clang/test/SemaCXX/warn-unsafe-buffer-usage-debug.cpp

@@ -53,7 +53,7 @@ void unclaimed_use() {
 void implied_unclaimed_var(int *b) {  // expected-warning{{'b' is an unsafe pointer used for buffer access}}
   int *a = new int[3];  // expected-warning{{'a' is an unsafe pointer used for buffer access}}
   a[4] = 7;  // expected-note{{used in buffer access here}}
-  a = b;  // debug-note{{safe buffers debug: gadget 'PointerAssignment' refused to produce a fix}}
+  a = b;  // debug-note{{safe buffers debug: gadget 'PtrToPtrAssignment' refused to produce a fix}}
   b++;  // expected-note{{used in pointer arithmetic here}} \
         // debug-note{{safe buffers debug: failed to produce fixit for 'b' : has an unclaimed use}}
 }

clang/test/SemaCXX/warn-unsafe-buffer-usage-fixits-array-assign-to-ptr.cpp

@@ -0,0 +1,45 @@
+// RUN: %clang_cc1 -std=c++20 -Wunsafe-buffer-usage \
+// RUN:            -fsafe-buffer-usage-suggestions \
+// RUN:            -fdiagnostics-parseable-fixits %s 2>&1 | FileCheck %s
+
+void safe_array_assigned_to_safe_ptr(unsigned idx) {
+  int buffer[10];
+  // CHECK-NOT: fix-it:"{{.*}}":{[[@LINE-1]]:
+  int* ptr;
+  // CHECK-NOT: fix-it:"{{.*}}":{[[@LINE-1]]:
+  ptr = buffer;
+  // CHECK-NOT: fix-it:"{{.*}}":{[[@LINE-1]]:
+}
+
+void safe_array_assigned_to_unsafe_ptr(unsigned idx) {
+  int buffer[10];
+  // CHECK-NOT: fix-it:"{{.*}}":{[[@LINE-1]]:
+  int* ptr;
+  // CHECK: fix-it:"{{.*}}":{[[@LINE-1]]:3-[[@LINE-1]]:11}:"std::span<int> ptr"
+  ptr = buffer;
+  // CHECK-NOT: fix-it:"{{.*}}":{[[@LINE-1]]:
+  ptr[idx] = 0;
+}
+
+void unsafe_array_assigned_to_safe_ptr(unsigned idx) {
+  int buffer[10];
+  // CHECK: fix-it:"{{.*}}":{[[@LINE-1]]:3-[[@LINE-1]]:17}:"std::array<int, 10> buffer"
+  int* ptr;
+  // CHECK-NOT: fix-it:"{{.*}}":{[[@LINE-1]]:
+  ptr = buffer;
+  // CHECK: fix-it:"{{.*}}":{[[@LINE-1]]:15-[[@LINE-1]]:15}:".data()"
+  buffer[idx] = 0;
+}
+
+// FIXME: Implement fixit/s for this case.
+// See comment in CArrayToPtrAssignmentGadget::getFixits to learn why this hasn't been implemented.
+void unsafe_array_assigned_to_unsafe_ptr(unsigned idx) {
+  int buffer[10];
+  // CHECK-NOT: fix-it:"{{.*}}":{[[@LINE-1]]:{{.*}}
+  int* ptr;
+  // CHECK-NOT: fix-it:"{{.*}}":{[[@LINE-1]]:{{.*}}
+  ptr = buffer;
+  // CHECK-NOT: fix-it:"{{.*}}":{[[@LINE-1]]:{{.*}}
+  buffer[idx] = 0;
+  ptr[idx] = 0;
+}