[clang-tools-extra] [clang-tidy] Add `modernize-use-bit-cast` check (PR #189962)

[Julian Schmidt via cfe-commits]

================
@@ -0,0 +1,291 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "UseBitCastCheck.h"
+#include "../utils/Matchers.h"
+#include "clang/AST/ASTContext.h"
+#include "clang/AST/Expr.h"
+#include "clang/AST/ExprCXX.h"
+#include "clang/AST/Type.h"
+#include "clang/ASTMatchers/ASTMatchFinder.h"
+#include "clang/Lex/Lexer.h"
+#include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/Twine.h"
+
+using namespace clang::ast_matchers;
+
+namespace clang::tidy::modernize {
+
+static bool isSupportedMemcpyObjectExpr(const Expr *ExprNode) {
+  ExprNode = ExprNode->IgnoreParenImpCasts();
+
+  if (isa<DeclRefExpr>(ExprNode))
+    return true;
+
+  const auto *Member = dyn_cast<MemberExpr>(ExprNode);
+  if (!Member || !isa<FieldDecl>(Member->getMemberDecl()))
+    if (const auto *MemberPointer = dyn_cast<BinaryOperator>(ExprNode))
+      if (MemberPointer->getOpcode() == BO_PtrMemD ||
+          MemberPointer->getOpcode() == BO_PtrMemI)
+        return isSupportedMemcpyObjectExpr(MemberPointer->getLHS());
+
+  return Member && isSupportedMemcpyObjectExpr(Member->getBase());
+}
+
+static const Expr *extractMemcpyObjectExpr(const Expr *ExprNode) {
+  ExprNode = ExprNode->IgnoreParenCasts();
+  const auto *AddressOf = dyn_cast<UnaryOperator>(ExprNode);
+  if (!AddressOf || AddressOf->getOpcode() != UO_AddrOf)
+    return nullptr;
+
+  const Expr *ObjectExpr = AddressOf->getSubExpr()->IgnoreParenImpCasts();
+  return isSupportedMemcpyObjectExpr(ObjectExpr) ? ObjectExpr : nullptr;
+}
+
+static bool isBitCastableMemcpyObjectType(QualType Type,
+                                          const ASTContext &Context) {
+  Type = Type.getCanonicalType().getNonReferenceType();
+  return !Type.isNull() && !Type.isVolatileQualified() &&
+         !Type->isAnyPointerType() && !Type->isFunctionType() &&
+         Type.isTriviallyCopyableType(Context) &&
+         Type.isBitwiseCloneableType(Context);
+}
+
+static bool canAssignBitCastResult(QualType Type) {
+  Type = Type.getCanonicalType().getNonReferenceType();
+  if (Type.isNull() || Type.isConstQualified() || Type->isArrayType())
+    return false;
+
+  const auto *Record = Type->getAsCXXRecordDecl();
+  return !Record || Record->hasSimpleCopyAssignment() ||
+         Record->hasSimpleMoveAssignment();
+}
+
+static bool isSameUnqualifiedCanonicalType(QualType LHS, QualType RHS) {
+  return LHS.getCanonicalType().getUnqualifiedType() ==
+         RHS.getCanonicalType().getUnqualifiedType();
+}
+
+static bool isMatchingSizeOfExpression(const Expr *SizeExpr, QualType SrcType,
+                                       QualType DstType,
+                                       const ASTContext &Context) {
+  const auto *UnaryExpr =
+      dyn_cast<UnaryExprOrTypeTraitExpr>(SizeExpr->IgnoreParenImpCasts());
+  if (!UnaryExpr || UnaryExpr->getKind() != UETT_SizeOf ||
+      SizeExpr->getBeginLoc().isMacroID())
+    return false;
+
+  const QualType SizeType = UnaryExpr->getTypeOfArgument();
+  if (SizeType.isNull())
+    return false;
+
+  const QualType SizeCanonical =
+      SizeType.getCanonicalType().getUnqualifiedType();
+  const QualType SrcCanonical = SrcType.getCanonicalType().getUnqualifiedType();
+  const QualType DstCanonical = DstType.getCanonicalType().getUnqualifiedType();
+  if (SizeCanonical != SrcCanonical && SizeCanonical != DstCanonical)
+    return false;
+
+  return Context.getTypeSizeInChars(SrcCanonical) ==
+         Context.getTypeSizeInChars(DstCanonical);
+}
+
+static bool isStatementBody(const Stmt *Current, const Stmt *Parent) {
+  if (const auto *Block = dyn_cast<CompoundStmt>(Parent))
+    return llvm::is_contained(Block->body(), Current);
+
+  if (const auto *If = dyn_cast<IfStmt>(Parent))
+    return If->getThen() == Current || If->getElse() == Current;
+  if (const auto *While = dyn_cast<WhileStmt>(Parent))
+    return While->getBody() == Current;
+  if (const auto *Do = dyn_cast<DoStmt>(Parent))
+    return Do->getBody() == Current;
+  if (const auto *For = dyn_cast<ForStmt>(Parent))
+    return For->getBody() == Current;
+  if (const auto *RangeFor = dyn_cast<CXXForRangeStmt>(Parent))
+    return RangeFor->getBody() == Current;
+  if (const auto *Label = dyn_cast<LabelStmt>(Parent))
+    return Label->getSubStmt() == Current;
+  if (const auto *Case = dyn_cast<SwitchCase>(Parent))
+    return Case->getSubStmt() == Current;
+  if (const auto *Attributed = dyn_cast<AttributedStmt>(Parent))
+    return Attributed->getSubStmt() == Current;
+
+  return false;
+}
+
+namespace {
+
+// These states describe how to spell the replacement when only the memcpy call
+// is replaced. An existing `(void)` cast is preserved by parenthesizing the
+// assignment, while comma/discarded subexpressions need an injected `(void)`.
+enum class MemcpyReplacementForm {
+  None,
+  StatementBody,
+  PreserveOuterVoidCast,
+  InjectVoidCast,
+};
+
+} // namespace
+
+static MemcpyReplacementForm getMemcpyReplacementForm(const Expr *ExprNode,
+                                                      ASTContext &Context) {
+  const Stmt *Current = ExprNode;
+  MemcpyReplacementForm Kind = MemcpyReplacementForm::StatementBody;
+
+  while (true) {
+    auto Parents = Context.getParents(*Current);
+    if (Parents.size() != 1)
+      return MemcpyReplacementForm::None;
+
+    if (const auto *ParentExpr = Parents[0].get<Expr>()) {
+      if (isa<ExprWithCleanups, ImplicitCastExpr, MaterializeTemporaryExpr,
+              CXXBindTemporaryExpr, ParenExpr>(ParentExpr)) {
+        Current = ParentExpr;
+        continue;
+      }
+
+      if (const auto *Cast = dyn_cast<CastExpr>(ParentExpr))
+        if (Cast->getCastKind() == CK_ToVoid)
+          return MemcpyReplacementForm::PreserveOuterVoidCast;
+
+      if (const auto *Comma = dyn_cast<BinaryOperator>(ParentExpr)) {
+        if (Comma->getOpcode() != BO_Comma)
+          return MemcpyReplacementForm::None;
+        if (Comma->getLHS() == Current)
+          return MemcpyReplacementForm::InjectVoidCast;
+        if (Comma->getRHS() == Current) {
+          Current = Comma;
+          Kind = MemcpyReplacementForm::InjectVoidCast;
+          continue;
+        }
+      }
+
+      return MemcpyReplacementForm::None;
+    }
+
+    const auto *ParentStmt = Parents[0].get<Stmt>();
+    if (!ParentStmt || !isStatementBody(Current, ParentStmt))
+      return MemcpyReplacementForm::None;
+    return Kind;
+  }
+}
+
+namespace {
+
+AST_MATCHER(CallExpr, isDiscardedValueContext) {
+  return getMemcpyReplacementForm(&Node, Finder->getASTContext()) !=
+         MemcpyReplacementForm::None;
+}
+
+AST_MATCHER(CallExpr, isBitCastMemcpyCandidate) {
+  if (Node.getNumArgs() != 3 || Node.getBeginLoc().isMacroID())
+    return false;
+
+  const auto *DstExpr = extractMemcpyObjectExpr(Node.getArg(0));
+  const auto *SrcExpr = extractMemcpyObjectExpr(Node.getArg(1));
+  if (!DstExpr || !SrcExpr || DstExpr->getBeginLoc().isMacroID() ||
+      SrcExpr->getBeginLoc().isMacroID())
+    return false;
+
+  const auto &Context = Finder->getASTContext();
+  const QualType DstType = DstExpr->getType().getNonReferenceType();
+  const QualType SrcType = SrcExpr->getType().getNonReferenceType();
+
+  return isBitCastableMemcpyObjectType(DstType, Context) &&
+         isBitCastableMemcpyObjectType(SrcType, Context) &&
+         canAssignBitCastResult(DstType) &&
+         !isSameUnqualifiedCanonicalType(SrcType, DstType) &&
+         isMatchingSizeOfExpression(Node.getArg(2), SrcType, DstType, Context);
+}
+
+} // namespace
+
+static StringRef getSourceText(const Expr *ExprNode, const SourceManager &SM,
+                               const LangOptions &LangOpts) {
+  return Lexer::getSourceText(
+      CharSourceRange::getTokenRange(ExprNode->getSourceRange()), SM, LangOpts);
+}
+
+UseBitCastCheck::UseBitCastCheck(StringRef Name, ClangTidyContext *Context)
+    : ClangTidyCheck(Name, Context),
+      IncludeInserter(Options.getLocalOrGlobal("IncludeStyle",
+                                               utils::IncludeSorter::IS_LLVM),
+                      areDiagsSelfContained()) {}
+
+void UseBitCastCheck::storeOptions(ClangTidyOptions::OptionMap &Opts) {
+  Options.store(Opts, "IncludeStyle", IncludeInserter.getStyle());
+}
+
+void UseBitCastCheck::registerPPCallbacks(const SourceManager &SM,
+                                          Preprocessor *PP,
+                                          Preprocessor *ModuleExpanderPP) {
+  IncludeInserter.registerPreprocessor(PP);
+}
+
+void UseBitCastCheck::registerMatchers(MatchFinder *Finder) {
+  Finder->addMatcher(
+      callExpr(callee(functionDecl(hasName("::memcpy"))),
+               isDiscardedValueContext(), isBitCastMemcpyCandidate(),
+               unless(hasAncestor(expr(matchers::hasUnevaluatedContext()))))
----------------
5chmidti wrote:

The callExpr is not in an unevaluated context already if the ancestor is?

https://github.com/llvm/llvm-project/pull/189962