[clang] [clang-tools-extra] [clang-tidy][dataflow] Add `bugprone-null-check-after-dereference` check (PR #84166)

[Dmitri Gribenko via cfe-commits]

================
@@ -0,0 +1,625 @@
+//===-- NullPointerAnalysisModel.cpp ----------------------------*- C++ -*-===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+//
+// This file defines a generic null-pointer analysis model, used for finding
+// pointer null-checks after the pointer has already been dereferenced.
+//
+// Only a limited set of operations are currently recognized. Notably, pointer
+// arithmetic, null-pointer assignments and _nullable/_nonnull attributes are
+// missing as of yet.
+//
+//===----------------------------------------------------------------------===//
+
+#include "clang/Analysis/FlowSensitive/Models/NullPointerAnalysisModel.h"
+#include "clang/AST/ASTContext.h"
+#include "clang/AST/Decl.h"
+#include "clang/AST/Expr.h"
+#include "clang/AST/Stmt.h"
+#include "clang/ASTMatchers/ASTMatchFinder.h"
+#include "clang/ASTMatchers/ASTMatchers.h"
+#include "clang/Analysis/CFG.h"
+#include "clang/Analysis/FlowSensitive/CFGMatchSwitch.h"
+#include "clang/Analysis/FlowSensitive/DataflowAnalysis.h"
+#include "clang/Analysis/FlowSensitive/DataflowEnvironment.h"
+#include "clang/Analysis/FlowSensitive/DataflowLattice.h"
+#include "clang/Analysis/FlowSensitive/MapLattice.h"
+#include "clang/Analysis/FlowSensitive/NoopLattice.h"
+#include "llvm/ADT/StringRef.h"
+#include "llvm/ADT/Twine.h"
+
+namespace clang::dataflow {
+
+namespace {
+using namespace ast_matchers;
+
+constexpr char kCond[] = "condition";
+constexpr char kVar[] = "var";
+constexpr char kValue[] = "value";
+constexpr char kIsNonnull[] = "is-nonnull";
+constexpr char kIsNull[] = "is-null";
+
+enum class SatisfiabilityResult {
+  // Returned when the value was not initialized yet.
+  Nullptr,
+  // Special value that signals that the boolean value can be anything.
+  // It signals that the underlying formulas are too complex to be calculated
+  // efficiently.
+  Top,
+  // Equivalent to the literal True in the current environment.
+  True,
+  // Equivalent to the literal False in the current environment.
+  False,
+  // Both True and False values could be produced with an appropriate set of
+  // conditions.
+  Unknown
+};
+
+using SR = SatisfiabilityResult;
+
+// FIXME: These AST matchers should also be exported via the
+// NullPointerAnalysisModel class, for tests
+auto ptrToVar(llvm::StringRef VarName = kVar) {
+  return traverse(TK_IgnoreUnlessSpelledInSource,
+                  declRefExpr(hasType(isAnyPointer())).bind(VarName));
+}
+
+auto derefMatcher() {
+  return traverse(
+      TK_IgnoreUnlessSpelledInSource,
+      unaryOperator(hasOperatorName("*"), hasUnaryOperand(ptrToVar())));
+}
+
+auto arrowMatcher() {
+  return traverse(
+      TK_IgnoreUnlessSpelledInSource,
+      memberExpr(allOf(isArrow(), hasObjectExpression(ptrToVar()))));
+}
+
+auto castExprMatcher() {
+  return castExpr(hasCastKind(CK_PointerToBoolean),
+                  hasSourceExpression(ptrToVar()))
----------------
gribozavr wrote:

Why limit the cast argument to a var? It shouldn't matter if the argument is a complex expression. As long as the framework models the value somehow (for example, one of the implicit casts that the core framework already models, like the lvalue to rvalue cast), the transfer function should not need to worry about how that ended up happening.

https://github.com/llvm/llvm-project/pull/84166