[Clang] add typo correction for unknown attribute names (#140629)

This patch enhances Clang's diagnosis for unknown attributes by
providing typo correction suggestions for known attributes.

```cpp
[[gmu::deprected]] // expected-warning {{unknown attribute 'gmu::deprected' ignored; did you mean 'gnu::deprecated'?}}
int f1(void) {
  return 0;
}

[[deprected]] // expected-warning {{unknown attribute 'deprected' ignored; did you mean 'deprecated'?}}
int f2(void) {
  return 0;
}
```

Author: a-tarasyuk

clang/docs/ReleaseNotes.rst

@@ -563,6 +563,8 @@ Improvements to Clang's diagnostics
 - Fixed a crash when checking a ``__thread``-specified variable declaration
   with a dependent type in C++. (#GH140509)
 
+- Clang now suggests corrections for unknown attribute names.
+
 Improvements to Clang's time-trace
 ----------------------------------
 

clang/include/clang/Basic/AttributeCommonInfo.h

@@ -21,6 +21,8 @@ namespace clang {
 
 class ASTRecordWriter;
 class IdentifierInfo;
+class LangOptions;
+class TargetInfo;
 
 class AttributeCommonInfo {
 public:
@@ -196,6 +198,9 @@ class AttributeCommonInfo {
   /// with surrounding underscores removed as appropriate (e.g.
   /// __gnu__::__attr__ will be normalized to gnu::attr).
   std::string getNormalizedFullName() const;
+  std::optional<std::string>
+  getCorrectedFullName(const TargetInfo &Target,
+                       const LangOptions &LangOpts) const;
   SourceRange getNormalizedRange() const;
 
   bool isDeclspecAttribute() const { return SyntaxUsed == AS_Declspec; }

clang/include/clang/Basic/Attributes.h

@@ -19,6 +19,10 @@ class TargetInfo;
 
 /// Return the version number associated with the attribute if we
 /// recognize and implement the attribute specified by the given information.
+int hasAttribute(AttributeCommonInfo::Syntax Syntax, llvm::StringRef ScopeName,
+                 llvm::StringRef AttrName, const TargetInfo &Target,
+                 const LangOptions &LangOpts, bool CheckPlugins);
+
 int hasAttribute(AttributeCommonInfo::Syntax Syntax,
                  const IdentifierInfo *Scope, const IdentifierInfo *Attr,
                  const TargetInfo &Target, const LangOptions &LangOpts);

clang/include/clang/Basic/CMakeLists.txt

@@ -79,6 +79,12 @@ clang_tablegen(CXX11AttributeInfo.inc -gen-cxx11-attribute-info
   TARGET CXX11AttributeInfo
   )
 
+  clang_tablegen(AttributeSpellingList.inc -gen-attribute-spelling-list
+  -I ${CMAKE_CURRENT_SOURCE_DIR}/../../
+  SOURCE Attr.td
+  TARGET AttributeSpellingList
+  )
+
 clang_tablegen(Builtins.inc -gen-clang-builtins
   SOURCE Builtins.td
   TARGET ClangBuiltins)

clang/include/clang/Basic/DiagnosticCommonKinds.td

@@ -181,6 +181,8 @@ def err_opencl_unknown_type_specifier : Error<
 
 def warn_unknown_attribute_ignored : Warning<
   "unknown attribute %0 ignored">, InGroup<UnknownAttributes>;
+def warn_unknown_attribute_ignored_suggestion : Warning<
+  "unknown attribute %0 ignored; did you mean '%1'?">, InGroup<UnknownAttributes>;
 def warn_attribute_ignored : Warning<"%0 attribute ignored">,
   InGroup<IgnoredAttributes>;
 def err_keyword_not_supported_on_target : Error<

clang/include/clang/Basic/SimpleTypoCorrection.h

@@ -0,0 +1,48 @@
+//===- SimpleTypoCorrection.h - Basic typo correction utility -*- 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 the SimpleTypoCorrection class, which performs basic
+// typo correction using string similarity based on edit distance.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_CLANG_BASIC_SIMPLETYPOCORRECTION_H
+#define LLVM_CLANG_BASIC_SIMPLETYPOCORRECTION_H
+
+#include "clang/Basic/LLVM.h"
+#include "llvm/ADT/StringRef.h"
+
+namespace clang {
+
+class IdentifierInfo;
+
+class SimpleTypoCorrection {
+  StringRef BestCandidate;
+  StringRef Typo;
+
+  const unsigned MaxEditDistance;
+  unsigned BestEditDistance;
+  unsigned BestIndex;
+  unsigned NextIndex;
+
+public:
+  explicit SimpleTypoCorrection(StringRef Typo)
+      : BestCandidate(), Typo(Typo), MaxEditDistance((Typo.size() + 2) / 3),
+        BestEditDistance(MaxEditDistance + 1), BestIndex(0), NextIndex(0) {}
+
+  void add(const StringRef Candidate);
+  void add(const char *Candidate);
+  void add(const IdentifierInfo *Candidate);
+
+  std::optional<StringRef> getCorrection() const;
+  bool hasCorrection() const;
+  unsigned getCorrectionIndex() const;
+};
+} // namespace clang
+
+#endif // LLVM_CLANG_BASIC_SIMPLETYPOCORRECTION_H

clang/include/clang/Sema/Sema.h

@@ -5033,6 +5033,8 @@ class Sema final : public SemaBase {
   /// which might be lying around on it.
   void checkUnusedDeclAttributes(Declarator &D);
 
+  void DiagnoseUnknownAttribute(const ParsedAttr &AL);
+
   /// DeclClonePragmaWeak - clone existing decl (maybe definition),
   /// \#pragma weak needs a non-definition decl and source may not have one.
   NamedDecl *DeclClonePragmaWeak(NamedDecl *ND, const IdentifierInfo *II,

clang/lib/AST/CommentSema.cpp

@@ -13,6 +13,7 @@
 #include "clang/AST/DeclTemplate.h"
 #include "clang/Basic/DiagnosticComment.h"
 #include "clang/Basic/LLVM.h"
+#include "clang/Basic/SimpleTypoCorrection.h"
 #include "clang/Basic/SourceManager.h"
 #include "clang/Lex/Preprocessor.h"
 #include "llvm/ADT/StringSwitch.h"
@@ -975,69 +976,22 @@ unsigned Sema::resolveParmVarReference(StringRef Name,
   return ParamCommandComment::InvalidParamIndex;
 }
 
-namespace {
-class SimpleTypoCorrector {
-  const NamedDecl *BestDecl;
-
-  StringRef Typo;
-  const unsigned MaxEditDistance;
-
-  unsigned BestEditDistance;
-  unsigned BestIndex;
-  unsigned NextIndex;
-
-public:
-  explicit SimpleTypoCorrector(StringRef Typo)
-      : BestDecl(nullptr), Typo(Typo), MaxEditDistance((Typo.size() + 2) / 3),
-        BestEditDistance(MaxEditDistance + 1), BestIndex(0), NextIndex(0) {}
-
-  void addDecl(const NamedDecl *ND);
-
-  const NamedDecl *getBestDecl() const {
-    if (BestEditDistance > MaxEditDistance)
-      return nullptr;
-
-    return BestDecl;
-  }
+unsigned
+Sema::correctTypoInParmVarReference(StringRef Typo,
+                                    ArrayRef<const ParmVarDecl *> ParamVars) {
+  SimpleTypoCorrection STC(Typo);
+  for (unsigned i = 0, e = ParamVars.size(); i != e; ++i) {
+    const ParmVarDecl *Param = ParamVars[i];
+    if (!Param)
+      continue;
 
-  unsigned getBestDeclIndex() const {
-    assert(getBestDecl());
-    return BestIndex;
+    STC.add(Param->getIdentifier());
   }
-};
-
-void SimpleTypoCorrector::addDecl(const NamedDecl *ND) {
-  unsigned CurrIndex = NextIndex++;
-
-  const IdentifierInfo *II = ND->getIdentifier();
-  if (!II)
-    return;
 
-  StringRef Name = II->getName();
-  unsigned MinPossibleEditDistance = abs((int)Name.size() - (int)Typo.size());
-  if (MinPossibleEditDistance > 0 &&
-      Typo.size() / MinPossibleEditDistance < 3)
-    return;
+  if (STC.hasCorrection())
+    return STC.getCorrectionIndex();
 
-  unsigned EditDistance = Typo.edit_distance(Name, true, MaxEditDistance);
-  if (EditDistance < BestEditDistance) {
-    BestEditDistance = EditDistance;
-    BestDecl = ND;
-    BestIndex = CurrIndex;
-  }
-}
-} // end anonymous namespace
-
-unsigned Sema::correctTypoInParmVarReference(
-                                    StringRef Typo,
-                                    ArrayRef<const ParmVarDecl *> ParamVars) {
-  SimpleTypoCorrector Corrector(Typo);
-  for (unsigned i = 0, e = ParamVars.size(); i != e; ++i)
-    Corrector.addDecl(ParamVars[i]);
-  if (Corrector.getBestDecl())
-    return Corrector.getBestDeclIndex();
-  else
-    return ParamCommandComment::InvalidParamIndex;
+  return ParamCommandComment::InvalidParamIndex;
 }
 
 namespace {
@@ -1079,30 +1033,31 @@ bool Sema::resolveTParamReference(
 
 namespace {
 void CorrectTypoInTParamReferenceHelper(
-                            const TemplateParameterList *TemplateParameters,
-                            SimpleTypoCorrector &Corrector) {
+    const TemplateParameterList *TemplateParameters,
+    SimpleTypoCorrection &STC) {
   for (unsigned i = 0, e = TemplateParameters->size(); i != e; ++i) {
     const NamedDecl *Param = TemplateParameters->getParam(i);
-    Corrector.addDecl(Param);
+    if (!Param)
+      continue;
+
+    STC.add(Param->getIdentifier());
 
     if (const TemplateTemplateParmDecl *TTP =
             dyn_cast<TemplateTemplateParmDecl>(Param))
-      CorrectTypoInTParamReferenceHelper(TTP->getTemplateParameters(),
-                                         Corrector);
+      CorrectTypoInTParamReferenceHelper(TTP->getTemplateParameters(), STC);
   }
 }
 } // end anonymous namespace
 
 StringRef Sema::correctTypoInTParamReference(
                             StringRef Typo,
                             const TemplateParameterList *TemplateParameters) {
-  SimpleTypoCorrector Corrector(Typo);
-  CorrectTypoInTParamReferenceHelper(TemplateParameters, Corrector);
-  if (const NamedDecl *ND = Corrector.getBestDecl()) {
-    const IdentifierInfo *II = ND->getIdentifier();
-    assert(II && "SimpleTypoCorrector should not return this decl");
-    return II->getName();
-  }
+  SimpleTypoCorrection STC(Typo);
+  CorrectTypoInTParamReferenceHelper(TemplateParameters, STC);
+
+  if (auto CorrectedTParamReference = STC.getCorrection())
+    return *CorrectedTParamReference;
+
   return StringRef();
 }