Merged main:ddf1de20a3f7 into amd-gfx:756ea0fac61f

Local branch amd-gfx 756ea0f Merged main:b0b8e83e668a into amd-gfx:b9f9dea74514
Remote branch main ddf1de2 [hwasan] Fix rare false negative (zero tag) in stack-uar.c (llvm#69374)

Author: SC llvm team

clang-tools-extra/docs/clang-tidy/checks/list.rst

@@ -241,7 +241,7 @@ Clang-Tidy Checks
    :doc:`llvmlibc-restrict-system-libc-headers <llvmlibc/restrict-system-libc-headers>`, "Yes"
    :doc:`misc-confusable-identifiers <misc/confusable-identifiers>`,
    :doc:`misc-const-correctness <misc/const-correctness>`, "Yes"
-   :doc:`misc-coroutine-hostile-raii <misc/coroutine-hostile-raii.html>`_,
+   :doc:`misc-coroutine-hostile-raii <misc/coroutine-hostile-raii>`,
    :doc:`misc-definitions-in-headers <misc/definitions-in-headers>`, "Yes"
    :doc:`misc-header-include-cycle <misc/header-include-cycle>`,
    :doc:`misc-include-cleaner <misc/include-cleaner>`, "Yes"

clang-tools-extra/docs/clang-tidy/checks/misc/coroutine-hostile-raii.rst

@@ -1,28 +1,28 @@
 .. title:: clang-tidy - misc-coroutine-hostile-raii
 
 misc-coroutine-hostile-raii
-====================
+===========================
 
 Detects when objects of certain hostile RAII types persists across suspension
 points in a coroutine. Such hostile types include scoped-lockable types and
 types belonging to a configurable denylist.
 
-Some objects require that they be destroyed on the same thread that created them. 
+Some objects require that they be destroyed on the same thread that created them.
 Traditionally this requirement was often phrased as "must be a local variable",
 under the assumption that local variables always work this way. However this is
 incorrect with C++20 coroutines, since an intervening ``co_await`` may cause the
 coroutine to suspend and later be resumed on another thread.
 
-The lifetime of an object that requires being destroyed on the same thread must 
+The lifetime of an object that requires being destroyed on the same thread must
 not encompass a ``co_await`` or ``co_yield`` point. If you create/destroy an object,
 you must do so without allowing the coroutine to suspend in the meantime.
 
 Following types are considered as hostile:
 
  - Scoped-lockable types: A scoped-lockable object persisting across a suspension
- point is problematic as the lock held by this object could be unlocked by a 
- different thread. This would be undefined behaviour.
- This includes all types annotated with the ``scoped_lockable`` attribute.
+   point is problematic as the lock held by this object could be unlocked by a
+   different thread. This would be undefined behaviour.
+   This includes all types annotated with the ``scoped_lockable`` attribute.
 
  - Types belonging to a configurable denylist.
 
@@ -44,7 +44,7 @@ Options
 
 .. option:: RAIITypesList
 
-    A semicolon-separated list of qualified types which should not be allowed to 
+    A semicolon-separated list of qualified types which should not be allowed to
     persist across suspension points.
     Eg: ``my::lockable; a::b;::my::other::lockable;``
-    The default value of this option is `"std::lock_guard;std::scoped_lock"`.
+    The default value of this option is `"std::lock_guard;std::scoped_lock"`.

clang/docs/ReleaseNotes.rst

@@ -117,8 +117,6 @@ C++ Language Changes
 
 C++20 Feature Support
 ^^^^^^^^^^^^^^^^^^^^^
-- Fix a bug in conversion sequence of arguments to a function with reversed parameter order.
-  Fixes `GH <https://github.com/llvm/llvm-project/issues/53954>`_.
 
 C++23 Feature Support
 ^^^^^^^^^^^^^^^^^^^^^
@@ -199,6 +197,9 @@ New Compiler Flags
   the preprocessed text to the output. This can greatly reduce the size of the
   preprocessed output, which can be helpful when trying to reduce a test case.
 
+* ``-Wbitfield-conversion`` was added to detect assignments of integral
+  types to a bitfield that may change the value.
+
 Deprecated Compiler Flags
 -------------------------
 
@@ -522,6 +523,10 @@ Bug Fixes to C++ Support
   with non-type template parameters of reference type. Fixes:
   (`#65153 <https://github.com/llvm/llvm-project/issues/65153>`_)
 
+- Clang now properly compares constraints on an out of line class template
+  declaration definition. Fixes:
+  (`#61763 <https://github.com/llvm/llvm-project/issues/61763>`_)
+
 Bug Fixes to AST Handling
 ^^^^^^^^^^^^^^^^^^^^^^^^^
 - Fixed an import failure of recursive friend class template.

clang/include/clang/AST/Expr.h

@@ -607,6 +607,13 @@ class Expr : public ValueStmt {
     /// foldable. If the expression is foldable, but not a constant expression,
     /// the notes will describes why it isn't a constant expression. If the
     /// expression *is* a constant expression, no notes will be produced.
+    ///
+    /// FIXME: this causes significant performance concerns and should be
+    /// refactored at some point. Not all evaluations of the constant
+    /// expression interpreter will display the given diagnostics, this means
+    /// those kinds of uses are paying the expense of generating a diagnostic
+    /// (which may include expensive operations like converting APValue objects
+    /// to a string representation).
     SmallVectorImpl<PartialDiagnosticAt> *Diag = nullptr;
 
     EvalStatus() = default;

clang/include/clang/Basic/DiagnosticGroups.td

@@ -53,6 +53,7 @@ def SingleBitBitFieldConstantConversion :
 def BitFieldConstantConversion : DiagGroup<"bitfield-constant-conversion",
                                            [SingleBitBitFieldConstantConversion]>;
 def BitFieldEnumConversion : DiagGroup<"bitfield-enum-conversion">;
+def BitFieldConversion : DiagGroup<"bitfield-conversion">;
 def BitFieldWidth : DiagGroup<"bitfield-width">;
 def CompoundTokenSplitByMacro : DiagGroup<"compound-token-split-by-macro">;
 def CompoundTokenSplitBySpace : DiagGroup<"compound-token-split-by-space">;
@@ -933,6 +934,7 @@ def Conversion : DiagGroup<"conversion",
                             ConstantConversion,
                             EnumConversion,
                             BitFieldEnumConversion,
+                            BitFieldConversion,
                             FloatConversion,
                             Shorten64To32,
                             IntConversion,

clang/include/clang/Basic/DiagnosticSemaKinds.td

@@ -6171,6 +6171,9 @@ def warn_signed_bitfield_enum_conversion : Warning<
   "signed bit-field %0 needs an extra bit to represent the largest positive "
   "enumerators of %1">,
   InGroup<BitFieldEnumConversion>, DefaultIgnore;
+def warn_bitfield_too_small_for_integral_type : Warning<
+  "conversion from %2 (%3 bits) to bit-field %0 (%1 bits) may change value">,
+  InGroup<BitFieldConversion>, DefaultIgnore;
 def note_change_bitfield_sign : Note<
   "consider making the bitfield type %select{unsigned|signed}0">;
 

clang/include/clang/Basic/arm_sve.td

@@ -1865,10 +1865,21 @@ def SVPTRUE_COUNT  : SInst<"svptrue_{d}", "}v", "QcQsQiQl", MergeNone, "aarch64_
 
 def SVPEXT_SINGLE : SInst<"svpext_lane_{d}", "P}i", "QcQsQiQl", MergeNone, "aarch64_sve_pext", [], [ImmCheck<1, ImmCheck0_3>]>;
 def SVPEXT_X2     : SInst<"svpext_lane_{d}_x2", "2.P}i", "QcQsQiQl", MergeNone, "aarch64_sve_pext_x2", [], [ImmCheck<1, ImmCheck0_1>]>;
+
+def SVPSEL_COUNT_ALIAS_B : SInst<"svpsel_lane_c8",  "}}Pm", "Pc", MergeNone, "", [], []>;
+def SVPSEL_COUNT_ALIAS_H : SInst<"svpsel_lane_c16", "}}Pm", "Ps", MergeNone, "", [], []>;
+def SVPSEL_COUNT_ALIAS_S : SInst<"svpsel_lane_c32", "}}Pm", "Pi", MergeNone, "", [], []>;
+def SVPSEL_COUNT_ALIAS_D : SInst<"svpsel_lane_c64", "}}Pm", "Pl", MergeNone, "", [], []>;
 }
 
 let TargetGuard = "sve2p1" in {
 def SVSCLAMP : SInst<"svclamp[_{d}]", "dddd", "csil",     MergeNone, "aarch64_sve_sclamp", [], []>;
 def SVUCLAMP : SInst<"svclamp[_{d}]", "dddd", "UcUsUiUl", MergeNone, "aarch64_sve_uclamp", [], []>;
+
+def SVPSEL_B : SInst<"svpsel_lane_b8",  "PPPm", "Pc", MergeNone, "", [], []>;
+def SVPSEL_H : SInst<"svpsel_lane_b16", "PPPm", "Ps", MergeNone, "", [], []>;
+def SVPSEL_S : SInst<"svpsel_lane_b32", "PPPm", "Pi", MergeNone, "", [], []>;
+def SVPSEL_D : SInst<"svpsel_lane_b64", "PPPm", "Pl", MergeNone, "", [], []>;
+
 def SVCNTP_COUNT : SInst<"svcntp_{d}", "n}i", "QcQsQiQl", MergeNone, "aarch64_sve_cntp_{d}", [IsOverloadNone], [ImmCheck<1, ImmCheck2_4_Mul2>]>;
 }

clang/include/clang/Sema/Sema.h

@@ -3809,17 +3809,6 @@ class Sema final {
   // the purposes of [temp.friend] p9.
   bool FriendConstraintsDependOnEnclosingTemplate(const FunctionDecl *FD);
 
-  // Calculates whether two constraint expressions are equal irrespective of a
-  // difference in 'depth'. This takes a pair of optional 'NamedDecl's 'Old' and
-  // 'New', which are the "source" of the constraint, since this is necessary
-  // for figuring out the relative 'depth' of the constraint. The depth of the
-  // 'primary template' and the 'instantiated from' templates aren't necessarily
-  // the same, such as a case when one is a 'friend' defined in a class.
-  bool AreConstraintExpressionsEqual(const NamedDecl *Old,
-                                     const Expr *OldConstr,
-                                     const NamedDecl *New,
-                                     const Expr *NewConstr);
-
   enum class AllowedExplicit {
     /// Allow no explicit functions to be used.
     None,
@@ -8615,8 +8604,48 @@ class Sema final {
     TPL_TemplateParamsEquivalent,
   };
 
+  // A struct to represent the 'new' declaration, which is either itself just
+  // the named decl, or the important information we need about it in order to
+  // do constraint comparisons.
+  class TemplateCompareNewDeclInfo {
+    const NamedDecl *ND = nullptr;
+    const DeclContext *DC = nullptr;
+    const DeclContext *LexicalDC = nullptr;
+    SourceLocation Loc;
+
+  public:
+    TemplateCompareNewDeclInfo(const NamedDecl *ND) : ND(ND) {}
+    TemplateCompareNewDeclInfo(const DeclContext *DeclCtx,
+                               const DeclContext *LexicalDeclCtx,
+                               SourceLocation Loc)
+
+        : DC(DeclCtx), LexicalDC(LexicalDeclCtx), Loc(Loc) {
+      assert(DC && LexicalDC &&
+             "Constructor only for cases where we have the information to put "
+             "in here");
+    }
+
+    // If this was constructed with no information, we cannot do substitution
+    // for constraint comparison, so make sure we can check that.
+    bool isInvalid() const { return !ND && !DC; }
+
+    const NamedDecl *getDecl() const { return ND; }
+
+    bool ContainsDecl(const NamedDecl *ND) const { return this->ND == ND; }
+
+    const DeclContext *getLexicalDeclContext() const {
+      return ND ? ND->getLexicalDeclContext() : LexicalDC;
+    }
+
+    const DeclContext *getDeclContext() const {
+      return ND ? ND->getDeclContext() : DC;
+    }
+
+    SourceLocation getLocation() const { return ND ? ND->getLocation() : Loc; }
+  };
+
   bool TemplateParameterListsAreEqual(
-      const NamedDecl *NewInstFrom, TemplateParameterList *New,
+      const TemplateCompareNewDeclInfo &NewInstFrom, TemplateParameterList *New,
       const NamedDecl *OldInstFrom, TemplateParameterList *Old, bool Complain,
       TemplateParameterListEqualKind Kind,
       SourceLocation TemplateArgLoc = SourceLocation());
@@ -8629,6 +8658,17 @@ class Sema final {
                                           Kind, TemplateArgLoc);
   }
 
+  // Calculates whether two constraint expressions are equal irrespective of a
+  // difference in 'depth'. This takes a pair of optional 'NamedDecl's 'Old' and
+  // 'New', which are the "source" of the constraint, since this is necessary
+  // for figuring out the relative 'depth' of the constraint. The depth of the
+  // 'primary template' and the 'instantiated from' templates aren't necessarily
+  // the same, such as a case when one is a 'friend' defined in a class.
+  bool AreConstraintExpressionsEqual(const NamedDecl *Old,
+                                     const Expr *OldConstr,
+                                     const TemplateCompareNewDeclInfo &New,
+                                     const Expr *NewConstr);
+
   bool CheckTemplateDeclScope(Scope *S, TemplateParameterList *TemplateParams);
 
   /// Called when the parser has parsed a C++ typename
@@ -9368,13 +9408,12 @@ class Sema final {
   // C++ Template Instantiation
   //
 
-  MultiLevelTemplateArgumentList
-  getTemplateInstantiationArgs(const NamedDecl *D, bool Final = false,
-                               const TemplateArgumentList *Innermost = nullptr,
-                               bool RelativeToPrimary = false,
-                               const FunctionDecl *Pattern = nullptr,
-                               bool ForConstraintInstantiation = false,
-                               bool SkipForSpecialization = false);
+  MultiLevelTemplateArgumentList getTemplateInstantiationArgs(
+      const NamedDecl *D, const DeclContext *DC = nullptr, bool Final = false,
+      const TemplateArgumentList *Innermost = nullptr,
+      bool RelativeToPrimary = false, const FunctionDecl *Pattern = nullptr,
+      bool ForConstraintInstantiation = false,
+      bool SkipForSpecialization = false);
 
   /// A context in which code is being synthesized (where a source location
   /// alone is not sufficient to identify the context). This covers template

clang/include/clang/Sema/Template.h

@@ -213,7 +213,9 @@ enum class TemplateSubstitutionKind : char {
              "substituted args outside retained args?");
       assert(getKind() == TemplateSubstitutionKind::Specialization);
       TemplateArgumentLists.push_back(
-          {{AssociatedDecl->getCanonicalDecl(), Final}, Args});
+          {{AssociatedDecl ? AssociatedDecl->getCanonicalDecl() : nullptr,
+            Final},
+           Args});
     }
 
     void addOuterTemplateArguments(ArgList Args) {

clang/lib/CodeGen/CGBuiltin.cpp

@@ -10007,7 +10007,33 @@ Value *CodeGenFunction::EmitAArch64SVEBuiltinExpr(unsigned BuiltinID,
   switch (BuiltinID) {
   default:
     return nullptr;
-
+  case SVE::BI__builtin_sve_svpsel_lane_b8:
+  case SVE::BI__builtin_sve_svpsel_lane_b16:
+  case SVE::BI__builtin_sve_svpsel_lane_b32:
+  case SVE::BI__builtin_sve_svpsel_lane_b64:
+  case SVE::BI__builtin_sve_svpsel_lane_c8:
+  case SVE::BI__builtin_sve_svpsel_lane_c16:
+  case SVE::BI__builtin_sve_svpsel_lane_c32:
+  case SVE::BI__builtin_sve_svpsel_lane_c64: {
+    bool IsSVCount = isa<TargetExtType>(Ops[0]->getType());
+    assert(((!IsSVCount || cast<TargetExtType>(Ops[0]->getType())->getName() ==
+                               "aarch64.svcount")) &&
+           "Unexpected TargetExtType");
+    auto SVCountTy =
+        llvm::TargetExtType::get(getLLVMContext(), "aarch64.svcount");
+    Function *CastFromSVCountF =
+        CGM.getIntrinsic(Intrinsic::aarch64_sve_convert_to_svbool, SVCountTy);
+    Function *CastToSVCountF =
+        CGM.getIntrinsic(Intrinsic::aarch64_sve_convert_from_svbool, SVCountTy);
+
+    auto OverloadedTy = getSVEType(SVETypeFlags(Builtin->TypeModifier));
+    Function *F = CGM.getIntrinsic(Intrinsic::aarch64_sve_psel, OverloadedTy);
+    llvm::Value *Ops0 =
+        IsSVCount ? Builder.CreateCall(CastFromSVCountF, Ops[0]) : Ops[0];
+    llvm::Value *Ops1 = EmitSVEPredicateCast(Ops[1], OverloadedTy);
+    llvm::Value *PSel = Builder.CreateCall(F, {Ops0, Ops1, Ops[2]});
+    return IsSVCount ? Builder.CreateCall(CastToSVCountF, PSel) : PSel;
+  }
   case SVE::BI__builtin_sve_svmov_b_z: {
     // svmov_b_z(pg, op) <=> svand_b_z(pg, op, op)
     SVETypeFlags TypeFlags(Builtin->TypeModifier);