Sync to upstream/release/582 (#960)

* Optimized operations like instantiation and module export for very
large types

In our new typechecker:
* Typechecking of function calls was rewritten to handle more cases
correctly
* Fixed a crash that can happen after self-referential type is exported
from a module
* Fixed a false positive error in string comparison
* Added handling of `for...in` variable type annotations and fixed
issues with the iterator call inside
* Self-referential 'hasProp' and 'setProp' constraints are now handled
correctly
 
In our native code generation (jit):
* Added '--target' argument to luau-compile to test multiple
architectures different from host architecture
* GC barrier tag check is skipped if type is already known to be
GC-collectable
* Added GET_TYPE/GET_TYPEOF instructions for type/typeof fast-calls
* Improved code size of interrupt handlers on X64

Analysis/include/Luau/Constraint.h

@@ -83,7 +83,7 @@ struct IterableConstraint
     TypePackId variables;
 
     const AstNode* nextAstFragment;
-    DenseHashMap<const AstNode*, TypeId>* astOverloadResolvedTypes;
+    DenseHashMap<const AstNode*, TypeId>* astForInNextTypes;
 };
 
 // name(namedType) = name

Analysis/include/Luau/ConstraintSolver.h

@@ -245,6 +245,17 @@ private:
     template <typename TID>
     bool tryUnify(NotNull<const Constraint> constraint, TID subTy, TID superTy);
 
+    /**
+     * Bind a BlockedType to another type while taking care not to bind it to
+     * itself in the case that resultTy == blockedTy.  This can happen if we
+     * have a tautological constraint.  When it does, we must instead bind
+     * blockedTy to a fresh type belonging to an appropriate scope.
+     *
+     * To determine which scope is appropriate, we also accept rootTy, which is
+     * to be the type that contains blockedTy.
+     */
+    void bindBlockedType(TypeId blockedTy, TypeId resultTy, TypeId rootTy, Location location);
+
     /**
      * Marks a constraint as being blocked on a type or type pack. The constraint
      * solver will not attempt to dispatch blocked constraints until their

Analysis/include/Luau/Module.h

@@ -81,13 +81,29 @@ struct Module
     DenseHashMap<const AstExpr*, TypePackId> astTypePacks{nullptr};
     DenseHashMap<const AstExpr*, TypeId> astExpectedTypes{nullptr};
 
+    // For AST nodes that are function calls, this map provides the
+    // unspecialized type of the function that was called. If a function call
+    // resolves to a __call metamethod application, this map will point at that
+    // metamethod.
+    //
+    // This is useful for type checking and Signature Help.
     DenseHashMap<const AstNode*, TypeId> astOriginalCallTypes{nullptr};
+
+    // The specialization of a function that was selected.  If the function is
+    // generic, those generic type parameters will be replaced with the actual
+    // types that were passed.  If the function is an overload, this map will
+    // point at the specific overloads that were selected.
     DenseHashMap<const AstNode*, TypeId> astOverloadResolvedTypes{nullptr};
 
+    // Only used with for...in loops.  The computed type of the next() function
+    // is kept here for type checking.
+    DenseHashMap<const AstNode*, TypeId> astForInNextTypes{nullptr};
+
     DenseHashMap<const AstType*, TypeId> astResolvedTypes{nullptr};
     DenseHashMap<const AstTypePack*, TypePackId> astResolvedTypePacks{nullptr};
 
-    // Map AST nodes to the scope they create.  Cannot be NotNull<Scope> because we need a sentinel value for the map.
+    // Map AST nodes to the scope they create.  Cannot be NotNull<Scope> because
+    // we need a sentinel value for the map.
     DenseHashMap<const AstNode*, Scope*> astScopes{nullptr};
 
     std::unordered_map<Name, TypeId> declaredGlobals;
@@ -103,8 +119,9 @@ struct Module
     bool hasModuleScope() const;
     ScopePtr getModuleScope() const;
 
-    // Once a module has been typechecked, we clone its public interface into a separate arena.
+    // Once a module has been typechecked, we clone its public interface into a
-    // This helps us to force Type ownership into a DAG rather than a DCG.
+    // separate arena. This helps us to force Type ownership into a DAG rather
+    // than a DCG.
     void clonePublicInterface(NotNull<BuiltinTypes> builtinTypes, InternalErrorReporter& ice);
 };
 

Analysis/include/Luau/Normalize.h

@@ -207,8 +207,6 @@ struct NormalizedFunctionType
 struct NormalizedType;
 using NormalizedTyvars = std::unordered_map<TypeId, std::unique_ptr<NormalizedType>>;
 
-bool isInhabited_DEPRECATED(const NormalizedType& norm);
-
 // A normalized type is either any, unknown, or one of the form P | T | F | G where
 // * P is a union of primitive types (including singletons, classes and the error type)
 // * T is a union of table types

Analysis/include/Luau/Substitution.h

@@ -69,6 +69,19 @@ struct TarjanWorklistVertex
     int lastEdge;
 };
 
+struct TarjanNode
+{
+    TypeId ty;
+    TypePackId tp;
+
+    bool onStack;
+    bool dirty;
+
+    // Tarjan calculates the lowlink for each vertex,
+    // which is the lowest ancestor index reachable from the vertex.
+    int lowlink;
+};
+
 // Tarjan's algorithm for finding the SCCs in a cyclic structure.
 // https://en.wikipedia.org/wiki/Tarjan%27s_strongly_connected_components_algorithm
 struct Tarjan
@@ -76,17 +89,12 @@ struct Tarjan
     // Vertices (types and type packs) are indexed, using pre-order traversal.
     DenseHashMap<TypeId, int> typeToIndex{nullptr};
     DenseHashMap<TypePackId, int> packToIndex{nullptr};
-    std::vector<TypeId> indexToType;
+
-    std::vector<TypePackId> indexToPack;
+    std::vector<TarjanNode> nodes;
 
     // Tarjan keeps a stack of vertices where we're still in the process
     // of finding their SCC.
     std::vector<int> stack;
-    std::vector<bool> onStack;
-
-    // Tarjan calculates the lowlink for each vertex,
-    // which is the lowest ancestor index reachable from the vertex.
-    std::vector<int> lowlink;
 
     int childCount = 0;
     int childLimit = 0;
@@ -98,6 +106,7 @@ struct Tarjan
     std::vector<TypeId> edgesTy;
     std::vector<TypePackId> edgesTp;
     std::vector<TarjanWorklistVertex> worklist;
+
     // This is hot code, so we optimize recursion to a stack.
     TarjanResult loop();
 
@@ -124,10 +133,22 @@ struct Tarjan
     TarjanResult visitRoot(TypeId ty);
     TarjanResult visitRoot(TypePackId ty);
 
-    // Each subclass gets called back once for each edge,
+    void clearTarjan();
-    // and once for each SCC.
+
-    virtual void visitEdge(int index, int parentIndex) {}
+    // Get/set the dirty bit for an index (grows the vector if needed)
-    virtual void visitSCC(int index) {}
+    bool getDirty(int index);
+    void setDirty(int index, bool d);
+
+    // Find all the dirty vertices reachable from `t`.
+    TarjanResult findDirty(TypeId t);
+    TarjanResult findDirty(TypePackId t);
+
+    // We find dirty vertices using Tarjan
+    void visitEdge(int index, int parentIndex);
+    void visitSCC(int index);
+
+    TarjanResult loop_DEPRECATED();
+    void visitSCC_DEPRECATED(int index);
 
     // Each subclass can decide to ignore some nodes.
     virtual bool ignoreChildren(TypeId ty)
@@ -150,27 +171,6 @@ struct Tarjan
     {
         return ignoreChildren(ty);
     }
-};
-
-// We use Tarjan to calculate dirty bits. We set `dirty[i]` true
-// if the vertex with index `i` can reach a dirty vertex.
-struct FindDirty : Tarjan
-{
-    std::vector<bool> dirty;
-
-    void clearTarjan();
-
-    // Get/set the dirty bit for an index (grows the vector if needed)
-    bool getDirty(int index);
-    void setDirty(int index, bool d);
-
-    // Find all the dirty vertices reachable from `t`.
-    TarjanResult findDirty(TypeId t);
-    TarjanResult findDirty(TypePackId t);
-
-    // We find dirty vertices using Tarjan
-    void visitEdge(int index, int parentIndex) override;
-    void visitSCC(int index) override;
 
     // Subclasses should say which vertices are dirty,
     // and what to do with dirty vertices.
@@ -178,11 +178,18 @@ struct FindDirty : Tarjan
     virtual bool isDirty(TypePackId tp) = 0;
     virtual void foundDirty(TypeId ty) = 0;
     virtual void foundDirty(TypePackId tp) = 0;
+
+    // TODO: remove with FFlagLuauTarjanSingleArr
+    std::vector<TypeId> indexToType;
+    std::vector<TypePackId> indexToPack;
+    std::vector<bool> onStack;
+    std::vector<int> lowlink;
+    std::vector<bool> dirty;
 };
 
 // And finally substitution, which finds all the reachable dirty vertices
 // and replaces them with clean ones.
-struct Substitution : FindDirty
+struct Substitution : Tarjan
 {
 protected:
     Substitution(const TxnLog* log_, TypeArena* arena)

Analysis/include/Luau/TypeInfer.h

@@ -19,8 +19,6 @@
 #include <unordered_map>
 #include <unordered_set>
 
-LUAU_FASTFLAG(LuauClassTypeVarsInSubstitution)
-
 namespace Luau
 {
 

Analysis/include/Luau/TypeUtils.h

@@ -64,4 +64,13 @@ const T* get(std::optional<Ty> ty)
         return nullptr;
 }
 
+template<typename Ty>
+std::optional<Ty> follow(std::optional<Ty> ty)
+{
+    if (ty)
+        return follow(*ty);
+    else
+        return std::nullopt;
+}
+
 } // namespace Luau

Analysis/src/Anyification.cpp

@@ -6,8 +6,6 @@
 #include "Luau/Normalize.h"
 #include "Luau/TxnLog.h"
 
-LUAU_FASTFLAG(LuauClassTypeVarsInSubstitution)
-
 namespace Luau
 {
 
@@ -78,7 +76,7 @@ TypePackId Anyification::clean(TypePackId tp)
 
 bool Anyification::ignoreChildren(TypeId ty)
 {
-    if (FFlag::LuauClassTypeVarsInSubstitution && get<ClassType>(ty))
+    if (get<ClassType>(ty))
         return true;
 
     return ty->persistent;

Analysis/src/ApplyTypeFunction.cpp

@@ -2,8 +2,6 @@
 
 #include "Luau/ApplyTypeFunction.h"
 
-LUAU_FASTFLAG(LuauClassTypeVarsInSubstitution)
-
 namespace Luau
 {
 
@@ -33,7 +31,7 @@ bool ApplyTypeFunction::ignoreChildren(TypeId ty)
 {
     if (get<GenericType>(ty))
         return true;
-    else if (FFlag::LuauClassTypeVarsInSubstitution && get<ClassType>(ty))
+    else if (get<ClassType>(ty))
         return true;
     else
         return false;

Analysis/src/ConstraintGraphBuilder.cpp

@@ -751,7 +751,12 @@ ControlFlow ConstraintGraphBuilder::visit(const ScopePtr& scope, AstStatForIn* f
     variableTypes.reserve(forIn->vars.size);
     for (AstLocal* var : forIn->vars)
     {
-        TypeId ty = freshType(loopScope);
+        TypeId ty = nullptr;
+        if (var->annotation)
+            ty = resolveType(loopScope, var->annotation, /*inTypeArguments*/ false);
+        else
+            ty = freshType(loopScope);
+
         loopScope->bindings[var] = Binding{ty, var->location};
         variableTypes.push_back(ty);
 
@@ -763,7 +768,7 @@ ControlFlow ConstraintGraphBuilder::visit(const ScopePtr& scope, AstStatForIn* f
     TypePackId variablePack = arena->addTypePack(std::move(variableTypes), arena->addTypePack(FreeTypePack{loopScope.get()}));
 
     addConstraint(
-        loopScope, getLocation(forIn->values), IterableConstraint{iterator, variablePack, forIn->values.data[0], &module->astOverloadResolvedTypes});
+        loopScope, getLocation(forIn->values), IterableConstraint{iterator, variablePack, forIn->values.data[0], &module->astForInNextTypes});
     visit(loopScope, forIn->body);
 
     return ControlFlow::None;

Analysis/src/ConstraintSolver.cpp

@@ -1453,7 +1453,7 @@ bool ConstraintSolver::tryDispatch(const HasPropConstraint& c, NotNull<const Con
         return false;
     }
 
-    asMutable(c.resultType)->ty.emplace<BoundType>(result.value_or(builtinTypes->anyType));
+    bindBlockedType(c.resultType, result.value_or(builtinTypes->anyType), c.subjectType, constraint->location);
     unblock(c.resultType, constraint->location);
     return true;
 }
@@ -1559,8 +1559,8 @@ bool ConstraintSolver::tryDispatch(const SetPropConstraint& c, NotNull<const Con
         existingPropType = result;
     }
 
-    auto bind = [](TypeId a, TypeId b) {
+    auto bind = [&](TypeId a, TypeId b) {
-        asMutable(a)->ty.emplace<BoundType>(b);
+        bindBlockedType(a, b, c.subjectType, constraint->location);
     };
 
     if (existingPropType)
@@ -2143,7 +2143,9 @@ bool ConstraintSolver::tryDispatchIterableFunction(
 
     // if there are no errors from unifying the two, we can pass forward the expected type as our selected resolution.
     if (errors.empty())
-        (*c.astOverloadResolvedTypes)[c.nextAstFragment] = expectedNextTy;
+    {
+        (*c.astForInNextTypes)[c.nextAstFragment] = expectedNextTy;
+    }
 
     auto it = begin(nextRetPack);
     std::vector<TypeId> modifiedNextRetHead;
@@ -2380,6 +2382,31 @@ bool ConstraintSolver::tryUnify(NotNull<const Constraint> constraint, TID subTy,
     return true;
 }
 
+void ConstraintSolver::bindBlockedType(TypeId blockedTy, TypeId resultTy, TypeId rootTy, Location location)
+{
+    resultTy = follow(resultTy);
+
+    LUAU_ASSERT(get<BlockedType>(blockedTy));
+
+    if (blockedTy == resultTy)
+    {
+        rootTy = follow(rootTy);
+        Scope* freeScope = nullptr;
+        if (auto ft = get<FreeType>(rootTy))
+            freeScope = ft->scope;
+        else if (auto tt = get<TableType>(rootTy); tt && tt->state == TableState::Free)
+            freeScope = tt->scope;
+        else
+            iceReporter.ice("bindBlockedType couldn't find an appropriate scope for a fresh type!", location);
+
+        LUAU_ASSERT(freeScope);
+
+        asMutable(blockedTy)->ty.emplace<BoundType>(arena->freshType(freeScope));
+    }
+    else
+        asMutable(blockedTy)->ty.emplace<BoundType>(resultTy);
+}
+
 void ConstraintSolver::block_(BlockedConstraintId target, NotNull<const Constraint> constraint)
 {
     blocked[target].push_back(constraint);

Analysis/src/Error.cpp

@@ -10,7 +10,6 @@
 #include <stdexcept>
 #include <type_traits>
 
-LUAU_FASTFLAGVARIABLE(LuauTypeMismatchInvarianceInError, false)
 
 static std::string wrongNumberOfArgsString(
     size_t expectedCount, std::optional<size_t> maximumCount, size_t actualCount, const char* argPrefix = nullptr, bool isVariadic = false)
@@ -106,7 +105,7 @@ struct ErrorConverter
         {
             result += "; " + tm.reason;
         }
-        else if (FFlag::LuauTypeMismatchInvarianceInError && tm.context == TypeMismatch::InvariantContext)
+        else if (tm.context == TypeMismatch::InvariantContext)
         {
             result += " in an invariant context";
         }

Analysis/src/Frontend.cpp

@@ -950,6 +950,7 @@ void Frontend::checkBuildQueueItem(BuildQueueItem& item)
         module->astExpectedTypes.clear();
         module->astOriginalCallTypes.clear();
         module->astOverloadResolvedTypes.clear();
+        module->astForInNextTypes.clear();
         module->astResolvedTypes.clear();
         module->astResolvedTypePacks.clear();
         module->astScopes.clear();

Analysis/src/Instantiation.cpp

@@ -4,8 +4,6 @@
 #include "Luau/TxnLog.h"
 #include "Luau/TypeArena.h"
 
-LUAU_FASTFLAG(LuauClassTypeVarsInSubstitution)
-
 namespace Luau
 {
 
@@ -33,7 +31,7 @@ bool Instantiation::ignoreChildren(TypeId ty)
 {
     if (log->getMutable<FunctionType>(ty))
         return true;
-    else if (FFlag::LuauClassTypeVarsInSubstitution && get<ClassType>(ty))
+    else if (get<ClassType>(ty))
         return true;
     else
         return false;
@@ -84,7 +82,7 @@ bool ReplaceGenerics::ignoreChildren(TypeId ty)
         // whenever we quantify, so the vectors overlap if and only if they are equal.
         return (!generics.empty() || !genericPacks.empty()) && (ftv->generics == generics) && (ftv->genericPacks == genericPacks);
     }
-    else if (FFlag::LuauClassTypeVarsInSubstitution && get<ClassType>(ty))
+    else if (get<ClassType>(ty))
         return true;
     else
     {

Analysis/src/Module.cpp

@@ -16,9 +16,6 @@
 
 LUAU_FASTFLAG(DebugLuauDeferredConstraintResolution);
 LUAU_FASTFLAGVARIABLE(LuauClonePublicInterfaceLess2, false);
-LUAU_FASTFLAG(LuauSubstitutionReentrant);
-LUAU_FASTFLAG(LuauClassTypeVarsInSubstitution);
-LUAU_FASTFLAG(LuauSubstitutionFixMissingFields);
 LUAU_FASTFLAGVARIABLE(LuauCloneSkipNonInternalVisit, false);
 
 namespace Luau
@@ -134,8 +131,6 @@ struct ClonePublicInterface : Substitution
 
     TypeId cloneType(TypeId ty)
     {
-        LUAU_ASSERT(FFlag::LuauSubstitutionReentrant && FFlag::LuauSubstitutionFixMissingFields);
-
         std::optional<TypeId> result = substitute(ty);
         if (result)
         {
@@ -150,8 +145,6 @@ struct ClonePublicInterface : Substitution
 
     TypePackId cloneTypePack(TypePackId tp)
     {
-        LUAU_ASSERT(FFlag::LuauSubstitutionReentrant && FFlag::LuauSubstitutionFixMissingFields);
-
         std::optional<TypePackId> result = substitute(tp);
         if (result)
         {
@@ -166,8 +159,6 @@ struct ClonePublicInterface : Substitution
 
     TypeFun cloneTypeFun(const TypeFun& tf)
     {
-        LUAU_ASSERT(FFlag::LuauSubstitutionReentrant && FFlag::LuauSubstitutionFixMissingFields);
-
         std::vector<GenericTypeDefinition> typeParams;
         std::vector<GenericTypePackDefinition> typePackParams;
 

Analysis/src/Normalize.cpp

@@ -19,7 +19,6 @@ LUAU_FASTINTVARIABLE(LuauNormalizeIterationLimit, 1200);
 LUAU_FASTINTVARIABLE(LuauNormalizeCacheLimit, 100000);
 LUAU_FASTFLAGVARIABLE(LuauNormalizeBlockedTypes, false);
 LUAU_FASTFLAG(DebugLuauDeferredConstraintResolution)
-LUAU_FASTFLAG(LuauUninhabitedSubAnything2)
 LUAU_FASTFLAG(LuauTransitiveSubtyping)
 LUAU_FASTFLAG(DebugLuauReadWriteProperties)
 
@@ -312,12 +311,6 @@ static bool isShallowInhabited(const NormalizedType& norm)
            !norm.functions.isNever() || !norm.tables.empty() || !norm.tyvars.empty();
 }
 
-bool isInhabited_DEPRECATED(const NormalizedType& norm)
-{
-    LUAU_ASSERT(!FFlag::LuauUninhabitedSubAnything2);
-    return isShallowInhabited(norm);
-}
-
 bool Normalizer::isInhabited(const NormalizedType* norm, std::unordered_set<TypeId> seen)
 {
     // If normalization failed, the type is complex, and so is more likely than not to be inhabited.

Analysis/src/Quantify.cpp

@@ -10,7 +10,6 @@
 
 LUAU_FASTFLAG(DebugLuauSharedSelf)
 LUAU_FASTFLAG(DebugLuauDeferredConstraintResolution);
-LUAU_FASTFLAG(LuauClassTypeVarsInSubstitution)
 
 namespace Luau
 {
@@ -244,7 +243,7 @@ struct PureQuantifier : Substitution
 
     bool ignoreChildren(TypeId ty) override
     {
-        if (FFlag::LuauClassTypeVarsInSubstitution && get<ClassType>(ty))
+        if (get<ClassType>(ty))
             return true;
 
         return ty->persistent;

Analysis/src/Substitution.cpp

@@ -8,13 +8,11 @@
 #include <algorithm>
 #include <stdexcept>
 
-LUAU_FASTFLAGVARIABLE(LuauSubstitutionFixMissingFields, false)
 LUAU_FASTFLAG(LuauClonePublicInterfaceLess2)
 LUAU_FASTINTVARIABLE(LuauTarjanChildLimit, 10000)
-LUAU_FASTFLAGVARIABLE(LuauClassTypeVarsInSubstitution, false)
-LUAU_FASTFLAGVARIABLE(LuauSubstitutionReentrant, false)
 LUAU_FASTFLAG(DebugLuauReadWriteProperties)
 LUAU_FASTFLAG(LuauCloneSkipNonInternalVisit)
+LUAU_FASTFLAGVARIABLE(LuauTarjanSingleArr, false)
 
 namespace Luau
 {
@@ -113,20 +111,35 @@ static TypeId shallowClone(TypeId ty, TypeArena& dest, const TxnLog* log, bool a
         else if constexpr (std::is_same_v<T, BlockedType>)
             return dest.addType(a);
         else if constexpr (std::is_same_v<T, PrimitiveType>)
+        {
+            LUAU_ASSERT(ty->persistent);
             return ty;
+        }
         else if constexpr (std::is_same_v<T, PendingExpansionType>)
         {
             PendingExpansionType clone = PendingExpansionType{a.prefix, a.name, a.typeArguments, a.packArguments};
             return dest.addType(std::move(clone));
         }
         else if constexpr (std::is_same_v<T, AnyType>)
+        {
+            LUAU_ASSERT(ty->persistent);
             return ty;
+        }
         else if constexpr (std::is_same_v<T, ErrorType>)
+        {
+            LUAU_ASSERT(ty->persistent);
             return ty;
+        }
         else if constexpr (std::is_same_v<T, UnknownType>)
+        {
+            LUAU_ASSERT(ty->persistent);
             return ty;
+        }
         else if constexpr (std::is_same_v<T, NeverType>)
+        {
+            LUAU_ASSERT(ty->persistent);
             return ty;
+        }
         else if constexpr (std::is_same_v<T, LazyType>)
             return ty;
         else if constexpr (std::is_same_v<T, SingletonType>)
@@ -227,13 +240,10 @@ void Tarjan::visitChildren(TypeId ty, int index)
 
     if (const FunctionType* ftv = get<FunctionType>(ty))
     {
-        if (FFlag::LuauSubstitutionFixMissingFields)
+        for (TypeId generic : ftv->generics)
-        {
+            visitChild(generic);
-            for (TypeId generic : ftv->generics)
+        for (TypePackId genericPack : ftv->genericPacks)
-                visitChild(generic);
+            visitChild(genericPack);
-            for (TypePackId genericPack : ftv->genericPacks)
-                visitChild(genericPack);
-        }
 
         visitChild(ftv->argTypes);
         visitChild(ftv->retTypes);
@@ -295,7 +305,7 @@ void Tarjan::visitChildren(TypeId ty, int index)
         for (TypePackId a : tfit->packArguments)
             visitChild(a);
     }
-    else if (const ClassType* ctv = get<ClassType>(ty); FFlag::LuauClassTypeVarsInSubstitution && ctv)
+    else if (const ClassType* ctv = get<ClassType>(ty))
     {
         for (const auto& [name, prop] : ctv->props)
             visitChild(prop.type());
@@ -348,36 +358,67 @@ std::pair<int, bool> Tarjan::indexify(TypeId ty)
 {
     ty = log->follow(ty);
 
-    bool fresh = !typeToIndex.contains(ty);
+    if (FFlag::LuauTarjanSingleArr)
-    int& index = typeToIndex[ty];
-
-    if (fresh)
     {
-        index = int(indexToType.size());
+        auto [index, fresh] = typeToIndex.try_insert(ty, false);
-        indexToType.push_back(ty);
+
-        indexToPack.push_back(nullptr);
+        if (fresh)
-        onStack.push_back(false);
+        {
-        lowlink.push_back(index);
+            index = int(nodes.size());
+            nodes.push_back({ty, nullptr, false, false, index});
+        }
+
+        return {index, fresh};
+    }
+    else
+    {
+        bool fresh = !typeToIndex.contains(ty);
+        int& index = typeToIndex[ty];
+
+        if (fresh)
+        {
+            index = int(indexToType.size());
+            indexToType.push_back(ty);
+            indexToPack.push_back(nullptr);
+            onStack.push_back(false);
+            lowlink.push_back(index);
+        }
+        return {index, fresh};
     }
-    return {index, fresh};
 }
 
 std::pair<int, bool> Tarjan::indexify(TypePackId tp)
 {
     tp = log->follow(tp);
 
-    bool fresh = !packToIndex.contains(tp);
+    if (FFlag::LuauTarjanSingleArr)
-    int& index = packToIndex[tp];
-
-    if (fresh)
     {
-        index = int(indexToPack.size());
+        auto [index, fresh] = packToIndex.try_insert(tp, false);
-        indexToType.push_back(nullptr);
+
-        indexToPack.push_back(tp);
+        if (fresh)
-        onStack.push_back(false);
+        {
-        lowlink.push_back(index);
+            index = int(nodes.size());
+            nodes.push_back({nullptr, tp, false, false, index});
+        }
+
+        return {index, fresh};
+    }
+    else
+    {
+
+        bool fresh = !packToIndex.contains(tp);
+        int& index = packToIndex[tp];
+
+        if (fresh)
+        {
+            index = int(indexToPack.size());
+            indexToType.push_back(nullptr);
+            indexToPack.push_back(tp);
+            onStack.push_back(false);
+            lowlink.push_back(index);
+        }
+        return {index, fresh};
     }
-    return {index, fresh};
 }
 
 void Tarjan::visitChild(TypeId ty)
@@ -397,6 +438,246 @@ void Tarjan::visitChild(TypePackId tp)
 }
 
 TarjanResult Tarjan::loop()
+{
+    if (!FFlag::LuauTarjanSingleArr)
+        return loop_DEPRECATED();
+
+    // Normally Tarjan is presented recursively, but this is a hot loop, so worth optimizing
+    while (!worklist.empty())
+    {
+        auto [index, currEdge, lastEdge] = worklist.back();
+
+        // First visit
+        if (currEdge == -1)
+        {
+            ++childCount;
+            if (childLimit > 0 && childLimit <= childCount)
+                return TarjanResult::TooManyChildren;
+
+            stack.push_back(index);
+
+            nodes[index].onStack = true;
+
+            currEdge = int(edgesTy.size());
+
+            // Fill in edge list of this vertex
+            if (TypeId ty = nodes[index].ty)
+                visitChildren(ty, index);
+            else if (TypePackId tp = nodes[index].tp)
+                visitChildren(tp, index);
+
+            lastEdge = int(edgesTy.size());
+        }
+
+        // Visit children
+        bool foundFresh = false;
+
+        for (; currEdge < lastEdge; currEdge++)
+        {
+            int childIndex = -1;
+            bool fresh = false;
+
+            if (auto ty = edgesTy[currEdge])
+                std::tie(childIndex, fresh) = indexify(ty);
+            else if (auto tp = edgesTp[currEdge])
+                std::tie(childIndex, fresh) = indexify(tp);
+            else
+                LUAU_ASSERT(false);
+
+            if (fresh)
+            {
+                // Original recursion point, update the parent continuation point and start the new element
+                worklist.back() = {index, currEdge + 1, lastEdge};
+                worklist.push_back({childIndex, -1, -1});
+
+                // We need to continue the top-level loop from the start with the new worklist element
+                foundFresh = true;
+                break;
+            }
+            else if (nodes[childIndex].onStack)
+            {
+                nodes[index].lowlink = std::min(nodes[index].lowlink, childIndex);
+            }
+
+            visitEdge(childIndex, index);
+        }
+
+        if (foundFresh)
+            continue;
+
+        if (nodes[index].lowlink == index)
+        {
+            visitSCC(index);
+            while (!stack.empty())
+            {
+                int popped = stack.back();
+                stack.pop_back();
+                nodes[popped].onStack = false;
+                if (popped == index)
+                    break;
+            }
+        }
+
+        worklist.pop_back();
+
+        // Original return from recursion into a child
+        if (!worklist.empty())
+        {
+            auto [parentIndex, _, parentEndEdge] = worklist.back();
+
+            // No need to keep child edges around
+            edgesTy.resize(parentEndEdge);
+            edgesTp.resize(parentEndEdge);
+
+            nodes[parentIndex].lowlink = std::min(nodes[parentIndex].lowlink, nodes[index].lowlink);
+            visitEdge(index, parentIndex);
+        }
+    }
+
+    return TarjanResult::Ok;
+}
+
+TarjanResult Tarjan::visitRoot(TypeId ty)
+{
+    childCount = 0;
+    if (childLimit == 0)
+        childLimit = FInt::LuauTarjanChildLimit;
+
+    ty = log->follow(ty);
+
+    auto [index, fresh] = indexify(ty);
+    worklist.push_back({index, -1, -1});
+    return loop();
+}
+
+TarjanResult Tarjan::visitRoot(TypePackId tp)
+{
+    childCount = 0;
+    if (childLimit == 0)
+        childLimit = FInt::LuauTarjanChildLimit;
+
+    tp = log->follow(tp);
+
+    auto [index, fresh] = indexify(tp);
+    worklist.push_back({index, -1, -1});
+    return loop();
+}
+
+void Tarjan::clearTarjan()
+{
+    if (FFlag::LuauTarjanSingleArr)
+    {
+        typeToIndex.clear();
+        packToIndex.clear();
+        nodes.clear();
+
+        stack.clear();
+    }
+    else
+    {
+        dirty.clear();
+
+        typeToIndex.clear();
+        packToIndex.clear();
+        indexToType.clear();
+        indexToPack.clear();
+
+        stack.clear();
+        onStack.clear();
+        lowlink.clear();
+    }
+
+    edgesTy.clear();
+    edgesTp.clear();
+    worklist.clear();
+}
+
+bool Tarjan::getDirty(int index)
+{
+    if (FFlag::LuauTarjanSingleArr)
+    {
+        LUAU_ASSERT(size_t(index) < nodes.size());
+        return nodes[index].dirty;
+    }
+    else
+    {
+        if (dirty.size() <= size_t(index))
+            dirty.resize(index + 1, false);
+        return dirty[index];
+    }
+}
+
+void Tarjan::setDirty(int index, bool d)
+{
+    if (FFlag::LuauTarjanSingleArr)
+    {
+        LUAU_ASSERT(size_t(index) < nodes.size());
+        nodes[index].dirty = d;
+    }
+    else
+    {
+        if (dirty.size() <= size_t(index))
+            dirty.resize(index + 1, false);
+        dirty[index] = d;
+    }
+}
+
+void Tarjan::visitEdge(int index, int parentIndex)
+{
+    if (getDirty(index))
+        setDirty(parentIndex, true);
+}
+
+void Tarjan::visitSCC(int index)
+{
+    if (!FFlag::LuauTarjanSingleArr)
+        return visitSCC_DEPRECATED(index);
+
+    bool d = getDirty(index);
+
+    for (auto it = stack.rbegin(); !d && it != stack.rend(); it++)
+    {
+        TarjanNode& node = nodes[*it];
+
+        if (TypeId ty = node.ty)
+            d = isDirty(ty);
+        else if (TypePackId tp = node.tp)
+            d = isDirty(tp);
+
+        if (*it == index)
+            break;
+    }
+
+    if (!d)
+        return;
+
+    for (auto it = stack.rbegin(); it != stack.rend(); it++)
+    {
+        setDirty(*it, true);
+
+        TarjanNode& node = nodes[*it];
+
+        if (TypeId ty = node.ty)
+            foundDirty(ty);
+        else if (TypePackId tp = node.tp)
+            foundDirty(tp);
+
+        if (*it == index)
+            return;
+    }
+}
+
+TarjanResult Tarjan::findDirty(TypeId ty)
+{
+    return visitRoot(ty);
+}
+
+TarjanResult Tarjan::findDirty(TypePackId tp)
+{
+    return visitRoot(tp);
+}
+
+TarjanResult Tarjan::loop_DEPRECATED()
 {
     // Normally Tarjan is presented recursively, but this is a hot loop, so worth optimizing
     while (!worklist.empty())
@@ -492,71 +773,8 @@ TarjanResult Tarjan::loop()
     return TarjanResult::Ok;
 }
 
-TarjanResult Tarjan::visitRoot(TypeId ty)
-{
-    childCount = 0;
-    if (childLimit == 0)
-        childLimit = FInt::LuauTarjanChildLimit;
 
-    ty = log->follow(ty);
+void Tarjan::visitSCC_DEPRECATED(int index)
-
-    auto [index, fresh] = indexify(ty);
-    worklist.push_back({index, -1, -1});
-    return loop();
-}
-
-TarjanResult Tarjan::visitRoot(TypePackId tp)
-{
-    childCount = 0;
-    if (childLimit == 0)
-        childLimit = FInt::LuauTarjanChildLimit;
-
-    tp = log->follow(tp);
-
-    auto [index, fresh] = indexify(tp);
-    worklist.push_back({index, -1, -1});
-    return loop();
-}
-
-void FindDirty::clearTarjan()
-{
-    dirty.clear();
-
-    typeToIndex.clear();
-    packToIndex.clear();
-    indexToType.clear();
-    indexToPack.clear();
-
-    stack.clear();
-    onStack.clear();
-    lowlink.clear();
-
-    edgesTy.clear();
-    edgesTp.clear();
-    worklist.clear();
-}
-
-bool FindDirty::getDirty(int index)
-{
-    if (dirty.size() <= size_t(index))
-        dirty.resize(index + 1, false);
-    return dirty[index];
-}
-
-void FindDirty::setDirty(int index, bool d)
-{
-    if (dirty.size() <= size_t(index))
-        dirty.resize(index + 1, false);
-    dirty[index] = d;
-}
-
-void FindDirty::visitEdge(int index, int parentIndex)
-{
-    if (getDirty(index))
-        setDirty(parentIndex, true);
-}
-
-void FindDirty::visitSCC(int index)
 {
     bool d = getDirty(index);
 
@@ -585,23 +803,12 @@ void FindDirty::visitSCC(int index)
     }
 }
 
-TarjanResult FindDirty::findDirty(TypeId ty)
-{
-    return visitRoot(ty);
-}
-
-TarjanResult FindDirty::findDirty(TypePackId tp)
-{
-    return visitRoot(tp);
-}
-
 std::optional<TypeId> Substitution::substitute(TypeId ty)
 {
     ty = log->follow(ty);
 
     // clear algorithm state for reentrancy
-    if (FFlag::LuauSubstitutionReentrant)
+    clearTarjan();
-        clearTarjan();
 
     auto result = findDirty(ty);
     if (result != TarjanResult::Ok)
@@ -609,34 +816,18 @@ std::optional<TypeId> Substitution::substitute(TypeId ty)
 
     for (auto [oldTy, newTy] : newTypes)
     {
-        if (FFlag::LuauSubstitutionReentrant)
+        if (!ignoreChildren(oldTy) && !replacedTypes.contains(newTy))
         {
-            if (!ignoreChildren(oldTy) && !replacedTypes.contains(newTy))
+            replaceChildren(newTy);
-            {
+            replacedTypes.insert(newTy);
-                replaceChildren(newTy);
-                replacedTypes.insert(newTy);
-            }
-        }
-        else
-        {
-            if (!ignoreChildren(oldTy))
-                replaceChildren(newTy);
         }
     }
     for (auto [oldTp, newTp] : newPacks)
     {
-        if (FFlag::LuauSubstitutionReentrant)
+        if (!ignoreChildren(oldTp) && !replacedTypePacks.contains(newTp))
         {
-            if (!ignoreChildren(oldTp) && !replacedTypePacks.contains(newTp))
+            replaceChildren(newTp);
-            {
+            replacedTypePacks.insert(newTp);
-                replaceChildren(newTp);
-                replacedTypePacks.insert(newTp);
-            }
-        }
-        else
-        {
-            if (!ignoreChildren(oldTp))
-                replaceChildren(newTp);
         }
     }
     TypeId newTy = replace(ty);
@@ -648,8 +839,7 @@ std::optional<TypePackId> Substitution::substitute(TypePackId tp)
     tp = log->follow(tp);
 
     // clear algorithm state for reentrancy
-    if (FFlag::LuauSubstitutionReentrant)
+    clearTarjan();
-        clearTarjan();
 
     auto result = findDirty(tp);
     if (result != TarjanResult::Ok)
@@ -657,34 +847,18 @@ std::optional<TypePackId> Substitution::substitute(TypePackId tp)
 
     for (auto [oldTy, newTy] : newTypes)
     {
-        if (FFlag::LuauSubstitutionReentrant)
+        if (!ignoreChildren(oldTy) && !replacedTypes.contains(newTy))
         {
-            if (!ignoreChildren(oldTy) && !replacedTypes.contains(newTy))
+            replaceChildren(newTy);
-            {
+            replacedTypes.insert(newTy);
-                replaceChildren(newTy);
-                replacedTypes.insert(newTy);
-            }
-        }
-        else
-        {
-            if (!ignoreChildren(oldTy))
-                replaceChildren(newTy);
         }
     }
     for (auto [oldTp, newTp] : newPacks)
     {
-        if (FFlag::LuauSubstitutionReentrant)
+        if (!ignoreChildren(oldTp) && !replacedTypePacks.contains(newTp))
         {
-            if (!ignoreChildren(oldTp) && !replacedTypePacks.contains(newTp))
+            replaceChildren(newTp);
-            {
+            replacedTypePacks.insert(newTp);
-                replaceChildren(newTp);
-                replacedTypePacks.insert(newTp);
-            }
-        }
-        else
-        {
-            if (!ignoreChildren(oldTp))
-                replaceChildren(newTp);
         }
     }
     TypePackId newTp = replace(tp);
@@ -714,8 +888,7 @@ TypePackId Substitution::clone(TypePackId tp)
     {
         VariadicTypePack clone;
         clone.ty = vtp->ty;
-        if (FFlag::LuauSubstitutionFixMissingFields)
+        clone.hidden = vtp->hidden;
-            clone.hidden = vtp->hidden;
         return addTypePack(std::move(clone));
     }
     else if (const TypeFamilyInstanceTypePack* tfitp = get<TypeFamilyInstanceTypePack>(tp))
@@ -738,7 +911,7 @@ void Substitution::foundDirty(TypeId ty)
 {
     ty = log->follow(ty);
 
-    if (FFlag::LuauSubstitutionReentrant && newTypes.contains(ty))
+    if (newTypes.contains(ty))
         return;
 
     if (isDirty(ty))
@@ -751,7 +924,7 @@ void Substitution::foundDirty(TypePackId tp)
 {
     tp = log->follow(tp);
 
-    if (FFlag::LuauSubstitutionReentrant && newPacks.contains(tp))
+    if (newPacks.contains(tp))
         return;
 
     if (isDirty(tp))
@@ -792,13 +965,10 @@ void Substitution::replaceChildren(TypeId ty)
 
     if (FunctionType* ftv = getMutable<FunctionType>(ty))
     {
-        if (FFlag::LuauSubstitutionFixMissingFields)
+        for (TypeId& generic : ftv->generics)
-        {
+            generic = replace(generic);
-            for (TypeId& generic : ftv->generics)
+        for (TypePackId& genericPack : ftv->genericPacks)
-                generic = replace(generic);
+            genericPack = replace(genericPack);
-            for (TypePackId& genericPack : ftv->genericPacks)
-                genericPack = replace(genericPack);
-        }
 
         ftv->argTypes = replace(ftv->argTypes);
         ftv->retTypes = replace(ftv->retTypes);
@@ -857,7 +1027,7 @@ void Substitution::replaceChildren(TypeId ty)
         for (TypePackId& a : tfit->packArguments)
             a = replace(a);
     }
-    else if (ClassType* ctv = getMutable<ClassType>(ty); FFlag::LuauClassTypeVarsInSubstitution && ctv)
+    else if (ClassType* ctv = getMutable<ClassType>(ty))
     {
         for (auto& [name, prop] : ctv->props)
             prop.setType(replace(prop.type()));

Analysis/src/TypeChecker2.cpp

@@ -7,6 +7,7 @@
 #include "Luau/Common.h"
 #include "Luau/DcrLogger.h"
 #include "Luau/Error.h"
+#include "Luau/InsertionOrderedMap.h"
 #include "Luau/Instantiation.h"
 #include "Luau/Metamethods.h"
 #include "Luau/Normalize.h"
@@ -656,7 +657,7 @@ struct TypeChecker2
         // if the initial and expected types from the iterator unified during constraint solving,
         // we'll have a resolved type to use here, but we'll only use it if either the iterator is
         // directly present in the for-in statement or if we have an iterator state constraining us
-        TypeId* resolvedTy = module->astOverloadResolvedTypes.find(firstValue);
+        TypeId* resolvedTy = module->astForInNextTypes.find(firstValue);
         if (resolvedTy && (!retPack || valueTypes.size() > 1))
             valueTypes[0] = *resolvedTy;
 
@@ -1062,83 +1063,21 @@ struct TypeChecker2
     // Note: this is intentionally separated from `visit(AstExprCall*)` for stack allocation purposes.
     void visitCall(AstExprCall* call)
     {
-        TypePackId expectedRetType = lookupExpectedPack(call, testArena);
         TypePack args;
         std::vector<Location> argLocs;
         argLocs.reserve(call->args.size + 1);
 
-        auto maybeOriginalCallTy = module->astOriginalCallTypes.find(call);
+        TypeId* originalCallTy = module->astOriginalCallTypes.find(call);
-        if (!maybeOriginalCallTy)
+        TypeId* selectedOverloadTy = module->astOverloadResolvedTypes.find(call);
+        if (!originalCallTy && !selectedOverloadTy)
             return;
 
-        TypeId originalCallTy = follow(*maybeOriginalCallTy);
+        TypeId fnTy = follow(selectedOverloadTy ? *selectedOverloadTy : *originalCallTy);
-        std::vector<TypeId> overloads = flattenIntersection(originalCallTy);
+        if (get<AnyType>(fnTy) || get<ErrorType>(fnTy) || get<NeverType>(fnTy))
-
-        if (get<AnyType>(originalCallTy) || get<ErrorType>(originalCallTy) || get<NeverType>(originalCallTy))
             return;
-        else if (std::optional<TypeId> callMm = findMetatableEntry(builtinTypes, module->errors, originalCallTy, "__call", call->func->location))
+        else if (isOptional(fnTy))
         {
-            if (get<FunctionType>(follow(*callMm)))
+            reportError(OptionalValueAccess{fnTy}, call->func->location);
-            {
-                args.head.push_back(originalCallTy);
-                argLocs.push_back(call->func->location);
-            }
-            else
-            {
-                // TODO: This doesn't flag the __call metamethod as the problem
-                // very clearly.
-                reportError(CannotCallNonFunction{*callMm}, call->func->location);
-                return;
-            }
-        }
-        else if (get<FunctionType>(originalCallTy))
-        {
-            // ok.
-        }
-        else if (get<IntersectionType>(originalCallTy))
-        {
-            auto norm = normalizer.normalize(originalCallTy);
-            if (!norm)
-                return reportError(CodeTooComplex{}, call->location);
-
-            // NormalizedType::hasFunction returns true if its' tops component is `unknown`, but for soundness we want the reverse.
-            if (get<UnknownType>(norm->tops) || !norm->hasFunctions())
-                return reportError(CannotCallNonFunction{originalCallTy}, call->func->location);
-        }
-        else if (auto utv = get<UnionType>(originalCallTy))
-        {
-            // Sometimes it's okay to call a union of functions, but only if all of the functions are the same.
-            // Another scenario we might run into it is if the union has a nil member. In this case, we want to throw an error
-            if (isOptional(originalCallTy))
-            {
-                reportError(OptionalValueAccess{originalCallTy}, call->location);
-                return;
-            }
-            std::optional<TypeId> fst;
-            for (TypeId ty : utv)
-            {
-                if (!fst)
-                    fst = follow(ty);
-                else if (fst != follow(ty))
-                {
-                    reportError(CannotCallNonFunction{originalCallTy}, call->func->location);
-                    return;
-                }
-            }
-
-            if (!fst)
-                ice->ice("UnionType had no elements, so fst is nullopt?");
-
-            originalCallTy = follow(*fst);
-            if (!get<FunctionType>(originalCallTy))
-            {
-                reportError(CannotCallNonFunction{originalCallTy}, call->func->location);
-                return;
-            }
-        }
-        else
-        {
-            reportError(CannotCallNonFunction{originalCallTy}, call->func->location);
             return;
         }
 
@@ -1161,9 +1100,12 @@ struct TypeChecker2
                 args.head.push_back(*argTy);
             else if (i == call->args.size - 1)
             {
-                TypePackId* argTail = module->astTypePacks.find(arg);
+                if (auto argTail = module->astTypePacks.find(arg))
-                if (argTail)
+                {
-                    args.tail = *argTail;
+                    auto [head, tail] = flatten(*argTail);
+                    args.head.insert(args.head.end(), head.begin(), head.end());
+                    args.tail = tail;
+                }
                 else
                     args.tail = builtinTypes->anyTypePack;
             }
@@ -1171,142 +1113,318 @@ struct TypeChecker2
                 args.head.push_back(builtinTypes->anyType);
         }
 
-        TypePackId expectedArgTypes = testArena.addTypePack(args);
+        FunctionCallResolver resolver{
+            builtinTypes,
+            NotNull{&testArena},
+            NotNull{&normalizer},
+            NotNull{stack.back()},
+            ice,
+            call->location,
+        };
 
-        if (auto maybeSelectedOverload = module->astOverloadResolvedTypes.find(call))
+        resolver.resolve(fnTy, &args, call->func->location, &argLocs);
+
+        if (!resolver.ok.empty())
+            return; // We found a call that works, so this is ok.
+        else if (auto norm = normalizer.normalize(fnTy); !norm || !normalizer.isInhabited(norm))
         {
-            // This overload might not work still: the constraint solver will
+            if (!norm)
-            // pass the type checker an instantiated function type that matches
+                reportError(NormalizationTooComplex{}, call->func->location);
-            // in arity, but not in subtyping, in order to allow the type
+            else
-            // checker to report better error messages.
+                return; // Ok. Calling an uninhabited type is no-op.
-
+        }
-            TypeId selectedOverload = follow(*maybeSelectedOverload);
+        else if (!resolver.nonviableOverloads.empty())
-            const FunctionType* ftv;
+        {
-
+            if (resolver.nonviableOverloads.size() == 1)
-            if (get<AnyType>(selectedOverload) || get<ErrorType>(selectedOverload) || get<NeverType>(selectedOverload))
+                reportErrors(resolver.nonviableOverloads.front().second);
-            {
-                return;
-            }
-            else if (const FunctionType* overloadFtv = get<FunctionType>(selectedOverload))
-            {
-                ftv = overloadFtv;
-            }
             else
             {
-                reportError(CannotCallNonFunction{selectedOverload}, call->func->location);
+                std::string s = "None of the overloads for function that accept ";
-                return;
+                s += std::to_string(args.head.size());
-            }
+                s += " arguments are compatible.";
-
+                reportError(GenericError{std::move(s)}, call->location);
-            TxnLog fake{};
-
-            LUAU_ASSERT(ftv);
-            reportErrors(tryUnify(stack.back(), call->location, ftv->retTypes, expectedRetType, CountMismatch::Context::Return, /* genericsOkay */ true));
-            reportErrors(
-                reduceFamilies(ftv->retTypes, call->location, NotNull{&testArena}, builtinTypes, stack.back(), NotNull{&normalizer}, &fake, true)
-                    .errors);
-
-            auto it = begin(expectedArgTypes);
-            size_t i = 0;
-            std::vector<TypeId> slice;
-            for (TypeId arg : ftv->argTypes)
-            {
-                if (it == end(expectedArgTypes))
-                {
-                    slice.push_back(arg);
-                    continue;
-                }
-
-                TypeId expectedArg = *it;
-
-                Location argLoc = argLocs.at(i >= argLocs.size() ? argLocs.size() - 1 : i);
-
-                reportErrors(tryUnify(stack.back(), argLoc, expectedArg, arg, CountMismatch::Context::Arg, /* genericsOkay */ true));
-                reportErrors(reduceFamilies(arg, argLoc, NotNull{&testArena}, builtinTypes, stack.back(), NotNull{&normalizer}, &fake, true).errors);
-
-                ++it;
-                ++i;
-            }
-
-            if (slice.size() > 0 && it == end(expectedArgTypes))
-            {
-                if (auto tail = it.tail())
-                {
-                    TypePackId remainingArgs = testArena.addTypePack(TypePack{std::move(slice), std::nullopt});
-                    reportErrors(tryUnify(stack.back(), argLocs.back(), *tail, remainingArgs, CountMismatch::Context::Arg, /* genericsOkay */ true));
-                    reportErrors(reduceFamilies(
-                        remainingArgs, argLocs.back(), NotNull{&testArena}, builtinTypes, stack.back(), NotNull{&normalizer}, &fake, true)
-                                     .errors);
-                }
             }
         }
-        else
+        else if (!resolver.arityMismatches.empty())
         {
-            // No overload worked, even when instantiated. We need to filter the
+            if (resolver.arityMismatches.size() == 1)
-            // set of overloads to those that match the arity of the incoming
+                reportErrors(resolver.arityMismatches.front().second);
-            // argument set, and then report only those as not matching.
+            else
-
-            std::vector<TypeId> arityMatchingOverloads;
-            ErrorVec empty;
-            for (TypeId overload : overloads)
             {
-                overload = follow(overload);
+                std::string s = "No overload for function accepts ";
-                if (const FunctionType* ftv = get<FunctionType>(overload))
+                s += std::to_string(args.head.size());
-                {
+                s += " arguments.";
-                    if (size(ftv->argTypes) == size(expectedArgTypes))
+                reportError(GenericError{std::move(s)}, call->location);
-                    {
-                        arityMatchingOverloads.push_back(overload);
-                    }
-                }
-                else if (const std::optional<TypeId> callMm = findMetatableEntry(builtinTypes, empty, overload, "__call", call->location))
-                {
-                    if (const FunctionType* ftv = get<FunctionType>(follow(*callMm)))
-                    {
-                        if (size(ftv->argTypes) == size(expectedArgTypes))
-                        {
-                            arityMatchingOverloads.push_back(overload);
-                        }
-                    }
-                    else
-                    {
-                        reportError(CannotCallNonFunction{}, call->location);
-                    }
-                }
             }
+        }
+        else if (!resolver.nonFunctions.empty())
+            reportError(CannotCallNonFunction{fnTy}, call->func->location);
+        else
+            LUAU_ASSERT(!"Generating the best possible error from this function call resolution was inexhaustive?");
 
-            if (arityMatchingOverloads.size() == 0)
+        if (resolver.arityMismatches.size() > 1 || resolver.nonviableOverloads.size() > 1)
+        {
+            std::string s = "Available overloads: ";
+
+            std::vector<TypeId> overloads;
+            if (resolver.nonviableOverloads.empty())
             {
-                reportError(
+                for (const auto& [ty, p] : resolver.resolution)
-                    GenericError{"No overload for function accepts " + std::to_string(size(expectedArgTypes)) + " arguments."}, call->location);
+                {
+                    if (p.first == FunctionCallResolver::TypeIsNotAFunction)
+                        continue;
+
+                    overloads.push_back(ty);
+                }
             }
             else
             {
-                // We have handled the case of a singular arity-matching
+                for (const auto& [ty, _] : resolver.nonviableOverloads)
-                // overload above, in the case where an overload was selected.
+                    overloads.push_back(ty);
-                // LUAU_ASSERT(arityMatchingOverloads.size() > 1);
-                reportError(GenericError{"None of the overloads for function that accept " + std::to_string(size(expectedArgTypes)) +
-                                         " arguments are compatible."},
-                    call->location);
             }
 
-            std::string s;
+            for (size_t i = 0; i < overloads.size(); ++i)
-            std::vector<TypeId>& stringifyOverloads = arityMatchingOverloads.size() == 0 ? overloads : arityMatchingOverloads;
-            for (size_t i = 0; i < stringifyOverloads.size(); ++i)
             {
-                TypeId overload = follow(stringifyOverloads[i]);
-
                 if (i > 0)
-                    s += "; ";
+                    s += (i == overloads.size() - 1) ? "; and " : "; ";
 
-                if (i > 0 && i == stringifyOverloads.size() - 1)
+                s += toString(overloads[i]);
-                    s += "and ";
-
-                s += toString(overload);
             }
 
-            reportError(ExtraInformation{"Available overloads: " + s}, call->func->location);
+            reportError(ExtraInformation{std::move(s)}, call->func->location);
         }
     }
 
+    struct FunctionCallResolver
+    {
+        enum Analysis
+        {
+            Ok,
+            TypeIsNotAFunction,
+            ArityMismatch,
+            OverloadIsNonviable, // Arguments were incompatible with the overload's parameters, but were otherwise compatible by arity.
+        };
+
+        NotNull<BuiltinTypes> builtinTypes;
+        NotNull<TypeArena> arena;
+        NotNull<Normalizer> normalizer;
+        NotNull<Scope> scope;
+        NotNull<InternalErrorReporter> ice;
+        Location callLoc;
+
+        std::vector<TypeId> ok;
+        std::vector<TypeId> nonFunctions;
+        std::vector<std::pair<TypeId, ErrorVec>> arityMismatches;
+        std::vector<std::pair<TypeId, ErrorVec>> nonviableOverloads;
+        InsertionOrderedMap<TypeId, std::pair<Analysis, size_t>> resolution;
+
+    private:
+        template<typename Ty>
+        std::optional<ErrorVec> tryUnify(const Location& location, Ty subTy, Ty superTy)
+        {
+            Unifier u{normalizer, scope, location, Covariant};
+            u.ctx = CountMismatch::Arg;
+            u.hideousFixMeGenericsAreActuallyFree = true;
+            u.enableScopeTests();
+            u.tryUnify(subTy, superTy);
+
+            if (u.errors.empty())
+                return std::nullopt;
+
+            return std::move(u.errors);
+        }
+
+        std::pair<Analysis, ErrorVec> checkOverload(TypeId fnTy, const TypePack* args, Location fnLoc, const std::vector<Location>* argLocs, bool callMetamethodOk = true)
+        {
+            fnTy = follow(fnTy);
+
+            ErrorVec discard;
+            if (get<AnyType>(fnTy) || get<ErrorType>(fnTy) || get<NeverType>(fnTy))
+                return {Ok, {}};
+            else if (auto fn = get<FunctionType>(fnTy))
+                return checkOverload_(fnTy, fn, args, fnLoc, argLocs); // Intentionally split to reduce the stack pressure of this function.
+            else if (auto callMm = findMetatableEntry(builtinTypes, discard, fnTy, "__call", callLoc); callMm && callMetamethodOk)
+            {
+                // Calling a metamethod forwards the `fnTy` as self.
+                TypePack withSelf = *args;
+                withSelf.head.insert(withSelf.head.begin(), fnTy);
+
+                std::vector<Location> withSelfLocs = *argLocs;
+                withSelfLocs.insert(withSelfLocs.begin(), fnLoc);
+
+                return checkOverload(*callMm, &withSelf, fnLoc, &withSelfLocs, /*callMetamethodOk=*/ false);
+            }
+            else
+                return {TypeIsNotAFunction, {}}; // Intentionally empty. We can just fabricate the type error later on.
+        }
+
+        LUAU_NOINLINE
+        std::pair<Analysis, ErrorVec> checkOverload_(TypeId fnTy, const FunctionType* fn, const TypePack* args, Location fnLoc, const std::vector<Location>* argLocs)
+        {
+            TxnLog fake;
+            FamilyGraphReductionResult result = reduceFamilies(fnTy, callLoc, arena, builtinTypes, scope, normalizer, &fake, /*force=*/ true);
+            if (!result.errors.empty())
+                return {OverloadIsNonviable, result.errors};
+
+            ErrorVec argumentErrors;
+
+            // Reminder: Functions have parameters. You provide arguments.
+            auto paramIter = begin(fn->argTypes);
+            size_t argOffset = 0;
+
+            while (paramIter != end(fn->argTypes))
+            {
+                if (argOffset >= args->head.size())
+                    break;
+
+                TypeId paramTy = *paramIter;
+                TypeId argTy = args->head[argOffset];
+                Location argLoc = argLocs->at(argOffset >= argLocs->size() ? argLocs->size() - 1 : argOffset);
+
+                if (auto errors = tryUnify(argLoc, argTy, paramTy))
+                {
+                    // Since we're stopping right here, we need to decide if this is a nonviable overload or if there is an arity mismatch.
+                    // If it's a nonviable overload, then we need to keep going to get all type errors.
+                    auto [minParams, optMaxParams] = getParameterExtents(TxnLog::empty(), fn->argTypes);
+                    if (args->head.size() < minParams)
+                        return {ArityMismatch, *errors};
+                    else
+                        argumentErrors.insert(argumentErrors.end(), errors->begin(), errors->end());
+                }
+
+                ++paramIter;
+                ++argOffset;
+            }
+
+            while (argOffset < args->head.size())
+            {
+                // If we can iterate over the head of arguments, then we have exhausted the head of the parameters.
+                LUAU_ASSERT(paramIter == end(fn->argTypes));
+
+                Location argLoc = argLocs->at(argOffset >= argLocs->size() ? argLocs->size() - 1 : argOffset);
+
+                if (!paramIter.tail())
+                {
+                    auto [minParams, optMaxParams] = getParameterExtents(TxnLog::empty(), fn->argTypes);
+                    TypeError error{argLoc, CountMismatch{minParams, optMaxParams, args->head.size(), CountMismatch::Arg, false}};
+                    return {ArityMismatch, {error}};
+                }
+                else if (auto vtp = get<VariadicTypePack>(follow(paramIter.tail())))
+                {
+                    if (auto errors = tryUnify(argLoc, args->head[argOffset], vtp->ty))
+                        argumentErrors.insert(argumentErrors.end(), errors->begin(), errors->end());
+                }
+
+                ++argOffset;
+            }
+
+            while (paramIter != end(fn->argTypes))
+            {
+                // If we can iterate over parameters, then we have exhausted the head of the arguments.
+                LUAU_ASSERT(argOffset == args->head.size());
+
+                // It may have a tail, however, so check that.
+                if (auto vtp = get<VariadicTypePack>(follow(args->tail)))
+                {
+                    Location argLoc = argLocs->at(argLocs->size() - 1);
+
+                    if (auto errors = tryUnify(argLoc, vtp->ty, *paramIter))
+                        argumentErrors.insert(argumentErrors.end(), errors->begin(), errors->end());
+                }
+                else if (!isOptional(*paramIter))
+                {
+                    Location argLoc = argLocs->empty() ? fnLoc : argLocs->at(argLocs->size() - 1);
+
+                    // It is ok to have excess parameters as long as they are all optional.
+                    auto [minParams, optMaxParams] = getParameterExtents(TxnLog::empty(), fn->argTypes);
+                    TypeError error{argLoc, CountMismatch{minParams, optMaxParams, args->head.size(), CountMismatch::Arg, false}};
+                    return {ArityMismatch, {error}};
+                }
+
+                ++paramIter;
+            }
+
+            // We hit the end of the heads for both parameters and arguments, so check their tails.
+            LUAU_ASSERT(paramIter == end(fn->argTypes));
+            LUAU_ASSERT(argOffset == args->head.size());
+
+            if (paramIter.tail() && args->tail)
+            {
+                Location argLoc = argLocs->at(argLocs->size() - 1);
+
+                if (auto errors = tryUnify(argLoc, *args->tail, *paramIter.tail()))
+                    argumentErrors.insert(argumentErrors.end(), errors->begin(), errors->end());
+            }
+
+            return {argumentErrors.empty() ? Ok : OverloadIsNonviable, argumentErrors};
+        }
+
+        size_t indexof(Analysis analysis)
+        {
+            switch (analysis)
+            {
+            case Ok:
+                return ok.size();
+            case TypeIsNotAFunction:
+                return nonFunctions.size();
+            case ArityMismatch:
+                return arityMismatches.size();
+            case OverloadIsNonviable:
+                return nonviableOverloads.size();
+            }
+
+            ice->ice("Inexhaustive switch in FunctionCallResolver::indexof");
+        }
+
+        void add(Analysis analysis, TypeId ty, ErrorVec&& errors)
+        {
+            resolution.insert(ty, {analysis, indexof(analysis)});
+
+            switch (analysis)
+            {
+            case Ok:
+                LUAU_ASSERT(errors.empty());
+                ok.push_back(ty);
+                break;
+            case TypeIsNotAFunction:
+                LUAU_ASSERT(errors.empty());
+                nonFunctions.push_back(ty);
+                break;
+            case ArityMismatch:
+                LUAU_ASSERT(!errors.empty());
+                arityMismatches.emplace_back(ty, std::move(errors));
+                break;
+            case OverloadIsNonviable:
+                LUAU_ASSERT(!errors.empty());
+                nonviableOverloads.emplace_back(ty, std::move(errors));
+                break;
+            }
+        }
+
+    public:
+        void resolve(TypeId fnTy, const TypePack* args, Location selfLoc, const std::vector<Location>* argLocs)
+        {
+            fnTy = follow(fnTy);
+
+            auto it = get<IntersectionType>(fnTy);
+            if (!it)
+            {
+                auto [analysis, errors] = checkOverload(fnTy, args, selfLoc, argLocs);
+                add(analysis, fnTy, std::move(errors));
+                return;
+            }
+
+            for (TypeId ty : it)
+            {
+                if (resolution.find(ty) != resolution.end())
+                    continue;
+
+                auto [analysis, errors] = checkOverload(ty, args, selfLoc, argLocs);
+                add(analysis, ty, std::move(errors));
+            }
+        }
+    };
+
     void visit(AstExprCall* call)
     {
         visit(call->func, ValueContext::RValue);
@@ -1584,7 +1702,11 @@ struct TypeChecker2
             leftType = stripNil(builtinTypes, testArena, leftType);
         }
 
-        bool isStringOperation = isString(leftType) && isString(rightType);
+        const NormalizedType* normLeft = normalizer.normalize(leftType);
+        const NormalizedType* normRight = normalizer.normalize(rightType);
+
+        bool isStringOperation =
+            (normLeft ? normLeft->isSubtypeOfString() : isString(leftType)) && (normRight ? normRight->isSubtypeOfString() : isString(rightType));
 
         if (get<AnyType>(leftType) || get<ErrorType>(leftType) || get<NeverType>(leftType))
             return leftType;
@@ -1630,15 +1752,16 @@ struct TypeChecker2
                 {
                     testUnion(utv, leftMt);
                 }
-
-                // If either left or right has no metatable (or both), we need to consider if
-                // there are values in common that could possibly inhabit the type (and thus equality could be considered)
-                if (!leftMt.has_value() || !rightMt.has_value())
-                {
-                    matches = matches || typesHaveIntersection;
-                }
             }
 
+            // If we're working with things that are not tables, the metatable comparisons above are a little excessive
+            // It's ok for one type to have a meta table and the other to not. In that case, we should fall back on
+            // checking if the intersection of the types is inhabited.
+            // TODO: Maybe add more checks here (e.g. for functions, classes, etc)
+            if (!(get<TableType>(leftType) || get<TableType>(rightType)))
+                if (!leftMt.has_value() || !rightMt.has_value())
+                    matches = matches || typesHaveIntersection;
+
             if (!matches && isComparison)
             {
                 reportError(GenericError{format("Types %s and %s cannot be compared with %s because they do not have the same metatable",
@@ -1663,15 +1786,15 @@ struct TypeChecker2
                 if (overrideKey != nullptr)
                     key = overrideKey;
 
-                TypeId instantiatedMm = module->astOverloadResolvedTypes[key];
+                TypeId* selectedOverloadTy = module->astOverloadResolvedTypes.find(key);
-                if (!instantiatedMm)
+                if (!selectedOverloadTy)
                 {
                     // reportError(CodeTooComplex{}, expr->location);
                     // was handled by a type family
                     return expectedResult;
                 }
 
-                else if (const FunctionType* ftv = get<FunctionType>(follow(instantiatedMm)))
+                else if (const FunctionType* ftv = get<FunctionType>(follow(*selectedOverloadTy)))
                 {
                     TypePackId expectedArgs;
                     // For >= and > we invoke __lt and __le respectively with
@@ -1803,13 +1926,12 @@ struct TypeChecker2
         case AstExprBinary::Op::CompareLe:
         case AstExprBinary::Op::CompareLt:
         {
-            const NormalizedType* leftTyNorm = normalizer.normalize(leftType);
+            if (normLeft && normLeft->isExactlyNumber())
-            if (leftTyNorm && leftTyNorm->isExactlyNumber())
             {
                 reportErrors(tryUnify(scope, expr->right->location, rightType, builtinTypes->numberType));
                 return builtinTypes->numberType;
             }
-            else if (leftTyNorm && leftTyNorm->isSubtypeOfString())
+            else if (normLeft && normLeft->isSubtypeOfString())
             {
                 reportErrors(tryUnify(scope, expr->right->location, rightType, builtinTypes->stringType));
                 return builtinTypes->stringType;

Analysis/src/TypeInfer.cpp

@@ -35,7 +35,6 @@ LUAU_FASTFLAGVARIABLE(DebugLuauFreezeDuringUnification, false)
 LUAU_FASTFLAGVARIABLE(DebugLuauSharedSelf, false)
 LUAU_FASTFLAG(LuauInstantiateInSubtyping)
 LUAU_FASTFLAGVARIABLE(LuauAllowIndexClassParameters, false)
-LUAU_FASTFLAG(LuauUninhabitedSubAnything2)
 LUAU_FASTFLAG(LuauOccursIsntAlwaysFailure)
 LUAU_FASTFLAGVARIABLE(LuauTypecheckTypeguards, false)
 LUAU_FASTFLAGVARIABLE(LuauTinyControlFlowAnalysis, false)
@@ -841,7 +840,7 @@ struct Demoter : Substitution
 
     bool ignoreChildren(TypeId ty) override
     {
-        if (FFlag::LuauClassTypeVarsInSubstitution && get<ClassType>(ty))
+        if (get<ClassType>(ty))
             return true;
 
         return false;
@@ -2648,10 +2647,7 @@ static std::optional<bool> areEqComparable(NotNull<TypeArena> arena, NotNull<Nor
     if (!n)
         return std::nullopt;
 
-    if (FFlag::LuauUninhabitedSubAnything2)
+    return normalizer->isInhabited(n);
-        return normalizer->isInhabited(n);
-    else
-        return isInhabited_DEPRECATED(*n);
 }
 
 TypeId TypeChecker::checkRelationalOperation(

Analysis/src/Unifier.cpp

@@ -19,12 +19,10 @@
 LUAU_FASTINT(LuauTypeInferTypePackLoopLimit)
 LUAU_FASTFLAG(LuauErrorRecoveryType)
 LUAU_FASTFLAGVARIABLE(LuauInstantiateInSubtyping, false)
-LUAU_FASTFLAGVARIABLE(LuauUninhabitedSubAnything2, false)
 LUAU_FASTFLAGVARIABLE(LuauVariadicAnyCanBeGeneric, false)
 LUAU_FASTFLAGVARIABLE(LuauMaintainScopesInUnifier, false)
 LUAU_FASTFLAGVARIABLE(LuauTransitiveSubtyping, false)
 LUAU_FASTFLAGVARIABLE(LuauOccursIsntAlwaysFailure, false)
-LUAU_FASTFLAG(LuauClassTypeVarsInSubstitution)
 LUAU_FASTFLAG(DebugLuauDeferredConstraintResolution)
 LUAU_FASTFLAG(LuauNormalizeBlockedTypes)
 LUAU_FASTFLAG(LuauAlwaysCommitInferencesOfFunctionCalls)
@@ -315,7 +313,7 @@ TypePackId Widen::clean(TypePackId)
 
 bool Widen::ignoreChildren(TypeId ty)
 {
-    if (FFlag::LuauClassTypeVarsInSubstitution && get<ClassType>(ty))
+    if (get<ClassType>(ty))
         return true;
 
     return !log->is<UnionType>(ty);
@@ -748,10 +746,9 @@ void Unifier::tryUnify_(TypeId subTy, TypeId superTy, bool isFunctionCall, bool
     else if (log.get<NegationType>(superTy) || log.get<NegationType>(subTy))
         tryUnifyNegations(subTy, superTy);
 
-    else if (FFlag::LuauUninhabitedSubAnything2 && checkInhabited && !normalizer->isInhabited(subTy))
+    else if (checkInhabited && !normalizer->isInhabited(subTy))
     {
     }
-
     else
         reportError(location, TypeMismatch{superTy, subTy, mismatchContext()});
 
@@ -2365,7 +2362,7 @@ void Unifier::tryUnifyScalarShape(TypeId subTy, TypeId superTy, bool reversed)
     TypeId osubTy = subTy;
     TypeId osuperTy = superTy;
 
-    if (FFlag::LuauUninhabitedSubAnything2 && checkInhabited && !normalizer->isInhabited(subTy))
+    if (checkInhabited && !normalizer->isInhabited(subTy))
         return;
 
     if (reversed)
@@ -2739,7 +2736,7 @@ void Unifier::tryUnifyVariadics(TypePackId subTp, TypePackId superTp, bool rever
             }
         }
     }
-    else if (FFlag::LuauVariadicAnyCanBeGeneric && get<AnyType>(variadicTy) && log.get<GenericTypePack>(subTp))
+    else if (get<AnyType>(variadicTy) && log.get<GenericTypePack>(subTp))
     {
         // Nothing to do.  This is ok.
     }
@@ -2893,7 +2890,7 @@ bool Unifier::occursCheck(TypeId needle, TypeId haystack, bool reversed)
         if (innerState.failure)
         {
             reportError(location, OccursCheckFailed{});
-            log.replace(needle, *builtinTypes->errorRecoveryType());
+            log.replace(needle, BoundType{builtinTypes->errorRecoveryType()});
         }
     }
 

CLI/Compile.cpp

@@ -129,7 +129,7 @@ static double recordDeltaTime(double& timer)
     return delta;
 }
 
-static bool compileFile(const char* name, CompileFormat format, CompileStats& stats)
+static bool compileFile(const char* name, CompileFormat format, Luau::CodeGen::AssemblyOptions::Target assemblyTarget, CompileStats& stats)
 {
     double currts = Luau::TimeTrace::getClock();
 
@@ -150,6 +150,7 @@ static bool compileFile(const char* name, CompileFormat format, CompileStats& st
         Luau::BytecodeBuilder bcb;
 
         Luau::CodeGen::AssemblyOptions options;
+        options.target = assemblyTarget;
         options.outputBinary = format == CompileFormat::CodegenNull;
 
         if (!options.outputBinary)
@@ -248,6 +249,7 @@ static void displayHelp(const char* argv0)
     printf("  -h, --help: Display this usage message.\n");
     printf("  -O<n>: compile with optimization level n (default 1, n should be between 0 and 2).\n");
     printf("  -g<n>: compile with debug level n (default 1, n should be between 0 and 2).\n");
+    printf("  --target=<target>: compile code for specific architecture (a64, x64, a64_nf, x64_ms).\n");
     printf("  --timetrace: record compiler time tracing information into trace.json\n");
 }
 
@@ -264,6 +266,7 @@ int main(int argc, char** argv)
     setLuauFlagsDefault();
 
     CompileFormat compileFormat = CompileFormat::Text;
+    Luau::CodeGen::AssemblyOptions::Target assemblyTarget = Luau::CodeGen::AssemblyOptions::Host;
 
     for (int i = 1; i < argc; i++)
     {
@@ -292,6 +295,24 @@ int main(int argc, char** argv)
             }
             globalOptions.debugLevel = level;
         }
+        else if (strncmp(argv[i], "--target=", 9) == 0)
+        {
+            const char* value = argv[i] + 9;
+
+            if (strcmp(value, "a64") == 0)
+                assemblyTarget = Luau::CodeGen::AssemblyOptions::A64;
+            else if (strcmp(value, "a64_nf") == 0)
+                assemblyTarget = Luau::CodeGen::AssemblyOptions::A64_NoFeatures;
+            else if (strcmp(value, "x64") == 0)
+                assemblyTarget = Luau::CodeGen::AssemblyOptions::X64_SystemV;
+            else if (strcmp(value, "x64_ms") == 0)
+                assemblyTarget = Luau::CodeGen::AssemblyOptions::X64_Windows;
+            else
+            {
+                fprintf(stderr, "Error: unknown target\n");
+                return 1;
+            }
+        }
         else if (strcmp(argv[i], "--timetrace") == 0)
         {
             FFlag::DebugLuauTimeTracing.value = true;
@@ -331,7 +352,7 @@ int main(int argc, char** argv)
     int failed = 0;
 
     for (const std::string& path : files)
-        failed += !compileFile(path.c_str(), compileFormat, stats);
+        failed += !compileFile(path.c_str(), compileFormat, assemblyTarget, stats);
 
     if (compileFormat == CompileFormat::Null)
         printf("Compiled %d KLOC into %d KB bytecode (read %.2fs, parse %.2fs, compile %.2fs)\n", int(stats.lines / 1000), int(stats.bytecode / 1024),

CodeGen/include/Luau/CodeGen.h

@@ -23,6 +23,17 @@ using AnnotatorFn = void (*)(void* context, std::string& result, int fid, int in
 
 struct AssemblyOptions
 {
+    enum Target
+    {
+        Host,
+        A64,
+        A64_NoFeatures,
+        X64_Windows,
+        X64_SystemV,
+    };
+
+    Target target = Host;
+
     bool outputBinary = false;
 
     bool includeAssembly = false;

CodeGen/include/Luau/IrData.h

@@ -414,6 +414,7 @@ enum class IrCmd : uint8_t
     // Handle GC write barrier (forward)
     // A: pointer (GCObject)
     // B: Rn (TValue that was written to the object)
+    // C: tag/undef (tag of the value that was written)
     BARRIER_OBJ,
 
     // Handle GC write barrier (backwards) for a write into a table
@@ -423,6 +424,7 @@ enum class IrCmd : uint8_t
     // Handle GC write barrier (forward) for a write into a table
     // A: pointer (Table)
     // B: Rn (TValue that was written to the object)
+    // C: tag/undef (tag of the value that was written)
     BARRIER_TABLE_FORWARD,
 
     // Update savedpc value
@@ -584,6 +586,14 @@ enum class IrCmd : uint8_t
     // B: double
     // C: double/int (optional, 2nd argument)
     INVOKE_LIBM,
+
+    // Returns the string name of a type based on tag, alternative for type(x)
+    // A: tag
+    GET_TYPE,
+
+    // Returns the string name of a type either from a __type metatable field or just based on the tag, alternative for typeof(x)
+    // A: Rn
+    GET_TYPEOF,
 };
 
 enum class IrConstKind : uint8_t

CodeGen/include/Luau/IrUtils.h

@@ -189,6 +189,8 @@ inline bool hasResult(IrCmd cmd)
     case IrCmd::BITCOUNTLZ_UINT:
     case IrCmd::BITCOUNTRZ_UINT:
     case IrCmd::INVOKE_LIBM:
+    case IrCmd::GET_TYPE:
+    case IrCmd::GET_TYPEOF:
         return true;
     default:
         break;

CodeGen/src/CodeGen.cpp

@@ -1,15 +1,12 @@
 // This file is part of the Luau programming language and is licensed under MIT License; see LICENSE.txt for details
 #include "Luau/CodeGen.h"
 
+#include "CodeGenLower.h"
+
 #include "Luau/Common.h"
 #include "Luau/CodeAllocator.h"
 #include "Luau/CodeBlockUnwind.h"
-#include "Luau/IrAnalysis.h"
 #include "Luau/IrBuilder.h"
-#include "Luau/IrDump.h"
-#include "Luau/IrUtils.h"
-#include "Luau/OptimizeConstProp.h"
-#include "Luau/OptimizeFinalX64.h"
 
 #include "Luau/UnwindBuilder.h"
 #include "Luau/UnwindBuilderDwarf2.h"
@@ -21,17 +18,10 @@
 #include "NativeState.h"
 
 #include "CodeGenA64.h"
-#include "EmitCommonA64.h"
-#include "IrLoweringA64.h"
-
 #include "CodeGenX64.h"
-#include "EmitCommonX64.h"
-#include "EmitInstructionX64.h"
-#include "IrLoweringX64.h"
 
 #include "lapi.h"
 
-#include <algorithm>
 #include <memory>
 #include <optional>
 
@@ -107,238 +97,14 @@ static void logPerfFunction(Proto* p, uintptr_t addr, unsigned size)
         gPerfLogFn(gPerfLogContext, addr, size, name);
 }
 
-template<typename AssemblyBuilder, typename IrLowering>
-static bool lowerImpl(AssemblyBuilder& build, IrLowering& lowering, IrFunction& function, int bytecodeid, AssemblyOptions options)
-{
-    // While we will need a better block ordering in the future, right now we want to mostly preserve build order with fallbacks outlined
-    std::vector<uint32_t> sortedBlocks;
-    sortedBlocks.reserve(function.blocks.size());
-    for (uint32_t i = 0; i < function.blocks.size(); i++)
-        sortedBlocks.push_back(i);
-
-    std::sort(sortedBlocks.begin(), sortedBlocks.end(), [&](uint32_t idxA, uint32_t idxB) {
-        const IrBlock& a = function.blocks[idxA];
-        const IrBlock& b = function.blocks[idxB];
-
-        // Place fallback blocks at the end
-        if ((a.kind == IrBlockKind::Fallback) != (b.kind == IrBlockKind::Fallback))
-            return (a.kind == IrBlockKind::Fallback) < (b.kind == IrBlockKind::Fallback);
-
-        // Try to order by instruction order
-        return a.sortkey < b.sortkey;
-    });
-
-    // For each IR instruction that begins a bytecode instruction, which bytecode instruction is it?
-    std::vector<uint32_t> bcLocations(function.instructions.size() + 1, ~0u);
-
-    for (size_t i = 0; i < function.bcMapping.size(); ++i)
-    {
-        uint32_t irLocation = function.bcMapping[i].irLocation;
-
-        if (irLocation != ~0u)
-            bcLocations[irLocation] = uint32_t(i);
-    }
-
-    bool outputEnabled = options.includeAssembly || options.includeIr;
-
-    IrToStringContext ctx{build.text, function.blocks, function.constants, function.cfg};
-
-    // We use this to skip outlined fallback blocks from IR/asm text output
-    size_t textSize = build.text.length();
-    uint32_t codeSize = build.getCodeSize();
-    bool seenFallback = false;
-
-    IrBlock dummy;
-    dummy.start = ~0u;
-
-    for (size_t i = 0; i < sortedBlocks.size(); ++i)
-    {
-        uint32_t blockIndex = sortedBlocks[i];
-        IrBlock& block = function.blocks[blockIndex];
-
-        if (block.kind == IrBlockKind::Dead)
-            continue;
-
-        LUAU_ASSERT(block.start != ~0u);
-        LUAU_ASSERT(block.finish != ~0u);
-
-        // If we want to skip fallback code IR/asm, we'll record when those blocks start once we see them
-        if (block.kind == IrBlockKind::Fallback && !seenFallback)
-        {
-            textSize = build.text.length();
-            codeSize = build.getCodeSize();
-            seenFallback = true;
-        }
-
-        if (options.includeIr)
-        {
-            build.logAppend("# ");
-            toStringDetailed(ctx, block, blockIndex, /* includeUseInfo */ true);
-        }
-
-        // Values can only reference restore operands in the current block
-        function.validRestoreOpBlockIdx = blockIndex;
-
-        build.setLabel(block.label);
-
-        for (uint32_t index = block.start; index <= block.finish; index++)
-        {
-            LUAU_ASSERT(index < function.instructions.size());
-
-            uint32_t bcLocation = bcLocations[index];
-
-            // If IR instruction is the first one for the original bytecode, we can annotate it with source code text
-            if (outputEnabled && options.annotator && bcLocation != ~0u)
-            {
-                options.annotator(options.annotatorContext, build.text, bytecodeid, bcLocation);
-            }
-
-            // If bytecode needs the location of this instruction for jumps, record it
-            if (bcLocation != ~0u)
-            {
-                Label label = (index == block.start) ? block.label : build.setLabel();
-                function.bcMapping[bcLocation].asmLocation = build.getLabelOffset(label);
-            }
-
-            IrInst& inst = function.instructions[index];
-
-            // Skip pseudo instructions, but make sure they are not used at this stage
-            // This also prevents them from getting into text output when that's enabled
-            if (isPseudo(inst.cmd))
-            {
-                LUAU_ASSERT(inst.useCount == 0);
-                continue;
-            }
-
-            // Either instruction result value is not referenced or the use count is not zero
-            LUAU_ASSERT(inst.lastUse == 0 || inst.useCount != 0);
-
-            if (options.includeIr)
-            {
-                build.logAppend("# ");
-                toStringDetailed(ctx, block, blockIndex, inst, index, /* includeUseInfo */ true);
-            }
-
-            IrBlock& next = i + 1 < sortedBlocks.size() ? function.blocks[sortedBlocks[i + 1]] : dummy;
-
-            lowering.lowerInst(inst, index, next);
-
-            if (lowering.hasError())
-            {
-                // Place labels for all blocks that we're skipping
-                // This is needed to avoid AssemblyBuilder assertions about jumps in earlier blocks with unplaced labels
-                for (size_t j = i + 1; j < sortedBlocks.size(); ++j)
-                {
-                    IrBlock& abandoned = function.blocks[sortedBlocks[j]];
-
-                    build.setLabel(abandoned.label);
-                }
-
-                lowering.finishFunction();
-
-                return false;
-            }
-        }
-
-        lowering.finishBlock();
-
-        if (options.includeIr)
-            build.logAppend("#\n");
-    }
-
-    if (!seenFallback)
-    {
-        textSize = build.text.length();
-        codeSize = build.getCodeSize();
-    }
-
-    lowering.finishFunction();
-
-    if (outputEnabled && !options.includeOutlinedCode && textSize < build.text.size())
-    {
-        build.text.resize(textSize);
-
-        if (options.includeAssembly)
-            build.logAppend("; skipping %u bytes of outlined code\n", unsigned((build.getCodeSize() - codeSize) * sizeof(build.code[0])));
-    }
-
-    return true;
-}
-
-[[maybe_unused]] static bool lowerIr(
-    X64::AssemblyBuilderX64& build, IrBuilder& ir, NativeState& data, ModuleHelpers& helpers, Proto* proto, AssemblyOptions options)
-{
-    optimizeMemoryOperandsX64(ir.function);
-
-    X64::IrLoweringX64 lowering(build, helpers, data, ir.function);
-
-    return lowerImpl(build, lowering, ir.function, proto->bytecodeid, options);
-}
-
-[[maybe_unused]] static bool lowerIr(
-    A64::AssemblyBuilderA64& build, IrBuilder& ir, NativeState& data, ModuleHelpers& helpers, Proto* proto, AssemblyOptions options)
-{
-    A64::IrLoweringA64 lowering(build, helpers, data, ir.function);
-
-    return lowerImpl(build, lowering, ir.function, proto->bytecodeid, options);
-}
-
 template<typename AssemblyBuilder>
-static std::optional<NativeProto> assembleFunction(AssemblyBuilder& build, NativeState& data, ModuleHelpers& helpers, Proto* proto, AssemblyOptions options)
+static std::optional<NativeProto> createNativeFunction(AssemblyBuilder& build, ModuleHelpers& helpers, Proto* proto)
 {
-    if (options.includeAssembly || options.includeIr)
-    {
-        if (proto->debugname)
-            build.logAppend("; function %s(", getstr(proto->debugname));
-        else
-            build.logAppend("; function(");
-
-        for (int i = 0; i < proto->numparams; i++)
-        {
-            LocVar* var = proto->locvars ? &proto->locvars[proto->sizelocvars - proto->numparams + i] : nullptr;
-
-            if (var && var->varname)
-                build.logAppend("%s%s", i == 0 ? "" : ", ", getstr(var->varname));
-            else
-                build.logAppend("%s$arg%d", i == 0 ? "" : ", ", i);
-        }
-
-        if (proto->numparams != 0 && proto->is_vararg)
-            build.logAppend(", ...)");
-        else
-            build.logAppend(")");
-
-        if (proto->linedefined >= 0)
-            build.logAppend(" line %d\n", proto->linedefined);
-        else
-            build.logAppend("\n");
-    }
-
     IrBuilder ir;
     ir.buildFunctionIr(proto);
 
-    computeCfgInfo(ir.function);
+    if (!lowerFunction(ir, build, helpers, proto, {}))
-
-    if (!FFlag::DebugCodegenNoOpt)
-    {
-        bool useValueNumbering = !FFlag::DebugCodegenSkipNumbering;
-
-        constPropInBlockChains(ir, useValueNumbering);
-
-        if (!FFlag::DebugCodegenOptSize)
-            createLinearBlocks(ir, useValueNumbering);
-    }
-
-    if (!lowerIr(build, ir, data, helpers, proto, options))
-    {
-        if (build.logText)
-            build.logAppend("; skipping (can't lower)\n\n");
-
         return std::nullopt;
-    }
-
-    if (build.logText)
-        build.logAppend("\n");
 
     return createNativeProto(proto, ir);
 }
@@ -384,7 +150,7 @@ static void onSetBreakpoint(lua_State* L, Proto* proto, int instruction)
 }
 
 #if defined(__aarch64__)
-static unsigned int getCpuFeaturesA64()
+unsigned int getCpuFeaturesA64()
 {
     unsigned int result = 0;
 
@@ -482,21 +248,6 @@ void create(lua_State* L)
     ecb->setbreakpoint = onSetBreakpoint;
 }
 
-static void gatherFunctions(std::vector<Proto*>& results, Proto* proto)
-{
-    if (results.size() <= size_t(proto->bytecodeid))
-        results.resize(proto->bytecodeid + 1);
-
-    // Skip protos that we've already compiled in this run: this happens because at -O2, inlined functions get their protos reused
-    if (results[proto->bytecodeid])
-        return;
-
-    results[proto->bytecodeid] = proto;
-
-    for (int i = 0; i < proto->sizep; i++)
-        gatherFunctions(results, proto->p[i]);
-}
-
 void compile(lua_State* L, int idx)
 {
     LUAU_ASSERT(lua_isLfunction(L, idx));
@@ -529,7 +280,7 @@ void compile(lua_State* L, int idx)
     // Skip protos that have been compiled during previous invocations of CodeGen::compile
     for (Proto* p : protos)
         if (p && p->execdata == nullptr)
-            if (std::optional<NativeProto> np = assembleFunction(build, *data, helpers, p, {}))
+            if (std::optional<NativeProto> np = createNativeFunction(build, helpers, p))
                 results.push_back(*np);
 
     // Very large modules might result in overflowing a jump offset; in this case we currently abandon the entire module
@@ -580,51 +331,6 @@ void compile(lua_State* L, int idx)
     }
 }
 
-std::string getAssembly(lua_State* L, int idx, AssemblyOptions options)
-{
-    LUAU_ASSERT(lua_isLfunction(L, idx));
-    const TValue* func = luaA_toobject(L, idx);
-
-#if defined(__aarch64__)
-    A64::AssemblyBuilderA64 build(/* logText= */ options.includeAssembly, getCpuFeaturesA64());
-#else
-    X64::AssemblyBuilderX64 build(/* logText= */ options.includeAssembly);
-#endif
-
-    NativeState data;
-    initFunctions(data);
-
-    std::vector<Proto*> protos;
-    gatherFunctions(protos, clvalue(func)->l.p);
-
-    ModuleHelpers helpers;
-#if defined(__aarch64__)
-    A64::assembleHelpers(build, helpers);
-#else
-    X64::assembleHelpers(build, helpers);
-#endif
-
-    if (!options.includeOutlinedCode && options.includeAssembly)
-    {
-        build.text.clear();
-        build.logAppend("; skipping %u bytes of outlined helpers\n", unsigned(build.getCodeSize() * sizeof(build.code[0])));
-    }
-
-    for (Proto* p : protos)
-        if (p)
-            if (std::optional<NativeProto> np = assembleFunction(build, data, helpers, p, options))
-                destroyExecData(np->execdata);
-
-    if (!build.finalize())
-        return std::string();
-
-    if (options.outputBinary)
-        return std::string(reinterpret_cast<const char*>(build.code.data()), reinterpret_cast<const char*>(build.code.data() + build.code.size())) +
-               std::string(build.data.begin(), build.data.end());
-    else
-        return build.text;
-}
-
 void setPerfLog(void* context, PerfLogFn logFn)
 {
     gPerfLogContext = context;

CodeGen/src/CodeGenAssembly.cpp

@@ -0,0 +1,146 @@
+// This file is part of the Luau programming language and is licensed under MIT License; see LICENSE.txt for details
+#include "Luau/CodeGen.h"
+
+#include "CodeGenLower.h"
+
+#include "CodeGenA64.h"
+#include "CodeGenX64.h"
+
+#include "lapi.h"
+
+namespace Luau
+{
+namespace CodeGen
+{
+
+template<typename AssemblyBuilder>
+static void logFunctionHeader(AssemblyBuilder& build, Proto* proto)
+{
+    if (proto->debugname)
+        build.logAppend("; function %s(", getstr(proto->debugname));
+    else
+        build.logAppend("; function(");
+
+    for (int i = 0; i < proto->numparams; i++)
+    {
+        LocVar* var = proto->locvars ? &proto->locvars[proto->sizelocvars - proto->numparams + i] : nullptr;
+
+        if (var && var->varname)
+            build.logAppend("%s%s", i == 0 ? "" : ", ", getstr(var->varname));
+        else
+            build.logAppend("%s$arg%d", i == 0 ? "" : ", ", i);
+    }
+
+    if (proto->numparams != 0 && proto->is_vararg)
+        build.logAppend(", ...)");
+    else
+        build.logAppend(")");
+
+    if (proto->linedefined >= 0)
+        build.logAppend(" line %d\n", proto->linedefined);
+    else
+        build.logAppend("\n");
+}
+
+template<typename AssemblyBuilder>
+static std::string getAssemblyImpl(AssemblyBuilder& build, const TValue* func, AssemblyOptions options)
+{
+    std::vector<Proto*> protos;
+    gatherFunctions(protos, clvalue(func)->l.p);
+
+    ModuleHelpers helpers;
+    assembleHelpers(build, helpers);
+
+    if (!options.includeOutlinedCode && options.includeAssembly)
+    {
+        build.text.clear();
+        build.logAppend("; skipping %u bytes of outlined helpers\n", unsigned(build.getCodeSize() * sizeof(build.code[0])));
+    }
+
+    for (Proto* p : protos)
+        if (p)
+        {
+            IrBuilder ir;
+            ir.buildFunctionIr(p);
+
+            if (options.includeAssembly || options.includeIr)
+                logFunctionHeader(build, p);
+
+            if (!lowerFunction(ir, build, helpers, p, options))
+            {
+                if (build.logText)
+                    build.logAppend("; skipping (can't lower)\n");
+            }
+
+            if (build.logText)
+                build.logAppend("\n");
+        }
+
+    if (!build.finalize())
+        return std::string();
+
+    if (options.outputBinary)
+        return std::string(reinterpret_cast<const char*>(build.code.data()), reinterpret_cast<const char*>(build.code.data() + build.code.size())) +
+               std::string(build.data.begin(), build.data.end());
+    else
+        return build.text;
+}
+
+#if defined(__aarch64__)
+unsigned int getCpuFeaturesA64();
+#endif
+
+std::string getAssembly(lua_State* L, int idx, AssemblyOptions options)
+{
+    LUAU_ASSERT(lua_isLfunction(L, idx));
+    const TValue* func = luaA_toobject(L, idx);
+
+    switch (options.target)
+    {
+    case AssemblyOptions::Host:
+    {
+#if defined(__aarch64__)
+        A64::AssemblyBuilderA64 build(/* logText= */ options.includeAssembly, getCpuFeaturesA64());
+#else
+        X64::AssemblyBuilderX64 build(/* logText= */ options.includeAssembly);
+#endif
+
+        return getAssemblyImpl(build, func, options);
+    }
+
+    case AssemblyOptions::A64:
+    {
+        A64::AssemblyBuilderA64 build(/* logText= */ options.includeAssembly, /* features= */ A64::Feature_JSCVT);
+
+        return getAssemblyImpl(build, func, options);
+    }
+
+    case AssemblyOptions::A64_NoFeatures:
+    {
+        A64::AssemblyBuilderA64 build(/* logText= */ options.includeAssembly, /* features= */ 0);
+
+        return getAssemblyImpl(build, func, options);
+    }
+
+    case AssemblyOptions::X64_Windows:
+    {
+        X64::AssemblyBuilderX64 build(/* logText= */ options.includeAssembly, X64::ABIX64::Windows);
+
+        return getAssemblyImpl(build, func, options);
+    }
+
+    case AssemblyOptions::X64_SystemV:
+    {
+        X64::AssemblyBuilderX64 build(/* logText= */ options.includeAssembly, X64::ABIX64::SystemV);
+
+        return getAssemblyImpl(build, func, options);
+    }
+
+    default:
+        LUAU_ASSERT(!"Unknown target");
+        return std::string();
+    }
+}
+
+} // namespace CodeGen
+} // namespace Luau

CodeGen/src/CodeGenLower.h

@@ -0,0 +1,240 @@
+// This file is part of the Luau programming language and is licensed under MIT License; see LICENSE.txt for details
+#pragma once
+
+#include "Luau/AssemblyBuilderA64.h"
+#include "Luau/AssemblyBuilderX64.h"
+#include "Luau/CodeGen.h"
+#include "Luau/IrBuilder.h"
+#include "Luau/IrDump.h"
+#include "Luau/IrUtils.h"
+#include "Luau/OptimizeConstProp.h"
+#include "Luau/OptimizeFinalX64.h"
+
+#include "EmitCommon.h"
+#include "IrLoweringA64.h"
+#include "IrLoweringX64.h"
+
+#include "lobject.h"
+#include "lstate.h"
+
+#include <algorithm>
+#include <vector>
+
+LUAU_FASTFLAG(DebugCodegenNoOpt)
+LUAU_FASTFLAG(DebugCodegenOptSize)
+LUAU_FASTFLAG(DebugCodegenSkipNumbering)
+
+namespace Luau
+{
+namespace CodeGen
+{
+
+inline void gatherFunctions(std::vector<Proto*>& results, Proto* proto)
+{
+    if (results.size() <= size_t(proto->bytecodeid))
+        results.resize(proto->bytecodeid + 1);
+
+    // Skip protos that we've already compiled in this run: this happens because at -O2, inlined functions get their protos reused
+    if (results[proto->bytecodeid])
+        return;
+
+    results[proto->bytecodeid] = proto;
+
+    for (int i = 0; i < proto->sizep; i++)
+        gatherFunctions(results, proto->p[i]);
+}
+
+template<typename AssemblyBuilder, typename IrLowering>
+inline bool lowerImpl(AssemblyBuilder& build, IrLowering& lowering, IrFunction& function, int bytecodeid, AssemblyOptions options)
+{
+    // While we will need a better block ordering in the future, right now we want to mostly preserve build order with fallbacks outlined
+    std::vector<uint32_t> sortedBlocks;
+    sortedBlocks.reserve(function.blocks.size());
+    for (uint32_t i = 0; i < function.blocks.size(); i++)
+        sortedBlocks.push_back(i);
+
+    std::sort(sortedBlocks.begin(), sortedBlocks.end(), [&](uint32_t idxA, uint32_t idxB) {
+        const IrBlock& a = function.blocks[idxA];
+        const IrBlock& b = function.blocks[idxB];
+
+        // Place fallback blocks at the end
+        if ((a.kind == IrBlockKind::Fallback) != (b.kind == IrBlockKind::Fallback))
+            return (a.kind == IrBlockKind::Fallback) < (b.kind == IrBlockKind::Fallback);
+
+        // Try to order by instruction order
+        return a.sortkey < b.sortkey;
+    });
+
+    // For each IR instruction that begins a bytecode instruction, which bytecode instruction is it?
+    std::vector<uint32_t> bcLocations(function.instructions.size() + 1, ~0u);
+
+    for (size_t i = 0; i < function.bcMapping.size(); ++i)
+    {
+        uint32_t irLocation = function.bcMapping[i].irLocation;
+
+        if (irLocation != ~0u)
+            bcLocations[irLocation] = uint32_t(i);
+    }
+
+    bool outputEnabled = options.includeAssembly || options.includeIr;
+
+    IrToStringContext ctx{build.text, function.blocks, function.constants, function.cfg};
+
+    // We use this to skip outlined fallback blocks from IR/asm text output
+    size_t textSize = build.text.length();
+    uint32_t codeSize = build.getCodeSize();
+    bool seenFallback = false;
+
+    IrBlock dummy;
+    dummy.start = ~0u;
+
+    for (size_t i = 0; i < sortedBlocks.size(); ++i)
+    {
+        uint32_t blockIndex = sortedBlocks[i];
+        IrBlock& block = function.blocks[blockIndex];
+
+        if (block.kind == IrBlockKind::Dead)
+            continue;
+
+        LUAU_ASSERT(block.start != ~0u);
+        LUAU_ASSERT(block.finish != ~0u);
+
+        // If we want to skip fallback code IR/asm, we'll record when those blocks start once we see them
+        if (block.kind == IrBlockKind::Fallback && !seenFallback)
+        {
+            textSize = build.text.length();
+            codeSize = build.getCodeSize();
+            seenFallback = true;
+        }
+
+        if (options.includeIr)
+        {
+            build.logAppend("# ");
+            toStringDetailed(ctx, block, blockIndex, /* includeUseInfo */ true);
+        }
+
+        // Values can only reference restore operands in the current block
+        function.validRestoreOpBlockIdx = blockIndex;
+
+        build.setLabel(block.label);
+
+        for (uint32_t index = block.start; index <= block.finish; index++)
+        {
+            LUAU_ASSERT(index < function.instructions.size());
+
+            uint32_t bcLocation = bcLocations[index];
+
+            // If IR instruction is the first one for the original bytecode, we can annotate it with source code text
+            if (outputEnabled && options.annotator && bcLocation != ~0u)
+            {
+                options.annotator(options.annotatorContext, build.text, bytecodeid, bcLocation);
+            }
+
+            // If bytecode needs the location of this instruction for jumps, record it
+            if (bcLocation != ~0u)
+            {
+                Label label = (index == block.start) ? block.label : build.setLabel();
+                function.bcMapping[bcLocation].asmLocation = build.getLabelOffset(label);
+            }
+
+            IrInst& inst = function.instructions[index];
+
+            // Skip pseudo instructions, but make sure they are not used at this stage
+            // This also prevents them from getting into text output when that's enabled
+            if (isPseudo(inst.cmd))
+            {
+                LUAU_ASSERT(inst.useCount == 0);
+                continue;
+            }
+
+            // Either instruction result value is not referenced or the use count is not zero
+            LUAU_ASSERT(inst.lastUse == 0 || inst.useCount != 0);
+
+            if (options.includeIr)
+            {
+                build.logAppend("# ");
+                toStringDetailed(ctx, block, blockIndex, inst, index, /* includeUseInfo */ true);
+            }
+
+            IrBlock& next = i + 1 < sortedBlocks.size() ? function.blocks[sortedBlocks[i + 1]] : dummy;
+
+            lowering.lowerInst(inst, index, next);
+
+            if (lowering.hasError())
+            {
+                // Place labels for all blocks that we're skipping
+                // This is needed to avoid AssemblyBuilder assertions about jumps in earlier blocks with unplaced labels
+                for (size_t j = i + 1; j < sortedBlocks.size(); ++j)
+                {
+                    IrBlock& abandoned = function.blocks[sortedBlocks[j]];
+
+                    build.setLabel(abandoned.label);
+                }
+
+                lowering.finishFunction();
+
+                return false;
+            }
+        }
+
+        lowering.finishBlock();
+
+        if (options.includeIr)
+            build.logAppend("#\n");
+    }
+
+    if (!seenFallback)
+    {
+        textSize = build.text.length();
+        codeSize = build.getCodeSize();
+    }
+
+    lowering.finishFunction();
+
+    if (outputEnabled && !options.includeOutlinedCode && textSize < build.text.size())
+    {
+        build.text.resize(textSize);
+
+        if (options.includeAssembly)
+            build.logAppend("; skipping %u bytes of outlined code\n", unsigned((build.getCodeSize() - codeSize) * sizeof(build.code[0])));
+    }
+
+    return true;
+}
+
+inline bool lowerIr(X64::AssemblyBuilderX64& build, IrBuilder& ir, ModuleHelpers& helpers, Proto* proto, AssemblyOptions options)
+{
+    optimizeMemoryOperandsX64(ir.function);
+
+    X64::IrLoweringX64 lowering(build, helpers, ir.function);
+
+    return lowerImpl(build, lowering, ir.function, proto->bytecodeid, options);
+}
+
+inline bool lowerIr(A64::AssemblyBuilderA64& build, IrBuilder& ir, ModuleHelpers& helpers, Proto* proto, AssemblyOptions options)
+{
+    A64::IrLoweringA64 lowering(build, helpers, ir.function);
+
+    return lowerImpl(build, lowering, ir.function, proto->bytecodeid, options);
+}
+
+template<typename AssemblyBuilder>
+inline bool lowerFunction(IrBuilder& ir, AssemblyBuilder& build, ModuleHelpers& helpers, Proto* proto, AssemblyOptions options)
+{
+    computeCfgInfo(ir.function);
+
+    if (!FFlag::DebugCodegenNoOpt)
+    {
+        bool useValueNumbering = !FFlag::DebugCodegenSkipNumbering;
+
+        constPropInBlockChains(ir, useValueNumbering);
+
+        if (!FFlag::DebugCodegenOptSize)
+            createLinearBlocks(ir, useValueNumbering);
+    }
+
+    return lowerIr(build, ir, helpers, proto, options);
+}
+
+} // namespace CodeGen
+} // namespace Luau

CodeGen/src/EmitBuiltinsX64.cpp

@@ -75,29 +75,6 @@ static void emitBuiltinMathSign(IrRegAllocX64& regs, AssemblyBuilderX64& build,
     build.vmovsd(luauRegValue(ra), tmp0.reg);
 }
 
-static void emitBuiltinType(IrRegAllocX64& regs, AssemblyBuilderX64& build, int ra, int arg)
-{
-    ScopedRegX64 tmp0{regs, SizeX64::qword};
-    ScopedRegX64 tag{regs, SizeX64::dword};
-
-    build.mov(tag.reg, luauRegTag(arg));
-
-    build.mov(tmp0.reg, qword[rState + offsetof(lua_State, global)]);
-    build.mov(tmp0.reg, qword[tmp0.reg + qwordReg(tag.reg) * sizeof(TString*) + offsetof(global_State, ttname)]);
-
-    build.mov(luauRegValue(ra), tmp0.reg);
-}
-
-static void emitBuiltinTypeof(IrRegAllocX64& regs, AssemblyBuilderX64& build, int ra, int arg)
-{
-    IrCallWrapperX64 callWrap(regs, build);
-    callWrap.addArgument(SizeX64::qword, rState);
-    callWrap.addArgument(SizeX64::qword, luauRegAddress(arg));
-    callWrap.call(qword[rNativeContext + offsetof(NativeContext, luaT_objtypenamestr)]);
-
-    build.mov(luauRegValue(ra), rax);
-}
-
 void emitBuiltin(IrRegAllocX64& regs, AssemblyBuilderX64& build, int bfid, int ra, int arg, OperandX64 arg2, int nparams, int nresults)
 {
     switch (bfid)
@@ -111,12 +88,6 @@ void emitBuiltin(IrRegAllocX64& regs, AssemblyBuilderX64& build, int bfid, int r
     case LBF_MATH_SIGN:
         LUAU_ASSERT(nparams == 1 && nresults == 1);
         return emitBuiltinMathSign(regs, build, ra, arg);
-    case LBF_TYPE:
-        LUAU_ASSERT(nparams == 1 && nresults == 1);
-        return emitBuiltinType(regs, build, ra, arg);
-    case LBF_TYPEOF:
-        LUAU_ASSERT(nparams == 1 && nresults == 1);
-        return emitBuiltinTypeof(regs, build, ra, arg);
     default:
         LUAU_ASSERT(!"Missing x64 lowering");
     }

CodeGen/src/EmitCommonX64.cpp

@@ -5,6 +5,7 @@
 #include "Luau/IrCallWrapperX64.h"
 #include "Luau/IrData.h"
 #include "Luau/IrRegAllocX64.h"
+#include "Luau/IrUtils.h"
 
 #include "NativeState.h"
 
@@ -179,11 +180,15 @@ void callSetTable(IrRegAllocX64& regs, AssemblyBuilderX64& build, int rb, Operan
     emitUpdateBase(build);
 }
 
-void checkObjectBarrierConditions(AssemblyBuilderX64& build, RegisterX64 tmp, RegisterX64 object, int ra, Label& skip)
+void checkObjectBarrierConditions(AssemblyBuilderX64& build, RegisterX64 tmp, RegisterX64 object, int ra, int ratag, Label& skip)
 {
-    // iscollectable(ra)
+    // Barrier should've been optimized away if we know that it's not collectable, checking for correctness
-    build.cmp(luauRegTag(ra), LUA_TSTRING);
+    if (ratag == -1 || !isGCO(ratag))
-    build.jcc(ConditionX64::Less, skip);
+    {
+        // iscollectable(ra)
+        build.cmp(luauRegTag(ra), LUA_TSTRING);
+        build.jcc(ConditionX64::Less, skip);
+    }
 
     // isblack(obj2gco(o))
     build.test(byte[object + offsetof(GCheader, marked)], bitmask(BLACKBIT));
@@ -195,12 +200,12 @@ void checkObjectBarrierConditions(AssemblyBuilderX64& build, RegisterX64 tmp, Re
     build.jcc(ConditionX64::Zero, skip);
 }
 
-void callBarrierObject(IrRegAllocX64& regs, AssemblyBuilderX64& build, RegisterX64 object, IrOp objectOp, int ra)
+void callBarrierObject(IrRegAllocX64& regs, AssemblyBuilderX64& build, RegisterX64 object, IrOp objectOp, int ra, int ratag)
 {
     Label skip;
 
     ScopedRegX64 tmp{regs, SizeX64::qword};
-    checkObjectBarrierConditions(build, tmp.reg, object, ra, skip);
+    checkObjectBarrierConditions(build, tmp.reg, object, ra, ratag, skip);
 
     {
         ScopedSpills spillGuard(regs);

CodeGen/src/EmitCommonX64.h

@@ -170,8 +170,8 @@ void callLengthHelper(IrRegAllocX64& regs, AssemblyBuilderX64& build, int ra, in
 void callPrepareForN(IrRegAllocX64& regs, AssemblyBuilderX64& build, int limit, int step, int init);
 void callGetTable(IrRegAllocX64& regs, AssemblyBuilderX64& build, int rb, OperandX64 c, int ra);
 void callSetTable(IrRegAllocX64& regs, AssemblyBuilderX64& build, int rb, OperandX64 c, int ra);
-void checkObjectBarrierConditions(AssemblyBuilderX64& build, RegisterX64 tmp, RegisterX64 object, int ra, Label& skip);
+void checkObjectBarrierConditions(AssemblyBuilderX64& build, RegisterX64 tmp, RegisterX64 object, int ra, int ratag, Label& skip);
-void callBarrierObject(IrRegAllocX64& regs, AssemblyBuilderX64& build, RegisterX64 object, IrOp objectOp, int ra);
+void callBarrierObject(IrRegAllocX64& regs, AssemblyBuilderX64& build, RegisterX64 object, IrOp objectOp, int ra, int ratag);
 void callBarrierTableFast(IrRegAllocX64& regs, AssemblyBuilderX64& build, RegisterX64 table, IrOp tableOp);
 void callStepGc(IrRegAllocX64& regs, AssemblyBuilderX64& build);
 

CodeGen/src/IrAnalysis.cpp

@@ -444,6 +444,9 @@ static RegisterSet computeBlockLiveInRegSet(IrFunction& function, const IrBlock&
         case IrCmd::ADJUST_STACK_TO_TOP:
             // While this can be considered to be a vararg consumer, it is already handled in fastcall instructions
             break;
+        case IrCmd::GET_TYPEOF:
+            use(inst.a);
+            break;
 
         default:
             // All instructions which reference registers have to be handled explicitly

CodeGen/src/IrDump.cpp

@@ -297,6 +297,10 @@ const char* getCmdName(IrCmd cmd)
         return "BITCOUNTRZ_UINT";
     case IrCmd::INVOKE_LIBM:
         return "INVOKE_LIBM";
+    case IrCmd::GET_TYPE:
+        return "GET_TYPE";
+    case IrCmd::GET_TYPEOF:
+        return "GET_TYPEOF";
     }
 
     LUAU_UNREACHABLE();

CodeGen/src/IrLoweringA64.cpp

@@ -60,14 +60,18 @@ inline ConditionA64 getConditionFP(IrCondition cond)
     }
 }
 
-static void checkObjectBarrierConditions(AssemblyBuilderA64& build, RegisterA64 object, RegisterA64 temp, int ra, Label& skip)
+static void checkObjectBarrierConditions(AssemblyBuilderA64& build, RegisterA64 object, RegisterA64 temp, int ra, int ratag, Label& skip)
 {
     RegisterA64 tempw = castReg(KindA64::w, temp);
 
-    // iscollectable(ra)
+    // Barrier should've been optimized away if we know that it's not collectable, checking for correctness
-    build.ldr(tempw, mem(rBase, ra * sizeof(TValue) + offsetof(TValue, tt)));
+    if (ratag == -1 || !isGCO(ratag))
-    build.cmp(tempw, LUA_TSTRING);
+    {
-    build.b(ConditionA64::Less, skip);
+        // iscollectable(ra)
+        build.ldr(tempw, mem(rBase, ra * sizeof(TValue) + offsetof(TValue, tt)));
+        build.cmp(tempw, LUA_TSTRING);
+        build.b(ConditionA64::Less, skip);
+    }
 
     // isblack(obj2gco(o))
     build.ldrb(tempw, mem(object, offsetof(GCheader, marked)));
@@ -162,33 +166,15 @@ static bool emitBuiltin(
         build.str(d0, mem(rBase, res * sizeof(TValue) + offsetof(TValue, value.n)));
         return true;
 
-    case LBF_TYPE:
-        build.ldr(w0, mem(rBase, arg * sizeof(TValue) + offsetof(TValue, tt)));
-        build.ldr(x1, mem(rState, offsetof(lua_State, global)));
-        LUAU_ASSERT(sizeof(TString*) == 8);
-        build.add(x1, x1, zextReg(w0), 3);
-        build.ldr(x0, mem(x1, offsetof(global_State, ttname)));
-        build.str(x0, mem(rBase, res * sizeof(TValue) + offsetof(TValue, value.gc)));
-        return true;
-
-    case LBF_TYPEOF:
-        build.mov(x0, rState);
-        build.add(x1, rBase, uint16_t(arg * sizeof(TValue)));
-        build.ldr(x2, mem(rNativeContext, offsetof(NativeContext, luaT_objtypenamestr)));
-        build.blr(x2);
-        build.str(x0, mem(rBase, res * sizeof(TValue) + offsetof(TValue, value.gc)));
-        return true;
-
     default:
         LUAU_ASSERT(!"Missing A64 lowering");
         return false;
     }
 }
 
-IrLoweringA64::IrLoweringA64(AssemblyBuilderA64& build, ModuleHelpers& helpers, NativeState& data, IrFunction& function)
+IrLoweringA64::IrLoweringA64(AssemblyBuilderA64& build, ModuleHelpers& helpers, IrFunction& function)
     : build(build)
     , helpers(helpers)
-    , data(data)
     , function(function)
     , regs(function, {{x0, x15}, {x16, x17}, {q0, q7}, {q16, q31}})
     , valueTracker(function)
@@ -1004,7 +990,7 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, IrBlock& next)
         build.str(temp3, temp2);
 
         Label skip;
-        checkObjectBarrierConditions(build, temp1, temp2, vmRegOp(inst.b), skip);
+        checkObjectBarrierConditions(build, temp1, temp2, vmRegOp(inst.b), /* ratag */ -1, skip);
 
         size_t spills = regs.spill(build, index, {temp1});
 
@@ -1210,7 +1196,7 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, IrBlock& next)
         RegisterA64 temp = regs.allocTemp(KindA64::x);
 
         Label skip;
-        checkObjectBarrierConditions(build, regOp(inst.a), temp, vmRegOp(inst.b), skip);
+        checkObjectBarrierConditions(build, regOp(inst.a), temp, vmRegOp(inst.b), inst.c.kind == IrOpKind::Undef ? -1 : tagOp(inst.c), skip);
 
         RegisterA64 reg = regOp(inst.a); // note: we need to call regOp before spill so that we don't do redundant reloads
         size_t spills = regs.spill(build, index, {reg});
@@ -1254,7 +1240,7 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, IrBlock& next)
         RegisterA64 temp = regs.allocTemp(KindA64::x);
 
         Label skip;
-        checkObjectBarrierConditions(build, regOp(inst.a), temp, vmRegOp(inst.b), skip);
+        checkObjectBarrierConditions(build, regOp(inst.a), temp, vmRegOp(inst.b), inst.c.kind == IrOpKind::Undef ? -1 : tagOp(inst.c), skip);
 
         RegisterA64 reg = regOp(inst.a); // note: we need to call regOp before spill so that we don't do redundant reloads
         size_t spills = regs.spill(build, index, {reg});
@@ -1710,6 +1696,34 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, IrBlock& next)
         inst.regA64 = regs.takeReg(d0, index);
         break;
     }
+    case IrCmd::GET_TYPE:
+    {
+        inst.regA64 = regs.allocReg(KindA64::x, index);
+
+        build.ldr(inst.regA64, mem(rState, offsetof(lua_State, global)));
+        LUAU_ASSERT(sizeof(TString*) == 8);
+
+        if (inst.a.kind == IrOpKind::Inst)
+            build.add(inst.regA64, inst.regA64, zextReg(regOp(inst.a)), 3);
+        else if (inst.a.kind == IrOpKind::Constant)
+            build.add(inst.regA64, inst.regA64, uint16_t(tagOp(inst.a)) * 8);
+        else
+            LUAU_ASSERT(!"Unsupported instruction form");
+
+        build.ldr(inst.regA64, mem(inst.regA64, offsetof(global_State, ttname)));
+        break;
+    }
+    case IrCmd::GET_TYPEOF:
+    {
+        regs.spill(build, index);
+        build.mov(x0, rState);
+        build.add(x1, rBase, uint16_t(vmRegOp(inst.a) * sizeof(TValue)));
+        build.ldr(x2, mem(rNativeContext, offsetof(NativeContext, luaT_objtypenamestr)));
+        build.blr(x2);
+
+        inst.regA64 = regs.takeReg(x0, index);
+        break;
+    }
 
         // To handle unsupported instructions, add "case IrCmd::OP" and make sure to set error = true!
     }

CodeGen/src/IrLoweringA64.h

@@ -15,7 +15,6 @@ namespace CodeGen
 {
 
 struct ModuleHelpers;
-struct NativeState;
 struct AssemblyOptions;
 
 namespace A64
@@ -23,7 +22,7 @@ namespace A64
 
 struct IrLoweringA64
 {
-    IrLoweringA64(AssemblyBuilderA64& build, ModuleHelpers& helpers, NativeState& data, IrFunction& function);
+    IrLoweringA64(AssemblyBuilderA64& build, ModuleHelpers& helpers, IrFunction& function);
 
     void lowerInst(IrInst& inst, uint32_t index, IrBlock& next);
     void finishBlock();
@@ -63,7 +62,6 @@ struct IrLoweringA64
 
     AssemblyBuilderA64& build;
     ModuleHelpers& helpers;
-    NativeState& data;
 
     IrFunction& function;
 

CodeGen/src/IrLoweringX64.cpp

@@ -22,10 +22,9 @@ namespace CodeGen
 namespace X64
 {
 
-IrLoweringX64::IrLoweringX64(AssemblyBuilderX64& build, ModuleHelpers& helpers, NativeState& data, IrFunction& function)
+IrLoweringX64::IrLoweringX64(AssemblyBuilderX64& build, ModuleHelpers& helpers, IrFunction& function)
     : build(build)
     , helpers(helpers)
-    , data(data)
     , function(function)
     , regs(build, function)
     , valueTracker(function)
@@ -872,7 +871,7 @@ void IrLoweringX64::lowerInst(IrInst& inst, uint32_t index, IrBlock& next)
 
         tmp1.free();
 
-        callBarrierObject(regs, build, tmp2.release(), {}, vmRegOp(inst.b));
+        callBarrierObject(regs, build, tmp2.release(), {}, vmRegOp(inst.b), /* ratag */ -1);
         break;
     }
     case IrCmd::PREPARE_FORN:
@@ -983,7 +982,7 @@ void IrLoweringX64::lowerInst(IrInst& inst, uint32_t index, IrBlock& next)
         callStepGc(regs, build);
         break;
     case IrCmd::BARRIER_OBJ:
-        callBarrierObject(regs, build, regOp(inst.a), inst.a, vmRegOp(inst.b));
+        callBarrierObject(regs, build, regOp(inst.a), inst.a, vmRegOp(inst.b), inst.c.kind == IrOpKind::Undef ? -1 : tagOp(inst.c));
         break;
     case IrCmd::BARRIER_TABLE_BACK:
         callBarrierTableFast(regs, build, regOp(inst.a), inst.a);
@@ -993,7 +992,7 @@ void IrLoweringX64::lowerInst(IrInst& inst, uint32_t index, IrBlock& next)
         Label skip;
 
         ScopedRegX64 tmp{regs, SizeX64::qword};
-        checkObjectBarrierConditions(build, tmp.reg, regOp(inst.a), vmRegOp(inst.b), skip);
+        checkObjectBarrierConditions(build, tmp.reg, regOp(inst.a), vmRegOp(inst.b), inst.c.kind == IrOpKind::Undef ? -1 : tagOp(inst.c), skip);
 
         {
             ScopedSpills spillGuard(regs);
@@ -1350,6 +1349,30 @@ void IrLoweringX64::lowerInst(IrInst& inst, uint32_t index, IrBlock& next)
         inst.regX64 = regs.takeReg(xmm0, index);
         break;
     }
+    case IrCmd::GET_TYPE:
+    {
+        inst.regX64 = regs.allocReg(SizeX64::qword, index);
+
+        build.mov(inst.regX64, qword[rState + offsetof(lua_State, global)]);
+
+        if (inst.a.kind == IrOpKind::Inst)
+            build.mov(inst.regX64, qword[inst.regX64 + qwordReg(regOp(inst.a)) * sizeof(TString*) + offsetof(global_State, ttname)]);
+        else if (inst.a.kind == IrOpKind::Constant)
+            build.mov(inst.regX64, qword[inst.regX64 + tagOp(inst.a) * sizeof(TString*) + offsetof(global_State, ttname)]);
+        else
+            LUAU_ASSERT(!"Unsupported instruction form");
+        break;
+    }
+    case IrCmd::GET_TYPEOF:
+    {
+        IrCallWrapperX64 callWrap(regs, build);
+        callWrap.addArgument(SizeX64::qword, rState);
+        callWrap.addArgument(SizeX64::qword, luauRegAddress(vmRegOp(inst.a)));
+        callWrap.call(qword[rNativeContext + offsetof(NativeContext, luaT_objtypenamestr)]);
+
+        inst.regX64 = regs.takeReg(rax, index);
+        break;
+    }
 
     // Pseudo instructions
     case IrCmd::NOP:
@@ -1376,7 +1399,7 @@ void IrLoweringX64::finishFunction()
     for (InterruptHandler& handler : interruptHandlers)
     {
         build.setLabel(handler.self);
-        build.mov(rax, handler.pcpos + 1);
+        build.mov(eax, handler.pcpos + 1);
         build.lea(rbx, handler.next);
         build.jmp(helpers.interrupt);
     }

CodeGen/src/IrLoweringX64.h

@@ -17,7 +17,6 @@ namespace CodeGen
 {
 
 struct ModuleHelpers;
-struct NativeState;
 struct AssemblyOptions;
 
 namespace X64
@@ -25,7 +24,7 @@ namespace X64
 
 struct IrLoweringX64
 {
-    IrLoweringX64(AssemblyBuilderX64& build, ModuleHelpers& helpers, NativeState& data, IrFunction& function);
+    IrLoweringX64(AssemblyBuilderX64& build, ModuleHelpers& helpers, IrFunction& function);
 
     void lowerInst(IrInst& inst, uint32_t index, IrBlock& next);
     void finishBlock();
@@ -63,7 +62,6 @@ struct IrLoweringX64
 
     AssemblyBuilderX64& build;
     ModuleHelpers& helpers;
-    NativeState& data;
 
     IrFunction& function;
 

CodeGen/src/IrTranslateBuiltins.cpp

@@ -344,8 +344,10 @@ static BuiltinImplResult translateBuiltinType(IrBuilder& build, int nparams, int
     if (nparams < 1 || nresults > 1)
         return {BuiltinImplType::None, -1};
 
-    build.inst(IrCmd::FASTCALL, build.constUint(LBF_TYPE), build.vmReg(ra), build.vmReg(arg), args, build.constInt(1), build.constInt(1));
+    IrOp tag = build.inst(IrCmd::LOAD_TAG, build.vmReg(arg));
+    IrOp name = build.inst(IrCmd::GET_TYPE, tag);
 
+    build.inst(IrCmd::STORE_POINTER, build.vmReg(ra), name);
     build.inst(IrCmd::STORE_TAG, build.vmReg(ra), build.constTag(LUA_TSTRING));
 
     return {BuiltinImplType::UsesFallback, 1};
@@ -356,8 +358,9 @@ static BuiltinImplResult translateBuiltinTypeof(IrBuilder& build, int nparams, i
     if (nparams < 1 || nresults > 1)
         return {BuiltinImplType::None, -1};
 
-    build.inst(IrCmd::FASTCALL, build.constUint(LBF_TYPEOF), build.vmReg(ra), build.vmReg(arg), args, build.constInt(1), build.constInt(1));
+    IrOp name = build.inst(IrCmd::GET_TYPEOF, build.vmReg(arg));
 
+    build.inst(IrCmd::STORE_POINTER, build.vmReg(ra), name);
     build.inst(IrCmd::STORE_TAG, build.vmReg(ra), build.constTag(LUA_TSTRING));
 
     return {BuiltinImplType::UsesFallback, 1};

CodeGen/src/IrTranslation.cpp

@@ -825,7 +825,7 @@ void translateInstSetTableN(IrBuilder& build, const Instruction* pc, int pcpos)
     IrOp tva = build.inst(IrCmd::LOAD_TVALUE, build.vmReg(ra));
     build.inst(IrCmd::STORE_TVALUE, arrEl, tva);
 
-    build.inst(IrCmd::BARRIER_TABLE_FORWARD, vb, build.vmReg(ra));
+    build.inst(IrCmd::BARRIER_TABLE_FORWARD, vb, build.vmReg(ra), build.undef());
 
     IrOp next = build.blockAtInst(pcpos + 1);
     FallbackStreamScope scope(build, fallback, next);
@@ -902,7 +902,7 @@ void translateInstSetTable(IrBuilder& build, const Instruction* pc, int pcpos)
     IrOp tva = build.inst(IrCmd::LOAD_TVALUE, build.vmReg(ra));
     build.inst(IrCmd::STORE_TVALUE, arrEl, tva);
 
-    build.inst(IrCmd::BARRIER_TABLE_FORWARD, vb, build.vmReg(ra));
+    build.inst(IrCmd::BARRIER_TABLE_FORWARD, vb, build.vmReg(ra), build.undef());
 
     IrOp next = build.blockAtInst(pcpos + 1);
     FallbackStreamScope scope(build, fallback, next);
@@ -989,7 +989,7 @@ void translateInstSetTableKS(IrBuilder& build, const Instruction* pc, int pcpos)
     IrOp tva = build.inst(IrCmd::LOAD_TVALUE, build.vmReg(ra));
     build.inst(IrCmd::STORE_NODE_VALUE_TV, addrSlotEl, tva);
 
-    build.inst(IrCmd::BARRIER_TABLE_FORWARD, vb, build.vmReg(ra));
+    build.inst(IrCmd::BARRIER_TABLE_FORWARD, vb, build.vmReg(ra), build.undef());
 
     IrOp next = build.blockAtInst(pcpos + 2);
     FallbackStreamScope scope(build, fallback, next);
@@ -1036,7 +1036,7 @@ void translateInstSetGlobal(IrBuilder& build, const Instruction* pc, int pcpos)
     IrOp tva = build.inst(IrCmd::LOAD_TVALUE, build.vmReg(ra));
     build.inst(IrCmd::STORE_NODE_VALUE_TV, addrSlotEl, tva);
 
-    build.inst(IrCmd::BARRIER_TABLE_FORWARD, env, build.vmReg(ra));
+    build.inst(IrCmd::BARRIER_TABLE_FORWARD, env, build.vmReg(ra), build.undef());
 
     IrOp next = build.blockAtInst(pcpos + 2);
     FallbackStreamScope scope(build, fallback, next);

CodeGen/src/IrUtils.cpp

@@ -159,6 +159,9 @@ IrValueKind getCmdValueKind(IrCmd cmd)
         return IrValueKind::Int;
     case IrCmd::INVOKE_LIBM:
         return IrValueKind::Double;
+    case IrCmd::GET_TYPE:
+    case IrCmd::GET_TYPEOF:
+        return IrValueKind::Pointer;
     }
 
     LUAU_UNREACHABLE();

CodeGen/src/IrValueLocationTracking.cpp

@@ -108,6 +108,7 @@ void IrValueLocationTracking::beforeInstLowering(IrInst& inst)
     case IrCmd::FALLBACK_SETTABLEKS:
     case IrCmd::FALLBACK_PREPVARARGS:
     case IrCmd::ADJUST_STACK_TO_TOP:
+    case IrCmd::GET_TYPEOF:
         break;
 
         // These instrucitons read VmReg only after optimizeMemoryOperandsX64

CodeGen/src/OptimizeConstProp.cpp

@@ -732,6 +732,8 @@ static void constPropInInst(ConstPropState& state, IrBuilder& build, IrFunction&
                 // If the written object is not collectable, barrier is not required
                 if (!isGCO(tag))
                     kill(function, inst);
+                else
+                    replace(function, inst.c, build.constTag(tag));
             }
         }
         break;
@@ -820,6 +822,8 @@ static void constPropInInst(ConstPropState& state, IrBuilder& build, IrFunction&
     case IrCmd::BITCOUNTLZ_UINT:
     case IrCmd::BITCOUNTRZ_UINT:
     case IrCmd::INVOKE_LIBM:
+    case IrCmd::GET_TYPE:
+    case IrCmd::GET_TYPEOF:
         break;
 
     case IrCmd::JUMP_CMP_ANY:

Common/include/Luau/DenseHash.h

@@ -33,7 +33,7 @@ public:
     class const_iterator;
     class iterator;
 
-    DenseHashTable(const Key& empty_key, size_t buckets = 0)
+    explicit DenseHashTable(const Key& empty_key, size_t buckets = 0)
         : data(nullptr)
         , capacity(0)
         , count(0)
@@ -477,7 +477,7 @@ public:
     typedef typename Impl::const_iterator const_iterator;
     typedef typename Impl::iterator iterator;
 
-    DenseHashSet(const Key& empty_key, size_t buckets = 0)
+    explicit DenseHashSet(const Key& empty_key, size_t buckets = 0)
         : impl(empty_key, buckets)
     {
     }
@@ -546,7 +546,7 @@ public:
     typedef typename Impl::const_iterator const_iterator;
     typedef typename Impl::iterator iterator;
 
-    DenseHashMap(const Key& empty_key, size_t buckets = 0)
+    explicit DenseHashMap(const Key& empty_key, size_t buckets = 0)
         : impl(empty_key, buckets)
     {
     }
@@ -584,6 +584,22 @@ public:
         return impl.find(key) != 0;
     }
 
+    std::pair<Value&, bool> try_insert(const Key& key, const Value& value)
+    {
+        impl.rehash_if_full(key);
+
+        size_t before = impl.size();
+        std::pair<Key, Value>* slot = impl.insert_unsafe(key);
+
+        // Value is fresh if container count has increased
+        bool fresh = impl.size() > before;
+
+        if (fresh)
+            slot->second = value;
+
+        return std::make_pair(std::ref(slot->second), fresh);
+    }
+
     size_t size() const
     {
         return impl.size();

Makefile

@@ -161,7 +161,7 @@ clean:
 	rm -rf $(BUILD)
 	rm -rf $(EXECUTABLE_ALIASES)
 
-coverage: $(TESTS_TARGET)
+coverage: $(TESTS_TARGET) $(COMPILE_CLI_TARGET)
 	$(TESTS_TARGET)
 	mv default.profraw tests.profraw
 	$(TESTS_TARGET) --fflags=true
@@ -170,7 +170,11 @@ coverage: $(TESTS_TARGET)
 	mv default.profraw codegen.profraw
 	$(TESTS_TARGET) -ts=Conformance --codegen --fflags=true
 	mv default.profraw codegen-flags.profraw
-	llvm-profdata merge tests.profraw tests-flags.profraw codegen.profraw codegen-flags.profraw -o default.profdata
+	$(COMPILE_CLI_TARGET) --codegennull --target=a64 tests/conformance
+	mv default.profraw codegen-a64.profraw
+	$(COMPILE_CLI_TARGET) --codegennull --target=x64 tests/conformance
+	mv default.profraw codegen-x64.profraw
+	llvm-profdata merge *.profraw -o default.profdata
 	rm *.profraw
 	llvm-cov show -format=html -show-instantiations=false -show-line-counts=true -show-region-summary=false -ignore-filename-regex=\(tests\|extern\|CLI\)/.* -output-dir=coverage --instr-profile default.profdata build/coverage/luau-tests
 	llvm-cov report -ignore-filename-regex=\(tests\|extern\|CLI\)/.* -show-region-summary=false --instr-profile default.profdata build/coverage/luau-tests

Sources.cmake

@@ -88,6 +88,7 @@ target_sources(Luau.CodeGen PRIVATE
     CodeGen/src/CodeAllocator.cpp
     CodeGen/src/CodeBlockUnwind.cpp
     CodeGen/src/CodeGen.cpp
+    CodeGen/src/CodeGenAssembly.cpp
     CodeGen/src/CodeGenUtils.cpp
     CodeGen/src/CodeGenA64.cpp
     CodeGen/src/CodeGenX64.cpp
@@ -115,6 +116,7 @@ target_sources(Luau.CodeGen PRIVATE
 
     CodeGen/src/BitUtils.h
     CodeGen/src/ByteUtils.h
+    CodeGen/src/CodeGenLower.h
     CodeGen/src/CodeGenUtils.h
     CodeGen/src/CodeGenA64.h
     CodeGen/src/CodeGenX64.h

tests/CostModel.test.cpp

@@ -31,7 +31,7 @@ static uint64_t modelFunction(const char* source)
     AstStatFunction* func = result.root->body.data[0]->as<AstStatFunction>();
     REQUIRE(func);
 
-    return Luau::Compile::modelCost(func->func->body, func->func->args.data, func->func->args.size, {nullptr});
+    return Luau::Compile::modelCost(func->func->body, func->func->args.data, func->func->args.size, DenseHashMap<AstExprCall*, int>{nullptr});
 }
 
 TEST_CASE("Expression")

tests/IrBuilder.test.cpp

@@ -1005,9 +1005,9 @@ TEST_CASE_FIXTURE(IrBuilderFixture, "SkipUselessBarriers")
 
     build.inst(IrCmd::STORE_TAG, build.vmReg(0), build.constTag(tnumber));
     IrOp table = build.inst(IrCmd::LOAD_POINTER, build.vmReg(1));
-    build.inst(IrCmd::BARRIER_TABLE_FORWARD, table, build.vmReg(0));
+    build.inst(IrCmd::BARRIER_TABLE_FORWARD, table, build.vmReg(0), build.undef());
     IrOp something = build.inst(IrCmd::LOAD_POINTER, build.vmReg(2));
-    build.inst(IrCmd::BARRIER_OBJ, something, build.vmReg(0));
+    build.inst(IrCmd::BARRIER_OBJ, something, build.vmReg(0), build.undef());
     build.inst(IrCmd::RETURN, build.constUint(0));
 
     updateUseCounts(build.function);

tests/Module.test.cpp

@@ -409,9 +409,6 @@ TEST_CASE_FIXTURE(BuiltinsFixture, "do_not_clone_reexports")
 {
     ScopedFastFlag flags[] = {
         {"LuauClonePublicInterfaceLess2", true},
-        {"LuauSubstitutionReentrant", true},
-        {"LuauClassTypeVarsInSubstitution", true},
-        {"LuauSubstitutionFixMissingFields", true},
     };
 
     fileResolver.source["Module/A"] = R"(
@@ -447,9 +444,6 @@ TEST_CASE_FIXTURE(BuiltinsFixture, "do_not_clone_types_of_reexported_values")
 {
     ScopedFastFlag flags[] = {
         {"LuauClonePublicInterfaceLess2", true},
-        {"LuauSubstitutionReentrant", true},
-        {"LuauClassTypeVarsInSubstitution", true},
-        {"LuauSubstitutionFixMissingFields", true},
     };
 
     fileResolver.source["Module/A"] = R"(

tests/TypeFamily.test.cpp

@@ -140,8 +140,8 @@ TEST_CASE_FIXTURE(FamilyFixture, "unsolvable_family")
         local b = impossible(true)
     )");
 
-    LUAU_REQUIRE_ERROR_COUNT(4, result);
+    LUAU_REQUIRE_ERROR_COUNT(2, result);
-    for (size_t i = 0; i < 4; ++i)
+    for (size_t i = 0; i < 2; ++i)
     {
         CHECK(toString(result.errors[i]) == "Type family instance Swap<a> is uninhabited");
     }

tests/TypeInfer.aliases.test.cpp

@@ -8,7 +8,6 @@
 using namespace Luau;
 
 LUAU_FASTFLAG(DebugLuauDeferredConstraintResolution)
-LUAU_FASTFLAG(LuauTypeMismatchInvarianceInError)
 
 TEST_SUITE_BEGIN("TypeAliases");
 
@@ -199,15 +198,9 @@ TEST_CASE_FIXTURE(Fixture, "generic_aliases")
 
     LUAU_REQUIRE_ERROR_COUNT(1, result);
 
-    const char* expectedError;
+    const char* expectedError = "Type 'bad' could not be converted into 'T<number>'\n"
-    if (FFlag::LuauTypeMismatchInvarianceInError)
+                                "caused by:\n"
-        expectedError = "Type 'bad' could not be converted into 'T<number>'\n"
+                                "  Property 'v' is not compatible. Type 'string' could not be converted into 'number' in an invariant context";
-                        "caused by:\n"
-                        "  Property 'v' is not compatible. Type 'string' could not be converted into 'number' in an invariant context";
-    else
-        expectedError = "Type 'bad' could not be converted into 'T<number>'\n"
-                        "caused by:\n"
-                        "  Property 'v' is not compatible. Type 'string' could not be converted into 'number'";
 
     CHECK(result.errors[0].location == Location{{4, 31}, {4, 44}});
     CHECK(toString(result.errors[0]) == expectedError);
@@ -226,19 +219,11 @@ TEST_CASE_FIXTURE(Fixture, "dependent_generic_aliases")
 
     LUAU_REQUIRE_ERROR_COUNT(1, result);
 
-    std::string expectedError;
+    std::string expectedError = "Type 'bad' could not be converted into 'U<number>'\n"
-    if (FFlag::LuauTypeMismatchInvarianceInError)
+                                "caused by:\n"
-        expectedError = "Type 'bad' could not be converted into 'U<number>'\n"
+                                "  Property 't' is not compatible. Type '{ v: string }' could not be converted into 'T<number>'\n"
-                        "caused by:\n"
+                                "caused by:\n"
-                        "  Property 't' is not compatible. Type '{ v: string }' could not be converted into 'T<number>'\n"
+                                "  Property 'v' is not compatible. Type 'string' could not be converted into 'number' in an invariant context";
-                        "caused by:\n"
-                        "  Property 'v' is not compatible. Type 'string' could not be converted into 'number' in an invariant context";
-    else
-        expectedError = "Type 'bad' could not be converted into 'U<number>'\n"
-                        "caused by:\n"
-                        "  Property 't' is not compatible. Type '{ v: string }' could not be converted into 'T<number>'\n"
-                        "caused by:\n"
-                        "  Property 'v' is not compatible. Type 'string' could not be converted into 'number'";
 
     CHECK(result.errors[0].location == Location{{4, 31}, {4, 52}});
     CHECK(toString(result.errors[0]) == expectedError);

tests/TypeInfer.anyerror.test.cpp

@@ -108,10 +108,7 @@ TEST_CASE_FIXTURE(Fixture, "for_in_loop_iterator_is_error2")
         end
     )");
 
-    if (FFlag::DebugLuauDeferredConstraintResolution)
+    LUAU_REQUIRE_ERROR_COUNT(1, result);
-        LUAU_REQUIRE_ERROR_COUNT(2, result);
-    else
-        LUAU_REQUIRE_ERROR_COUNT(1, result);
 
     CHECK_EQ("*error-type*", toString(requireType("a")));
 }

tests/TypeInfer.classes.test.cpp

@@ -12,8 +12,6 @@
 using namespace Luau;
 using std::nullopt;
 
-LUAU_FASTFLAG(LuauTypeMismatchInvarianceInError);
-
 TEST_SUITE_BEGIN("TypeInferClasses");
 
 TEST_CASE_FIXTURE(ClassFixture, "call_method_of_a_class")
@@ -462,14 +460,9 @@ local b: B = a
     )");
 
     LUAU_REQUIRE_ERRORS(result);
-    if (FFlag::LuauTypeMismatchInvarianceInError)
+    CHECK_EQ(toString(result.errors[0]), R"(Type 'A' could not be converted into 'B'
-        CHECK_EQ(toString(result.errors[0]), R"(Type 'A' could not be converted into 'B'
 caused by:
   Property 'x' is not compatible. Type 'ChildClass' could not be converted into 'BaseClass' in an invariant context)");
-    else
-        CHECK_EQ(toString(result.errors[0]), R"(Type 'A' could not be converted into 'B'
-caused by:
-  Property 'x' is not compatible. Type 'ChildClass' could not be converted into 'BaseClass')");
 }
 
 TEST_CASE_FIXTURE(ClassFixture, "callable_classes")

tests/TypeInfer.functions.test.cpp

@@ -1312,10 +1312,6 @@ f(function(x) return x * 2 end)
 
 TEST_CASE_FIXTURE(Fixture, "variadic_any_is_compatible_with_a_generic_TypePack")
 {
-    ScopedFastFlag sff[] = {
-        {"LuauVariadicAnyCanBeGeneric", true}
-    };
-
     CheckResult result = check(R"(
         --!strict
         local function f(...) return ... end
@@ -1328,8 +1324,6 @@ TEST_CASE_FIXTURE(Fixture, "variadic_any_is_compatible_with_a_generic_TypePack")
 // https://github.com/Roblox/luau/issues/767
 TEST_CASE_FIXTURE(BuiltinsFixture, "variadic_any_is_compatible_with_a_generic_TypePack_2")
 {
-    ScopedFastFlag sff{"LuauVariadicAnyCanBeGeneric", true};
-
     CheckResult result = check(R"(
         local function somethingThatsAny(...: any)
             print(...)
@@ -1920,8 +1914,6 @@ TEST_CASE_FIXTURE(BuiltinsFixture, "dont_assert_when_the_tarjan_limit_is_exceede
     ScopedFastFlag sff[] = {
         {"DebugLuauDeferredConstraintResolution", true},
         {"LuauClonePublicInterfaceLess2", true},
-        {"LuauSubstitutionReentrant", true},
-        {"LuauSubstitutionFixMissingFields", true},
         {"LuauCloneSkipNonInternalVisit", true},
     };
 
@@ -2089,4 +2081,19 @@ TEST_CASE_FIXTURE(Fixture, "attempt_to_call_an_intersection_of_tables")
         CHECK_EQ(toString(result.errors[0]), "Cannot call non-function {| x: number |}");
 }
 
+TEST_CASE_FIXTURE(BuiltinsFixture, "attempt_to_call_an_intersection_of_tables_with_call_metamethod")
+{
+    CheckResult result = check(R"(
+        type Callable = typeof(setmetatable({}, {
+            __call = function(self, ...) return ... end
+        }))
+
+        local function f(t: Callable & { x: number })
+            t()
+        end
+    )");
+
+    LUAU_REQUIRE_NO_ERRORS(result);
+}
+
 TEST_SUITE_END();

tests/TypeInfer.generics.test.cpp

@@ -10,7 +10,6 @@
 #include "doctest.h"
 
 LUAU_FASTFLAG(LuauInstantiateInSubtyping)
-LUAU_FASTFLAG(LuauTypeMismatchInvarianceInError)
 
 using namespace Luau;
 
@@ -725,24 +724,12 @@ y.a.c = y
     )");
 
     LUAU_REQUIRE_ERRORS(result);
-    if (FFlag::LuauTypeMismatchInvarianceInError)
+    CHECK_EQ(toString(result.errors[0]),
-    {
+        R"(Type 'y' could not be converted into 'T<string>'
-        CHECK_EQ(toString(result.errors[0]),
-            R"(Type 'y' could not be converted into 'T<string>'
 caused by:
   Property 'a' is not compatible. Type '{ c: T<string>?, d: number }' could not be converted into 'U<string>'
 caused by:
   Property 'd' is not compatible. Type 'number' could not be converted into 'string' in an invariant context)");
-    }
-    else
-    {
-        CHECK_EQ(toString(result.errors[0]),
-            R"(Type 'y' could not be converted into 'T<string>'
-caused by:
-  Property 'a' is not compatible. Type '{ c: T<string>?, d: number }' could not be converted into 'U<string>'
-caused by:
-  Property 'd' is not compatible. Type 'number' could not be converted into 'string')");
-    }
 }
 
 TEST_CASE_FIXTURE(Fixture, "generic_type_pack_unification1")

tests/TypeInfer.intersectionTypes.test.cpp

@@ -539,10 +539,6 @@ TEST_CASE_FIXTURE(Fixture, "intersection_of_tables_with_top_properties")
 
 TEST_CASE_FIXTURE(Fixture, "intersection_of_tables_with_never_properties")
 {
-    ScopedFastFlag sffs[]{
-        {"LuauUninhabitedSubAnything2", true},
-    };
-
     CheckResult result = check(R"(
         local x : { p : number?, q : never } & { p : never, q : string? } -- OK
         local y : { p : never, q : never } = x -- OK

tests/TypeInfer.modules.test.cpp

@@ -11,7 +11,6 @@
 #include "doctest.h"
 
 LUAU_FASTFLAG(LuauInstantiateInSubtyping)
-LUAU_FASTFLAG(LuauTypeMismatchInvarianceInError)
 LUAU_FASTFLAG(DebugLuauDeferredConstraintResolution)
 
 using namespace Luau;
@@ -410,14 +409,9 @@ local b: B.T = a
     CheckResult result = frontend.check("game/C");
     LUAU_REQUIRE_ERROR_COUNT(1, result);
 
-    if (FFlag::LuauTypeMismatchInvarianceInError)
+    CHECK_EQ(toString(result.errors[0]), R"(Type 'T' from 'game/A' could not be converted into 'T' from 'game/B'
-        CHECK_EQ(toString(result.errors[0]), R"(Type 'T' from 'game/A' could not be converted into 'T' from 'game/B'
 caused by:
   Property 'x' is not compatible. Type 'number' could not be converted into 'string' in an invariant context)");
-    else
-        CHECK_EQ(toString(result.errors[0]), R"(Type 'T' from 'game/A' could not be converted into 'T' from 'game/B'
-caused by:
-  Property 'x' is not compatible. Type 'number' could not be converted into 'string')");
 }
 
 TEST_CASE_FIXTURE(BuiltinsFixture, "module_type_conflict_instantiated")
@@ -449,14 +443,9 @@ local b: B.T = a
     CheckResult result = frontend.check("game/D");
     LUAU_REQUIRE_ERROR_COUNT(1, result);
 
-    if (FFlag::LuauTypeMismatchInvarianceInError)
+    CHECK_EQ(toString(result.errors[0]), R"(Type 'T' from 'game/B' could not be converted into 'T' from 'game/C'
-        CHECK_EQ(toString(result.errors[0]), R"(Type 'T' from 'game/B' could not be converted into 'T' from 'game/C'
 caused by:
   Property 'x' is not compatible. Type 'number' could not be converted into 'string' in an invariant context)");
-    else
-        CHECK_EQ(toString(result.errors[0]), R"(Type 'T' from 'game/B' could not be converted into 'T' from 'game/C'
-caused by:
-  Property 'x' is not compatible. Type 'number' could not be converted into 'string')");
 }
 
 TEST_CASE_FIXTURE(BuiltinsFixture, "constrained_anyification_clone_immutable_types")

tests/TypeInfer.oop.test.cpp

@@ -26,17 +26,8 @@ TEST_CASE_FIXTURE(Fixture, "dont_suggest_using_colon_rather_than_dot_if_not_defi
         someTable.Function1() -- Argument count mismatch
     )");
 
-    if (FFlag::DebugLuauDeferredConstraintResolution)
+    LUAU_REQUIRE_ERROR_COUNT(1, result);
-    {
+    REQUIRE(get<CountMismatch>(result.errors[0]));
-        LUAU_REQUIRE_ERROR_COUNT(2, result);
-        CHECK(toString(result.errors[0]) == "No overload for function accepts 0 arguments.");
-        CHECK(toString(result.errors[1]) == "Available overloads: <a>(a) -> ()");
-    }
-    else
-    {
-        LUAU_REQUIRE_ERROR_COUNT(1, result);
-        REQUIRE(get<CountMismatch>(result.errors[0]));
-    }
 }
 
 TEST_CASE_FIXTURE(Fixture, "dont_suggest_using_colon_rather_than_dot_if_it_wont_help_2")
@@ -50,17 +41,8 @@ TEST_CASE_FIXTURE(Fixture, "dont_suggest_using_colon_rather_than_dot_if_it_wont_
         someTable.Function2() -- Argument count mismatch
     )");
 
-    if (FFlag::DebugLuauDeferredConstraintResolution)
+    LUAU_REQUIRE_ERROR_COUNT(1, result);
-    {
+    REQUIRE(get<CountMismatch>(result.errors[0]));
-        LUAU_REQUIRE_ERROR_COUNT(2, result);
-        CHECK(toString(result.errors[0]) == "No overload for function accepts 0 arguments.");
-        CHECK(toString(result.errors[1]) == "Available overloads: <a, b>(a, b) -> ()");
-    }
-    else
-    {
-        LUAU_REQUIRE_ERROR_COUNT(1, result);
-        REQUIRE(get<CountMismatch>(result.errors[0]));
-    }
 }
 
 TEST_CASE_FIXTURE(Fixture, "dont_suggest_using_colon_rather_than_dot_if_another_overload_works")

tests/TypeInfer.operators.test.cpp

@@ -1238,4 +1238,21 @@ TEST_CASE_FIXTURE(Fixture, "add_type_family_works")
     CHECK(toString(result.errors[0]) == "Type family instance Add<string, string> is uninhabited");
 }
 
+TEST_CASE_FIXTURE(BuiltinsFixture, "normalize_strings_comparison")
+{
+    CheckResult result = check(R"(
+local function sortKeysForPrinting(a: any, b)
+	local typeofA = type(a)
+	local typeofB = type(b)
+	-- strings and numbers are sorted numerically/alphabetically
+	if typeofA == typeofB and (typeofA == "number" or typeofA == "string") then
+		return a < b
+	end
+	-- sort the rest by type name
+	return typeofA < typeofB
+end
+)");
+    LUAU_REQUIRE_NO_ERRORS(result);
+}
+
 TEST_SUITE_END();

tests/TypeInfer.provisional.test.cpp

@@ -9,8 +9,6 @@
 
 using namespace Luau;
 
-LUAU_FASTFLAG(LuauTypeMismatchInvarianceInError)
-
 TEST_SUITE_BEGIN("ProvisionalTests");
 
 // These tests check for behavior that differs from the final behavior we'd
@@ -793,20 +791,10 @@ TEST_CASE_FIXTURE(Fixture, "assign_table_with_refined_property_with_a_similar_ty
 
     LUAU_REQUIRE_ERROR_COUNT(1, result);
 
-    if (FFlag::LuauTypeMismatchInvarianceInError)
+    CHECK_EQ(R"(Type '{| x: number? |}' could not be converted into '{| x: number |}'
-    {
-        CHECK_EQ(R"(Type '{| x: number? |}' could not be converted into '{| x: number |}'
 caused by:
   Property 'x' is not compatible. Type 'number?' could not be converted into 'number' in an invariant context)",
-            toString(result.errors[0]));
+        toString(result.errors[0]));
-    }
-    else
-    {
-        CHECK_EQ(R"(Type '{| x: number? |}' could not be converted into '{| x: number |}'
-caused by:
-  Property 'x' is not compatible. Type 'number?' could not be converted into 'number')",
-            toString(result.errors[0]));
-    }
 }
 
 TEST_CASE_FIXTURE(BuiltinsFixture, "table_insert_with_a_singleton_argument")
@@ -856,10 +844,6 @@ TEST_CASE_FIXTURE(Fixture, "lookup_prop_of_intersection_containing_unions_of_tab
 
 TEST_CASE_FIXTURE(Fixture, "expected_type_should_be_a_helpful_deduction_guide_for_function_calls")
 {
-    ScopedFastFlag sffs[]{
-        {"LuauTypeMismatchInvarianceInError", true},
-    };
-
     CheckResult result = check(R"(
         type Ref<T> = { val: T }
 
@@ -947,10 +931,6 @@ TEST_CASE_FIXTURE(Fixture, "unify_more_complex_unions_that_include_nil")
 
 TEST_CASE_FIXTURE(Fixture, "optional_class_instances_are_invariant")
 {
-    ScopedFastFlag sff[] = {
-        {"LuauTypeMismatchInvarianceInError", true}
-    };
-
     createSomeClasses(&frontend);
 
     CheckResult result = check(R"(

tests/TypeInfer.tables.test.cpp

@@ -17,7 +17,6 @@ using namespace Luau;
 LUAU_FASTFLAG(LuauLowerBoundsCalculation);
 LUAU_FASTFLAG(DebugLuauDeferredConstraintResolution);
 LUAU_FASTFLAG(LuauInstantiateInSubtyping)
-LUAU_FASTFLAG(LuauTypeMismatchInvarianceInError)
 
 TEST_SUITE_BEGIN("TableTests");
 
@@ -2077,14 +2076,9 @@ local b: B = a
     )");
 
     LUAU_REQUIRE_ERRORS(result);
-    if (FFlag::LuauTypeMismatchInvarianceInError)
+    CHECK_EQ(toString(result.errors[0]), R"(Type 'A' could not be converted into 'B'
-        CHECK_EQ(toString(result.errors[0]), R"(Type 'A' could not be converted into 'B'
 caused by:
   Property 'y' is not compatible. Type 'number' could not be converted into 'string' in an invariant context)");
-    else
-        CHECK_EQ(toString(result.errors[0]), R"(Type 'A' could not be converted into 'B'
-caused by:
-  Property 'y' is not compatible. Type 'number' could not be converted into 'string')");
 }
 
 TEST_CASE_FIXTURE(Fixture, "error_detailed_prop_nested")
@@ -2101,18 +2095,11 @@ local b: B = a
     )");
 
     LUAU_REQUIRE_ERRORS(result);
-    if (FFlag::LuauTypeMismatchInvarianceInError)
+    CHECK_EQ(toString(result.errors[0]), R"(Type 'A' could not be converted into 'B'
-        CHECK_EQ(toString(result.errors[0]), R"(Type 'A' could not be converted into 'B'
 caused by:
   Property 'b' is not compatible. Type 'AS' could not be converted into 'BS'
 caused by:
   Property 'y' is not compatible. Type 'number' could not be converted into 'string' in an invariant context)");
-    else
-        CHECK_EQ(toString(result.errors[0]), R"(Type 'A' could not be converted into 'B'
-caused by:
-  Property 'b' is not compatible. Type 'AS' could not be converted into 'BS'
-caused by:
-  Property 'y' is not compatible. Type 'number' could not be converted into 'string')");
 }
 
 TEST_CASE_FIXTURE(BuiltinsFixture, "error_detailed_metatable_prop")
@@ -2128,18 +2115,11 @@ local c2: typeof(a2) = b2
     )");
 
     LUAU_REQUIRE_ERROR_COUNT(2, result);
-    if (FFlag::LuauTypeMismatchInvarianceInError)
+    CHECK_EQ(toString(result.errors[0]), R"(Type 'b1' could not be converted into 'a1'
-        CHECK_EQ(toString(result.errors[0]), R"(Type 'b1' could not be converted into 'a1'
 caused by:
   Type '{ x: number, y: string }' could not be converted into '{ x: number, y: number }'
 caused by:
   Property 'y' is not compatible. Type 'string' could not be converted into 'number' in an invariant context)");
-    else
-        CHECK_EQ(toString(result.errors[0]), R"(Type 'b1' could not be converted into 'a1'
-caused by:
-  Type '{ x: number, y: string }' could not be converted into '{ x: number, y: number }'
-caused by:
-  Property 'y' is not compatible. Type 'string' could not be converted into 'number')");
 
     if (FFlag::LuauInstantiateInSubtyping)
     {
@@ -2170,14 +2150,9 @@ TEST_CASE_FIXTURE(Fixture, "error_detailed_indexer_key")
     )");
 
     LUAU_REQUIRE_ERRORS(result);
-    if (FFlag::LuauTypeMismatchInvarianceInError)
+    CHECK_EQ(toString(result.errors[0]), R"(Type 'A' could not be converted into 'B'
-        CHECK_EQ(toString(result.errors[0]), R"(Type 'A' could not be converted into 'B'
 caused by:
   Property '[indexer key]' is not compatible. Type 'number' could not be converted into 'string' in an invariant context)");
-    else
-        CHECK_EQ(toString(result.errors[0]), R"(Type 'A' could not be converted into 'B'
-caused by:
-  Property '[indexer key]' is not compatible. Type 'number' could not be converted into 'string')");
 }
 
 TEST_CASE_FIXTURE(Fixture, "error_detailed_indexer_value")
@@ -2191,14 +2166,9 @@ TEST_CASE_FIXTURE(Fixture, "error_detailed_indexer_value")
     )");
 
     LUAU_REQUIRE_ERRORS(result);
-    if (FFlag::LuauTypeMismatchInvarianceInError)
+    CHECK_EQ(toString(result.errors[0]), R"(Type 'A' could not be converted into 'B'
-        CHECK_EQ(toString(result.errors[0]), R"(Type 'A' could not be converted into 'B'
 caused by:
   Property '[indexer value]' is not compatible. Type 'number' could not be converted into 'string' in an invariant context)");
-    else
-        CHECK_EQ(toString(result.errors[0]), R"(Type 'A' could not be converted into 'B'
-caused by:
-  Property '[indexer value]' is not compatible. Type 'number' could not be converted into 'string')");
 }
 
 TEST_CASE_FIXTURE(Fixture, "explicitly_typed_table")
@@ -2871,10 +2841,20 @@ TEST_CASE_FIXTURE(BuiltinsFixture, "table_call_metamethod_must_be_callable")
     )");
 
     LUAU_REQUIRE_ERROR_COUNT(1, result);
-    CHECK(result.errors[0] == TypeError{
+
-                                  Location{{5, 20}, {5, 21}},
+    if (FFlag::DebugLuauDeferredConstraintResolution)
-                                  CannotCallNonFunction{builtinTypes->numberType},
+    {
-                              });
+        CHECK("Cannot call non-function { @metatable { __call: number }, {  } }" == toString(result.errors[0]));
+    }
+    else
+    {
+        TypeError e{
+            Location{{5, 20}, {5, 21}},
+            CannotCallNonFunction{builtinTypes->numberType},
+        };
+
+        CHECK(result.errors[0] == e);
+    }
 }
 
 TEST_CASE_FIXTURE(BuiltinsFixture, "table_call_metamethod_generic")

tests/TypeInfer.test.cpp

@@ -1291,4 +1291,45 @@ TEST_CASE_FIXTURE(BuiltinsFixture, "convoluted_case_where_two_TypeVars_were_boun
     // If this code does not crash, we are in good shape.
 }
 
+/*
+ * Under DCR we had an issue where constraint resolution resulted in the
+ * following:
+ *
+ * *blocked-55* ~ hasProp {- name: *blocked-55* -}, "name"
+ *
+ * This is a perfectly reasonable constraint, but one that doesn't actually
+ * constrain anything.  When we encounter a constraint like this, we need to
+ * replace the result type by a free type that is scoped to the enclosing table.
+ *
+ * Conceptually, it's simplest to think of this constraint as one that is
+ * tautological.  It does not actually contribute any new information.
+ */
+TEST_CASE_FIXTURE(Fixture, "handle_self_referential_HasProp_constraints")
+{
+    CheckResult result = check(R"(
+        local function calculateTopBarHeight(props)
+        end
+        local function isTopPage(props)
+            local topMostOpaquePage
+            if props.avatarRoute then
+                topMostOpaquePage = props.avatarRoute.opaque.name
+            else
+                topMostOpaquePage = props.opaquePage
+            end
+        end
+
+        function TopBarContainer:updateTopBarHeight(prevProps, prevState)
+            calculateTopBarHeight(self.props)
+            isTopPage(self.props)
+            local topMostOpaquePage
+            if self.props.avatarRoute then
+                topMostOpaquePage = self.props.avatarRoute.opaque.name
+                --                  ^--------------------------------^
+            else
+                topMostOpaquePage = self.props.opaquePage
+            end
+        end
+    )");
+}
+
 TEST_SUITE_END();

tests/TypeInfer.tryUnify.test.cpp

@@ -161,10 +161,6 @@ TEST_CASE_FIXTURE(TryUnifyFixture, "uninhabited_intersection_sub_anything")
 
 TEST_CASE_FIXTURE(TryUnifyFixture, "uninhabited_table_sub_never")
 {
-    ScopedFastFlag sffs[]{
-        {"LuauUninhabitedSubAnything2", true},
-    };
-
     CheckResult result = check(R"(
         function f(arg : { prop : string & number }) : never
           return arg
@@ -175,10 +171,6 @@ TEST_CASE_FIXTURE(TryUnifyFixture, "uninhabited_table_sub_never")
 
 TEST_CASE_FIXTURE(TryUnifyFixture, "uninhabited_table_sub_anything")
 {
-    ScopedFastFlag sffs[]{
-        {"LuauUninhabitedSubAnything2", true},
-    };
-
     CheckResult result = check(R"(
         function f(arg : { prop : string & number }) : boolean
           return arg

tools/faillist.txt

@@ -5,8 +5,6 @@ BuiltinTests.assert_removes_falsy_types2
 BuiltinTests.assert_removes_falsy_types_even_from_type_pack_tail_but_only_for_the_first_type
 BuiltinTests.assert_returns_false_and_string_iff_it_knows_the_first_argument_cannot_be_truthy
 BuiltinTests.bad_select_should_not_crash
-BuiltinTests.gmatch_definition
-BuiltinTests.math_max_checks_for_numbers
 BuiltinTests.select_slightly_out_of_range
 BuiltinTests.select_way_out_of_range
 BuiltinTests.set_metatable_needs_arguments
@@ -16,6 +14,10 @@ BuiltinTests.string_format_correctly_ordered_types
 BuiltinTests.string_format_report_all_type_errors_at_correct_positions
 BuiltinTests.string_format_tostring_specifier_type_constraint
 BuiltinTests.string_format_use_correct_argument2
+BuiltinTests.table_pack
+BuiltinTests.table_pack_reduce
+BuiltinTests.table_pack_variadic
+DefinitionTests.class_definition_indexer
 DefinitionTests.class_definition_overload_metamethods
 DefinitionTests.class_definition_string_props
 GenericsTests.better_mismatch_error_messages
@@ -71,6 +73,7 @@ TableTests.expected_indexer_value_type_extra_2
 TableTests.explicitly_typed_table
 TableTests.explicitly_typed_table_with_indexer
 TableTests.fuzz_table_unify_instantiated_table
+TableTests.fuzz_table_unify_instantiated_table_with_prop_realloc
 TableTests.generic_table_instantiation_potential_regression
 TableTests.give_up_after_one_metatable_index_look_up
 TableTests.indexer_on_sealed_table_must_unify_with_free_table
@@ -93,6 +96,7 @@ TableTests.shared_selfs
 TableTests.shared_selfs_from_free_param
 TableTests.shared_selfs_through_metatables
 TableTests.table_call_metamethod_basic
+TableTests.table_call_metamethod_generic
 TableTests.table_simple_call
 TableTests.table_subtyping_with_missing_props_dont_report_multiple_errors
 TableTests.used_colon_instead_of_dot
@@ -127,6 +131,7 @@ TypeInfer.tc_after_error_recovery_no_replacement_name_in_error
 TypeInfer.type_infer_recursion_limit_no_ice
 TypeInfer.type_infer_recursion_limit_normalizer
 TypeInferAnyError.for_in_loop_iterator_is_any2
+TypeInferClasses.callable_classes
 TypeInferClasses.class_type_mismatch_with_name_conflict
 TypeInferClasses.index_instance_property
 TypeInferFunctions.cannot_hoist_interior_defns_into_signature
@@ -161,8 +166,6 @@ TypeInferModules.module_type_conflict
 TypeInferModules.module_type_conflict_instantiated
 TypeInferOOP.inferring_hundreds_of_self_calls_should_not_suffocate_memory
 TypeInferOOP.methods_are_topologically_sorted
-TypeInferOperators.CallAndOrOfFunctions
-TypeInferOperators.CallOrOfFunctions
 TypeInferOperators.cli_38355_recursive_union
 TypeInferOperators.compound_assign_mismatch_metatable
 TypeInferOperators.disallow_string_and_types_without_metatables_from_arithmetic_binary_ops
@@ -179,8 +182,6 @@ TypePackTests.detect_cyclic_typepacks2
 TypePackTests.pack_tail_unification_check
 TypePackTests.type_alias_backwards_compatible
 TypePackTests.type_alias_default_type_errors
-TypePackTests.unify_variadic_tails_in_arguments
-TypePackTests.variadic_packs
 TypeSingletons.function_call_with_singletons
 TypeSingletons.function_call_with_singletons_mismatch
 TypeSingletons.no_widening_from_callsites
@@ -192,5 +193,4 @@ TypeSingletons.widening_happens_almost_everywhere
 UnionTypes.dont_allow_cyclic_unions_to_be_inferred
 UnionTypes.generic_function_with_optional_arg
 UnionTypes.index_on_a_union_type_with_missing_property
-UnionTypes.optional_union_follow
 UnionTypes.table_union_write_indirect