ClickHouse · nikitamikhaylov · Dec 23, 2024 · Sep 16, 2024 · Sep 17, 2024 · Sep 17, 2024
diff --git a/base/base/AlignedUnion.h b/base/base/AlignedUnion.h
@@ -0,0 +1,16 @@
+#pragma once
+
+#include <algorithm>
+
+template<std::size_t len, class... Types>
+struct AlignedUnion
+{
+    static constexpr std::size_t alignment_value = std::max({alignof(Types)...});
+    struct Type
+    {
+        alignas(alignment_value) char s[std::max({len, sizeof(Types)...})];
+    };
+};
+
+template<std::size_t len, class... Types>
+using AlignedUnionT = typename AlignedUnion<len, Types...>::Type;
diff --git a/base/poco/Foundation/include/Poco/UUID.h b/base/poco/Foundation/include/Poco/UUID.h
@@ -21,6 +21,7 @@
 #include "Poco/Foundation.h"
 #include <Poco/Types.h>
 
+#include <array>
 
 namespace Poco
 {

diff --git a/contrib/llvm-project b/contrib/llvm-project
diff --git a/contrib/llvm-project-cmake/CMakeLists.txt b/contrib/llvm-project-cmake/CMakeLists.txt
@@ -27,7 +27,7 @@ set (LLVM_INCLUDE_DIRS
 set (LLVM_LIBRARY_DIRS "${ClickHouse_BINARY_DIR}/contrib/llvm-project/llvm")
 # NOTE: You should not remove this line since otherwise it will use default 20,
 # and llvm cannot be compiled with bundled libcxx and 20 standard.
-set (CMAKE_CXX_STANDARD 14)
+set (CMAKE_CXX_STANDARD 17)
 
 if (ARCH_AMD64)
     set (LLVM_TARGETS_TO_BUILD "X86" CACHE INTERNAL "")

diff --git a/contrib/rocksdb b/contrib/rocksdb
diff --git a/src/AggregateFunctions/SingleValueData.h b/src/AggregateFunctions/SingleValueData.h
@@ -5,6 +5,7 @@
 #include <Columns/ColumnDecimal.h>
 #include <DataTypes/DataTypeDate.h>
 #include <DataTypes/DataTypeDateTime.h>
+#include <base/AlignedUnion.h>
 #include <base/StringRef.h>
 
 namespace DB
@@ -374,7 +375,7 @@ createAggregateFunctionSingleValue(const String & name, const DataTypes & argume
 /// Helper to allocate enough memory to store any derived class
 struct SingleValueDataBaseMemoryBlock
 {
-    std::aligned_union_t<
+    AlignedUnionT<
         SingleValueDataBase::MAX_STORAGE_SIZE,
         SingleValueDataNumeric<Decimal256>, /// We check all types in generateSingleValueFromTypeIndex
         SingleValueDataString,

diff --git a/src/Common/BSONCXXHelper.h b/src/Common/BSONCXXHelper.h
@@ -59,7 +59,7 @@ static bsoncxx::types::bson_value::value fieldAsBSONValue(const Field & field, c
         case TypeIndex::DateTime:
             return bsoncxx::types::b_date{std::chrono::seconds{field.safeGet<UInt32 &>()}};
         case TypeIndex::UUID:
-            return bsoncxx::types::b_string{static_cast<String>(formatUUID(field.safeGet<UUID &>()))};
+            return bsoncxx::types::b_string{String{formatUUID(field.safeGet<UUID &>()).data()}};
         case TypeIndex::Tuple: {
             auto arr = array();
             for (const auto & elem : field.safeGet<Tuple &>())

diff --git a/src/Common/GWPAsan.cpp b/src/Common/GWPAsan.cpp
@@ -159,7 +159,7 @@ void printReport([[maybe_unused]] uintptr_t fault_address)
 {
     const auto logger = getLogger("GWPAsan");
     const auto * state = GuardedAlloc.getAllocatorState();
-    if (uintptr_t internal_error_ptr = __gwp_asan_get_internal_crash_address(state); internal_error_ptr)
+    if (uintptr_t internal_error_ptr = __gwp_asan_get_internal_crash_address(state, fault_address); internal_error_ptr)
         fault_address = internal_error_ptr;
 
     const gwp_asan::AllocationMetadata * allocation_meta = __gwp_asan_get_metadata(state, GuardedAlloc.getMetadataRegion(), fault_address);

diff --git a/src/Common/HashTable/HashTable.h b/src/Common/HashTable/HashTable.h
@@ -371,7 +371,7 @@ struct ZeroValueStorage<true, Cell>
 {
 private:
     bool has_zero = false;
-    std::aligned_storage_t<sizeof(Cell), alignof(Cell)> zero_value_storage; /// Storage of element with zero key.
+    alignas(Cell) std::byte zero_value_storage[sizeof(Cell)]; /// Storage of element with zero key.
 
 public:
     bool hasZero() const { return has_zero; }

diff --git a/src/Common/HashTable/StringHashTable.h b/src/Common/HashTable/StringHashTable.h
@@ -141,7 +141,7 @@ struct StringHashTableEmpty
     using Self = StringHashTableEmpty;
 
     bool has_zero = false;
-    std::aligned_storage_t<sizeof(Cell), alignof(Cell)> zero_value_storage; /// Storage of element with zero key.
+    alignas(Cell) std::byte zero_value_storage[sizeof(Cell)]; /// Storage of element with zero key.
 
 public:
     bool hasZero() const { return has_zero; }

diff --git a/src/Common/MatchGenerator.h b/src/Common/MatchGenerator.h
@@ -1,6 +1,7 @@
 #pragma once
 
 #include <base/types.h>
+#include <functional>
 #include <memory>
 
 namespace re2

diff --git a/src/Common/ZooKeeper/KeeperFeatureFlags.h b/src/Common/ZooKeeper/KeeperFeatureFlags.h
@@ -2,6 +2,8 @@
 
 #include <Coordination/KeeperConstants.h>
 
+#include <memory>
+
 namespace Poco
 {
 class Logger;

diff --git a/src/Coordination/Changelog.h b/src/Coordination/Changelog.h
@@ -9,6 +9,7 @@
 #include <unordered_map>
 #include <unordered_set>
 #include <future>
+#include <vector>
 
 namespace nuraft
 {

diff --git a/src/Core/Field.h b/src/Core/Field.h
@@ -7,6 +7,7 @@
 #include <type_traits>
 #include <functional>
 
+#include <base/AlignedUnion.h>
 #include <Core/CompareHelper.h>
 #include <Core/Defines.h>
 #include <Core/Types.h>
@@ -301,7 +302,6 @@ concept not_field_or_bool_or_stringlike
   */
 static constexpr auto DBMS_MIN_FIELD_SIZE = 32;
 
-
 /** Discriminated union of several types.
   * Made for replacement of `boost::variant`
   *  is not generalized,
@@ -672,7 +672,7 @@ class Field
     static Field restoreFromDump(std::string_view dump_);
 
 private:
-    std::aligned_union_t<DBMS_MIN_FIELD_SIZE - sizeof(Types::Which),
+    AlignedUnionT<DBMS_MIN_FIELD_SIZE - sizeof(Types::Which),
         Null, UInt64, UInt128, UInt256, Int64, Int128, Int256, UUID, IPv4, IPv6, Float64, String, Array, Tuple, Map,
         DecimalField<Decimal32>, DecimalField<Decimal64>, DecimalField<Decimal128>, DecimalField<Decimal256>,
         AggregateFunctionStateData, CustomType

diff --git a/src/Core/QualifiedTableName.h b/src/Core/QualifiedTableName.h
@@ -2,9 +2,10 @@
 
 #include <base/types.h>
 
+#include <optional>
 #include <string>
 #include <tuple>
-#include <optional>
+#include <vector>
 
 namespace DB
 {

diff --git a/src/Formats/MarkInCompressedFile.h b/src/Formats/MarkInCompressedFile.h
@@ -1,6 +1,7 @@
 #pragma once
 
 #include <tuple>
+#include <unordered_map>
 
 #include <base/types.h>
 #include <Common/PODArray.h>

diff --git a/src/Formats/ReadSchemaUtils.h b/src/Formats/ReadSchemaUtils.h
@@ -1,5 +1,6 @@
 #pragma once
 
+#include <IO/ReadBuffer.h>
 #include <Formats/FormatSettings.h>
 #include <Storages/Cache/SchemaCache.h>
 #include <Storages/ColumnsDescription.h>

diff --git a/src/Interpreters/ClusterProxy/executeQuery.cpp b/src/Interpreters/ClusterProxy/executeQuery.cpp
@@ -729,7 +729,7 @@ void executeQueryWithParallelReplicas(
         context, query_ast, storage_id.database_name, storage_id.table_name, /*remote_table_function_ptr*/ nullptr);
     auto header = InterpreterSelectQuery(modified_query_ast, context, SelectQueryOptions(processed_stage).analyze()).getSampleBlock();
 
-    executeQueryWithParallelReplicas(query_plan, storage_id, header, processed_stage, modified_query_ast, context, storage_limits);
+    executeQueryWithParallelReplicas(query_plan, storage_id, header, processed_stage, modified_query_ast, context, storage_limits, nullptr);
 }
 
 void executeQueryWithParallelReplicasCustomKey(

diff --git a/src/Interpreters/ClusterProxy/executeQuery.h b/src/Interpreters/ClusterProxy/executeQuery.h
@@ -87,7 +87,7 @@ void executeQueryWithParallelReplicas(
     const ASTPtr & query_ast,
     ContextPtr context,
     std::shared_ptr<const StorageLimitsList> storage_limits,
-    QueryPlanStepPtr read_from_merge_tree = nullptr);
+    QueryPlanStepPtr read_from_merge_tree);
 
 void executeQueryWithParallelReplicas(
     QueryPlan & query_plan,

diff --git a/src/Interpreters/JIT/CHJIT.cpp b/src/Interpreters/JIT/CHJIT.cpp
@@ -6,7 +6,10 @@
 
 #include <boost/noncopyable.hpp>
 
+#include <llvm/Analysis/CGSCCPassManager.h>
 #include <llvm/Analysis/TargetTransformInfo.h>
+#include <llvm/Analysis/LoopAnalysisManager.h>
+#include <llvm/Passes/PassBuilder.h>
 #include <llvm/IR/BasicBlock.h>
 #include <llvm/IR/DataLayout.h>
 #include <llvm/IR/DerivedTypes.h>
@@ -483,29 +486,21 @@ std::string CHJIT::getMangledName(const std::string & name_to_mangle) const
 
 void CHJIT::runOptimizationPassesOnModule(llvm::Module & module) const
 {
-    llvm::PassManagerBuilder pass_manager_builder;
-    llvm::legacy::PassManager mpm;
-    llvm::legacy::FunctionPassManager fpm(&module);
-    pass_manager_builder.OptLevel = 3;
-    pass_manager_builder.SLPVectorize = true;
-    pass_manager_builder.LoopVectorize = true;
-    pass_manager_builder.RerollLoops = true;
-    pass_manager_builder.VerifyInput = true;
-    pass_manager_builder.VerifyOutput = true;
-    machine->adjustPassManager(pass_manager_builder);
-
-    fpm.add(llvm::createTargetTransformInfoWrapperPass(machine->getTargetIRAnalysis()));
-    mpm.add(llvm::createTargetTransformInfoWrapperPass(machine->getTargetIRAnalysis()));
-
-    pass_manager_builder.populateFunctionPassManager(fpm);
-    pass_manager_builder.populateModulePassManager(mpm);
-
-    fpm.doInitialization();
-    for (auto & function : module)
-        fpm.run(function);
-    fpm.doFinalization();
-
-    mpm.run(module);
+    llvm::LoopAnalysisManager lam;
+    llvm::FunctionAnalysisManager fam;
+    llvm::CGSCCAnalysisManager cgam;
+    llvm::ModuleAnalysisManager mam;
+
+    llvm::PassBuilder pb;
+
+    pb.registerModuleAnalyses(mam);
+    pb.registerCGSCCAnalyses(cgam);
+    pb.registerFunctionAnalyses(fam);
+    pb.registerLoopAnalyses(lam);
+    pb.crossRegisterProxies(lam, fam, cgam, mam);
+
+    llvm::ModulePassManager mpm = pb.buildPerModuleDefaultPipeline(llvm::OptimizationLevel::O3);
+    mpm.run(module, mam);
 }
 
 std::unique_ptr<llvm::TargetMachine> CHJIT::getTargetMachine()
@@ -538,8 +533,8 @@ std::unique_ptr<llvm::TargetMachine> CHJIT::getTargetMachine()
         cpu,
         features.getString(),
         options,
-        llvm::None,
-        llvm::None,
+        std::nullopt,
+        std::nullopt,
         llvm::CodeGenOpt::Aggressive,
         jit);
 

diff --git a/src/Planner/findParallelReplicasQuery.cpp b/src/Planner/findParallelReplicasQuery.cpp
@@ -494,7 +494,8 @@ JoinTreeQueryPlan buildQueryPlanForParallelReplicas(
         processed_stage,
         modified_query_ast,
         context,
-        storage_limits);
+        storage_limits,
+        nullptr);
 
     auto converting = ActionsDAG::makeConvertingActions(
         header.getColumnsWithTypeAndName(),

diff --git a/src/Processors/Formats/Impl/DWARFBlockInputFormat.cpp b/src/Processors/Formats/Impl/DWARFBlockInputFormat.cpp
@@ -837,7 +837,7 @@ void DWARFBlockInputFormat::parseRanges(
     uint64_t offset, bool form_rnglistx, const UnitState & unit, const ColumnVector<UInt64>::MutablePtr & col_ranges_start,
     const ColumnVector<UInt64>::MutablePtr & col_ranges_end) const
 {
-    llvm::Optional<llvm::object::SectionedAddress> base_addr;
+    std::optional<llvm::object::SectionedAddress> base_addr;
     if (unit.base_address != UINT64_MAX)
         base_addr = llvm::object::SectionedAddress{.Address = unit.base_address};
 
@@ -882,7 +882,7 @@ void DWARFBlockInputFormat::parseRanges(
         if (err)
             throw Exception(ErrorCodes::CANNOT_PARSE_DWARF, "Error parsing .debug_rnglists list: {}", llvm::toString(std::move(err)));
 
-        auto lookup_addr = [&](uint32_t idx) -> llvm::Optional<llvm::object::SectionedAddress>
+        auto lookup_addr = [&](uint32_t idx) -> std::optional<llvm::object::SectionedAddress>
             {
                 uint64_t addr = fetchFromDebugAddr(unit.debug_addr_base, idx);
                 return llvm::object::SectionedAddress{.Address = addr};

diff --git a/src/Storages/MergeTree/IDataPartStorage.h b/src/Storages/MergeTree/IDataPartStorage.h
@@ -1,6 +1,7 @@
 #pragma once
 #include <IO/WriteSettings.h>
 #include <IO/WriteBufferFromFileBase.h>
+#include <IO/ReadBufferFromFileBase.h>
 #include <base/types.h>
 #include <Core/NamesAndTypes.h>
 #include <Interpreters/TransactionVersionMetadata.h>

diff --git a/src/Storages/TableZnodeInfo.h b/src/Storages/TableZnodeInfo.h
@@ -4,6 +4,8 @@
 #include <Storages/RenamingRestrictions.h>
 #include <Databases/LoadingStrictnessLevel.h>
 
+#include <memory>
+
 namespace zkutil
 {
 class ZooKeeper;