https://github.com/mentlerd updated https://github.com/llvm/llvm-project/pull/111892
>From b1e6178b1130135262884d99262716fcc0ada86e Mon Sep 17 00:00:00 2001 From: David Mentler <david.ment...@shapr3d.com> Date: Mon, 7 Oct 2024 21:46:50 +0200 Subject: [PATCH 1/5] Make existing tests break --- .../data-formatter-stl/libcxx/function/main.cpp | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/function/main.cpp b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/function/main.cpp index ef7c97470652fc..b17aba3fbbc70e 100644 --- a/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/function/main.cpp +++ b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/function/main.cpp @@ -1,5 +1,10 @@ #include <functional> +template<typename = bool, typename = int> +struct Dummy { + // Used to make lambda host function's symbol more complex +}; + int foo(int x, int y) { return x + y - 1; } @@ -18,7 +23,7 @@ struct Bar { } } ; -int foo2() { +int foo2(Dummy<> dummy = {}) { auto f = [](int x) { return x+1; }; >From d976756f47781d58c0367835a56b5b8e9130191b Mon Sep 17 00:00:00 2001 From: David Mentler <david.ment...@shapr3d.com> Date: Wed, 9 Oct 2024 20:40:03 +0200 Subject: [PATCH 2/5] Expose conversion from DeclContext to Decl --- lldb/include/lldb/Symbol/CompilerDeclContext.h | 2 ++ lldb/include/lldb/Symbol/TypeSystem.h | 4 ++++ lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp | 7 +++++++ lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.h | 2 ++ lldb/source/Symbol/CompilerDeclContext.cpp | 6 ++++++ 5 files changed, 21 insertions(+) diff --git a/lldb/include/lldb/Symbol/CompilerDeclContext.h b/lldb/include/lldb/Symbol/CompilerDeclContext.h index 89b4a9787688bc..3954ddd8e52087 100644 --- a/lldb/include/lldb/Symbol/CompilerDeclContext.h +++ b/lldb/include/lldb/Symbol/CompilerDeclContext.h @@ -110,6 +110,8 @@ class CompilerDeclContext { ConstString GetScopeQualifiedName() const; + CompilerDecl GetDecl() const; + private: TypeSystem *m_type_system = nullptr; void *m_opaque_decl_ctx = nullptr; diff --git a/lldb/include/lldb/Symbol/TypeSystem.h b/lldb/include/lldb/Symbol/TypeSystem.h index 7d48f9b316138c..4b531a83dd5fbb 100644 --- a/lldb/include/lldb/Symbol/TypeSystem.h +++ b/lldb/include/lldb/Symbol/TypeSystem.h @@ -134,6 +134,10 @@ class TypeSystem : public PluginInterface, virtual lldb::LanguageType DeclContextGetLanguage(void *opaque_decl_ctx) = 0; + virtual CompilerDecl DeclContextGetDecl(void *opaque_decl_ctx) { + return CompilerDecl(); + } + /// Returns the direct parent context of specified type virtual CompilerDeclContext GetCompilerDeclContextForType(const CompilerType &type); diff --git a/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp b/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp index fe0c53a7e9a3ea..678ac8d65331b7 100644 --- a/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp +++ b/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp @@ -9595,6 +9595,13 @@ TypeSystemClang::DeclContextGetLanguage(void *opaque_decl_ctx) { return eLanguageTypeUnknown; } +CompilerDecl TypeSystemClang::DeclContextGetDecl(void *opaque_decl_ctx) { + if (auto *decl_ctx = (clang::DeclContext *)opaque_decl_ctx) + if (auto* decl = dyn_cast_or_null<clang::Decl>(decl_ctx)) + return CompilerDecl(this, decl); + return CompilerDecl(); +} + static bool IsClangDeclContext(const CompilerDeclContext &dc) { return dc.IsValid() && isa<TypeSystemClang>(dc.GetTypeSystem()); } diff --git a/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.h b/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.h index e39aedec7e3902..e7ac6e320f20b7 100644 --- a/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.h +++ b/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.h @@ -595,6 +595,8 @@ class TypeSystemClang : public TypeSystem { lldb::LanguageType DeclContextGetLanguage(void *opaque_decl_ctx) override; + CompilerDecl DeclContextGetDecl(void *opaque_decl_ctx) override; + std::vector<lldb_private::CompilerContext> DeclContextGetCompilerContext(void *opaque_decl_ctx) override; diff --git a/lldb/source/Symbol/CompilerDeclContext.cpp b/lldb/source/Symbol/CompilerDeclContext.cpp index b40a08e9b1953d..7a72c4301b0b5c 100644 --- a/lldb/source/Symbol/CompilerDeclContext.cpp +++ b/lldb/source/Symbol/CompilerDeclContext.cpp @@ -34,6 +34,12 @@ ConstString CompilerDeclContext::GetScopeQualifiedName() const { return ConstString(); } +CompilerDecl CompilerDeclContext::GetDecl() const { + if (IsValid()) + return m_type_system->DeclContextGetDecl(m_opaque_decl_ctx); + return CompilerDecl(); +} + bool CompilerDeclContext::IsClassMethod() { if (IsValid()) return m_type_system->DeclContextIsClassMethod(m_opaque_decl_ctx); >From 2e75bb0aac608d626fba5878315e9118bfcf9c8a Mon Sep 17 00:00:00 2001 From: David Mentler <david.ment...@shapr3d.com> Date: Wed, 9 Oct 2024 20:41:04 +0200 Subject: [PATCH 3/5] IsPolymorphicClass already checks for record types and handles pointers to typenames --- .../CPlusPlus/ItaniumABI/ItaniumABILanguageRuntime.cpp | 8 -------- 1 file changed, 8 deletions(-) diff --git a/lldb/source/Plugins/LanguageRuntime/CPlusPlus/ItaniumABI/ItaniumABILanguageRuntime.cpp b/lldb/source/Plugins/LanguageRuntime/CPlusPlus/ItaniumABI/ItaniumABILanguageRuntime.cpp index 4c547afe30fe81..4ac1e565cb2619 100644 --- a/lldb/source/Plugins/LanguageRuntime/CPlusPlus/ItaniumABI/ItaniumABILanguageRuntime.cpp +++ b/lldb/source/Plugins/LanguageRuntime/CPlusPlus/ItaniumABI/ItaniumABILanguageRuntime.cpp @@ -191,14 +191,6 @@ llvm::Error ItaniumABILanguageRuntime::TypeHasVTable(CompilerType type) { type = pointee_type; } - // Make sure this is a class or a struct first by checking the type class - // bitfield that gets returned. - if ((type.GetTypeClass() & (eTypeClassStruct | eTypeClassClass)) == 0) { - return llvm::createStringError(std::errc::invalid_argument, - "type \"%s\" is not a class or struct or a pointer to one", - original_type.GetTypeName().AsCString("<invalid>")); - } - // Check if the type has virtual functions by asking it if it is polymorphic. if (!type.IsPolymorphicClass()) { return llvm::createStringError(std::errc::invalid_argument, >From 248be8c68221edc03ee0691b3b2c83eda0789324 Mon Sep 17 00:00:00 2001 From: David Mentler <david.ment...@shapr3d.com> Date: Wed, 9 Oct 2024 22:23:48 +0200 Subject: [PATCH 4/5] Digging for wrapped callable works --- .../CPlusPlus/CPPLanguageRuntime.cpp | 132 +++++++++++++++++- 1 file changed, 131 insertions(+), 1 deletion(-) diff --git a/lldb/source/Plugins/LanguageRuntime/CPlusPlus/CPPLanguageRuntime.cpp b/lldb/source/Plugins/LanguageRuntime/CPlusPlus/CPPLanguageRuntime.cpp index faa05e8f834ea1..7167cd4dc75f90 100644 --- a/lldb/source/Plugins/LanguageRuntime/CPlusPlus/CPPLanguageRuntime.cpp +++ b/lldb/source/Plugins/LanguageRuntime/CPlusPlus/CPPLanguageRuntime.cpp @@ -174,6 +174,137 @@ CPPLanguageRuntime::FindLibCppStdFunctionCallableInfo( if (!valobj_sp) return optional_info; + // std::function has many variants, try to disambiguate + ValueObjectSP func_as_base_ptr; + { + ValueObjectSP outer_f = valobj_sp->GetChildMemberWithName("__f_"); + + if (!outer_f) + return optional_info; // Unrecognized implementation + + if (outer_f->IsPointerType()) { + // git: 3e519524c118651123eecf60c2bbc5d65ad9bac3 + // + // class function<_Rp()> { + // aligned_storage<3*sizeof(void*)>::type __buf_; + // __base<_Rp>* __f_; + // } + + func_as_base_ptr = std::move(outer_f); + } else if (auto inner_f = outer_f->GetChildMemberWithName("__f_")) { + // git: 050b064f15ee56ee0b42c9b957a3dd0f32532394 + // + // class function<_Rp(_ArgTypes...)> { + // __value_func<_Rp(_ArgTypes...)> __f_; + // } + // + // class __value_func<_Rp(_ArgTypes...)> { + // aligned_storage<3 * sizeof(void*)>::type __buf_; + // __base<_Rp(_ArgTypes...)>* __f_; + // } + + func_as_base_ptr = std::move(inner_f); + } else + return optional_info; // Unrecognized implementation + } + + // __base<...> is a pure virtual class with an interface to create/copy/destroy/invoke + // the underlying value. This interface is implemented by partial specializations of the + // __func<_Fp, _Alloc, ...> template where _Fp is the wrapped functor object + Status status; + ValueObjectSP func_as_base = func_as_base_ptr->Dereference(status); + if (status.Fail()) + return optional_info; + + // First we'll try to extract the __func<...> template instantiation's type by looking up + // the declarations of the member function pointers in it's vtable + CompilerType func_type; + { + ValueObjectSP vtable = func_as_base->GetVTable(); + + llvm::Expected<uint32_t> num_entries = vtable->GetNumChildren(); + if (num_entries.takeError()) + return optional_info; + + // __base is pure virtual, __func is final. All member function pointers are equally + // good candidates to find the enclosing class. + // + // In practice the first two vtable entries point to artificial destructors which the + // type system refuses to elaborate as their artificial specifications are not added + // to the enclosing class' declaration context. This causes various warnings, and dont + // get us any closer to the concrete type thus we skip them. + for (uint32_t idx = 2; idx < *num_entries; idx++) { + ValueObjectSP entry = vtable->GetChildAtIndex(idx); + + // Points to a potentially interesting member function + addr_t mfunc_load_addr = entry->GetValueAsUnsigned(0); + if (!mfunc_load_addr) + continue; + + Address mfunc_symbol_addr; + if (!valobj_sp->GetTargetSP()->ResolveLoadAddress(mfunc_load_addr, mfunc_symbol_addr)) + continue; + + Function* func = mfunc_symbol_addr.CalculateSymbolContextFunction(); + if (!func) + continue; + + CompilerDeclContext mfunc_decl_ctx = func->GetDeclContext(); + if (!mfunc_decl_ctx.IsClassMethod()) + continue; + + // Member functions are contained in their enclosing class' decl context + CompilerDeclContext mfunc_parent = mfunc_decl_ctx.GetDecl().GetDeclContext(); + if (!mfunc_parent.IsValid()) + continue; + + func_type = mfunc_parent.GetDecl().GetType(); + break; + } + } + + // Regardless of what std::function wraps we are looking for the load address of a function to call + std::optional<addr_t> target_func_load_addr; + + if (CompilerType callable_type = func_type.GetTypeTemplateArgument(0)) { + if (callable_type.IsFunctionPointerType() || callable_type.IsMemberFunctionPointerType()) { + // TODO: The previous implementation just does raw pointer arithmetic and reads + // 'a pointer' to a function right after the vtable. + // + // What is the preferred approach? Go digging for the compressed_pair.first in __func + // or assume layout citing ABI compatibility requirements? + } else if (callable_type.IsRecordType()) { + // Target is a lambda, or a generic callable. Search for a single operator() overload + std::optional<ConstString> mangled_func_name; + + for (uint32_t idx = 0; idx < callable_type.GetNumMemberFunctions(); idx++) { + TypeMemberFunctionImpl mfunc = callable_type.GetMemberFunctionAtIndex(idx); + + if (mfunc.GetKind() != eMemberFunctionKindInstanceMethod) + continue; + + if (mfunc.GetName() != "operator()") + continue; + + if (mangled_func_name) + return optional_info; // Cannot resolve ambiguous target + + mangled_func_name = mfunc.GetMangledName(); + } + + // TODO: The SymbolFile did a bunch of work to reconstruct `callable_type`, + // including it's member functions. Surely it knows there they are loaded? + } + } else { + // TODO: What if we don't have debug info for callable_type? Do we fallback to + // treating the std::function as wrapping a function/member function pointer + // due to lack of options, or give up to avoid guessing wrong? + } + + if (!target_func_load_addr) + return optional_info; + + // Member __f_ has type __base*, the contents of which will hold: // 1) a vtable entry which may hold type information needed to discover the // lambda being called @@ -232,7 +363,6 @@ CPPLanguageRuntime::FindLibCppStdFunctionCallableInfo( return optional_info; uint32_t address_size = process->GetAddressByteSize(); - Status status; // First item pointed to by __f_ should be the pointer to the vtable for // a __base object. >From e7612f5c6417ababebfb7509d1dba697f07250db Mon Sep 17 00:00:00 2001 From: David Mentler <david.ment...@shapr3d.com> Date: Thu, 10 Oct 2024 23:08:01 +0200 Subject: [PATCH 5/5] RecordType callable identification works, more questions --- .../CPlusPlus/CPPLanguageRuntime.cpp | 113 ++++++++++++------ 1 file changed, 76 insertions(+), 37 deletions(-) diff --git a/lldb/source/Plugins/LanguageRuntime/CPlusPlus/CPPLanguageRuntime.cpp b/lldb/source/Plugins/LanguageRuntime/CPlusPlus/CPPLanguageRuntime.cpp index 7167cd4dc75f90..a3c719528e4e4b 100644 --- a/lldb/source/Plugins/LanguageRuntime/CPlusPlus/CPPLanguageRuntime.cpp +++ b/lldb/source/Plugins/LanguageRuntime/CPlusPlus/CPPLanguageRuntime.cpp @@ -219,6 +219,7 @@ CPPLanguageRuntime::FindLibCppStdFunctionCallableInfo( // First we'll try to extract the __func<...> template instantiation's type by looking up // the declarations of the member function pointers in it's vtable CompilerType func_type; + Address func_method_addr; { ValueObjectSP vtable = func_as_base->GetVTable(); @@ -241,11 +242,10 @@ CPPLanguageRuntime::FindLibCppStdFunctionCallableInfo( if (!mfunc_load_addr) continue; - Address mfunc_symbol_addr; - if (!valobj_sp->GetTargetSP()->ResolveLoadAddress(mfunc_load_addr, mfunc_symbol_addr)) + if (!valobj_sp->GetTargetSP()->ResolveLoadAddress(mfunc_load_addr, func_method_addr)) continue; - Function* func = mfunc_symbol_addr.CalculateSymbolContextFunction(); + Function* func = func_method_addr.CalculateSymbolContextFunction(); if (!func) continue; @@ -263,47 +263,86 @@ CPPLanguageRuntime::FindLibCppStdFunctionCallableInfo( } } - // Regardless of what std::function wraps we are looking for the load address of a function to call - std::optional<addr_t> target_func_load_addr; - - if (CompilerType callable_type = func_type.GetTypeTemplateArgument(0)) { - if (callable_type.IsFunctionPointerType() || callable_type.IsMemberFunctionPointerType()) { - // TODO: The previous implementation just does raw pointer arithmetic and reads - // 'a pointer' to a function right after the vtable. - // - // What is the preferred approach? Go digging for the compressed_pair.first in __func - // or assume layout citing ABI compatibility requirements? - } else if (callable_type.IsRecordType()) { - // Target is a lambda, or a generic callable. Search for a single operator() overload - std::optional<ConstString> mangled_func_name; + CompilerType callable_type = func_type.GetTypeTemplateArgument(0); + if (!callable_type) + return optional_info; + + if (callable_type.IsFunctionPointerType() || callable_type.IsMemberFunctionPointerType()) { + // TODO: The previous implementation just does raw pointer arithmetic and reads + // 'a pointer' to a function right after the vtable. + // + // What is the preferred approach? Go digging for the compressed_pair.first in __func + // or assume layout citing ABI compatibility requirements? + } else if (callable_type.IsRecordType()) { + // Target is a lambda, or a generic callable. Search for a single operator() overload + std::optional<ConstString> mangled_func_name; + + // TODO: I am still not sure whether it is a good idea to reconstruct the full type + // here.. it seems there are handy FindFunctions that could perhaps to a good job + // at locating candidates. However even when limiting the search to the decl_ctx of + // the class the code seems to iterate over way more DIEs than I expected. What to do? + + // TODO: Because we have access to the type we know a _lot_ about callable_type, we + // could even extract a ValueObjectSP to it if we wanted. It would be cool to make + // std::function have a synt children provider showing the wrapped lambda/callable! + + for (uint32_t idx = 0; idx < callable_type.GetNumMemberFunctions(); idx++) { + TypeMemberFunctionImpl mfunc = callable_type.GetMemberFunctionAtIndex(idx); - for (uint32_t idx = 0; idx < callable_type.GetNumMemberFunctions(); idx++) { - TypeMemberFunctionImpl mfunc = callable_type.GetMemberFunctionAtIndex(idx); - - if (mfunc.GetKind() != eMemberFunctionKindInstanceMethod) - continue; - - if (mfunc.GetName() != "operator()") - continue; - - if (mangled_func_name) - return optional_info; // Cannot resolve ambiguous target + if (mfunc.GetKind() != eMemberFunctionKindInstanceMethod) + continue; + + if (mfunc.GetName() != "operator()") + continue; - mangled_func_name = mfunc.GetMangledName(); - } + if (mangled_func_name) + return optional_info; // Cannot resolve ambiguous target + + mangled_func_name = mfunc.GetMangledName(); + } + + // Locate the symbol context corresponding to the target function + SymbolContext sc; + { + // We'll assume that callable_type is in the same module as the vtable + ModuleSP mod = func_method_addr.CalculateSymbolContextModule(); + + // Limit our lookup to callable_type + CompilerDeclContext decl_ctx = callable_type.GetTypeSystem()->GetCompilerDeclContextForType(callable_type); - // TODO: The SymbolFile did a bunch of work to reconstruct `callable_type`, - // including it's member functions. Surely it knows there they are loaded? + SymbolContextList list; + mod->FindFunctions(*mangled_func_name, decl_ctx, eFunctionNameTypeFull, {}, list); + + if (list.GetSize() != 1) + return optional_info; + + list.GetContextAtIndex(0, sc); } - } else { - // TODO: What if we don't have debug info for callable_type? Do we fallback to - // treating the std::function as wrapping a function/member function pointer - // due to lack of options, or give up to avoid guessing wrong? + + // TODO: This feels a bit clunky, I am probably misusing the API? FindFunctions returns me + // SymbolContexts with the .function set but not .symbol ... At first glance it seemed like + // if we know the function there must be a symbol too! + if (!sc.function) + return optional_info; + + Symbol* symbol = sc.function->GetAddressRange().GetBaseAddress().CalculateSymbolContextSymbol(); + if (!symbol) + return optional_info; + + return LibCppStdFunctionCallableInfo { + .callable_symbol = *symbol, + .callable_address = symbol->GetAddress(), + .callable_line_entry = sc.GetFunctionStartLineEntry(), + + // TODO: Can't tell lambdas apart from generic callables.. do we really need to? + // Is it important to have the correct qualification in the summary? + .callable_case = LibCppStdFunctionCallableCase::Lambda + }; } - if (!target_func_load_addr) + // Unrecognized callable type - skip the original implementation for now + if (!callable_type.IsVoidType()) return optional_info; - // Member __f_ has type __base*, the contents of which will hold: // 1) a vtable entry which may hold type information needed to discover the _______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits