llvm-project/lldb/source/Expression/ClangExpressionDeclMap.cpp

1474 lines
47 KiB
C++
Raw Normal View History

//===-- ClangExpressionDeclMap.cpp -----------------------------*- C++ -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
#include "lldb/Expression/ClangExpressionDeclMap.h"
// C Includes
// C++ Includes
// Other libraries and framework includes
// Project includes
#include "clang/AST/DeclarationName.h"
#include "clang/AST/Decl.h"
#include "lldb/lldb-private.h"
#include "lldb/Core/Address.h"
#include "lldb/Core/Error.h"
#include "lldb/Core/Log.h"
#include "lldb/Core/Module.h"
#include "lldb/Expression/ClangASTSource.h"
#include "lldb/Expression/ClangPersistentVariables.h"
#include "lldb/Symbol/ClangASTContext.h"
#include "lldb/Symbol/ClangNamespaceDecl.h"
#include "lldb/Symbol/CompileUnit.h"
#include "lldb/Symbol/Function.h"
#include "lldb/Symbol/ObjectFile.h"
#include "lldb/Symbol/SymbolContext.h"
#include "lldb/Symbol/Type.h"
#include "lldb/Symbol/TypeList.h"
#include "lldb/Symbol/Variable.h"
#include "lldb/Symbol/VariableList.h"
#include "lldb/Target/ExecutionContext.h"
#include "lldb/Target/Process.h"
#include "lldb/Target/RegisterContext.h"
#include "lldb/Target/StackFrame.h"
#include "lldb/Target/Target.h"
#include "llvm/Support/raw_ostream.h"
using namespace lldb;
using namespace lldb_private;
using namespace clang;
ClangExpressionDeclMap::ClangExpressionDeclMap (ExecutionContext *exe_ctx) :
m_found_entities (),
m_struct_members (),
m_exe_ctx (),
m_sym_ctx (),
m_persistent_vars (NULL),
m_struct_alignment (0),
m_struct_size (0),
m_struct_laid_out (false),
m_enable_lookups (false),
m_allocated_area (0),
m_materialized_location (0),
m_result_name (),
m_object_pointer_type (),
m_lookedup_types ()
{
if (exe_ctx)
{
m_exe_ctx = *exe_ctx;
if (exe_ctx->frame)
m_sym_ctx = exe_ctx->frame->GetSymbolContext(lldb::eSymbolContextEverything);
if (exe_ctx->process)
m_persistent_vars = &exe_ctx->process->GetPersistentVariables();
}
}
ClangExpressionDeclMap::~ClangExpressionDeclMap()
{
for (uint64_t entity_index = 0, num_entities = m_found_entities.Size();
entity_index < num_entities;
++entity_index)
{
ClangExpressionVariable &entity(m_found_entities.VariableAtIndex(entity_index));
if (entity.m_parser_vars.get() &&
entity.m_parser_vars->m_lldb_value)
delete entity.m_parser_vars->m_lldb_value;
entity.DisableParserVars();
}
for (uint64_t pvar_index = 0, num_pvars = m_persistent_vars->Size();
pvar_index < num_pvars;
++pvar_index)
{
ClangExpressionVariable &pvar(m_persistent_vars->VariableAtIndex(pvar_index));
pvar.DisableParserVars();
}
if (m_materialized_location)
{
//#define SINGLE_STEP_EXPRESSIONS
#ifndef SINGLE_STEP_EXPRESSIONS
m_exe_ctx.process->DeallocateMemory(m_materialized_location);
#endif
m_materialized_location = 0;
}
}
// Interface for IRForTarget
const ConstString &
ClangExpressionDeclMap::GetPersistentResultName ()
{
if (!m_result_name)
m_persistent_vars->GetNextResultName(m_result_name);
return m_result_name;
}
bool
ClangExpressionDeclMap::AddPersistentVariable
(
const clang::NamedDecl *decl,
const ConstString &name,
TypeFromParser parser_type
)
{
clang::ASTContext *context(m_exe_ctx.target->GetScratchClangASTContext()->getASTContext());
TypeFromUser user_type(ClangASTContext::CopyType(context,
parser_type.GetASTContext(),
parser_type.GetOpaqueQualType()),
context);
if (!m_persistent_vars->CreatePersistentVariable (name, user_type))
return false;
ClangExpressionVariable *var = m_persistent_vars->GetVariable(name);
if (!var)
return false;
var->EnableParserVars();
var->m_parser_vars->m_named_decl = decl;
var->m_parser_vars->m_parser_type = parser_type;
return true;
}
bool
ClangExpressionDeclMap::AddValueToStruct
(
const clang::NamedDecl *decl,
const ConstString &name,
llvm::Value *value,
size_t size,
off_t alignment
)
{
lldb::LogSP log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS));
m_struct_laid_out = false;
if (m_struct_members.GetVariable(decl))
return true;
ClangExpressionVariable *var = m_found_entities.GetVariable(decl);
if (!var)
var = m_persistent_vars->GetVariable(decl);
if (!var)
return false;
if (log)
log->Printf("Adding value for decl %p [%s - %s] to the structure",
decl,
name.GetCString(),
var->m_name.GetCString());
// We know entity->m_parser_vars is valid because we used a parser variable
// to find it
var->m_parser_vars->m_llvm_value = value;
var->EnableJITVars();
var->m_jit_vars->m_alignment = alignment;
var->m_jit_vars->m_size = size;
m_struct_members.AddVariable(*var);
return true;
}
bool
ClangExpressionDeclMap::DoStructLayout ()
{
if (m_struct_laid_out)
return true;
off_t cursor = 0;
m_struct_alignment = 0;
m_struct_size = 0;
for (uint64_t member_index = 0, num_members = m_struct_members.Size();
member_index < num_members;
++member_index)
{
ClangExpressionVariable &member(m_struct_members.VariableAtIndex(member_index));
if (!member.m_jit_vars.get())
return false;
if (member_index == 0)
m_struct_alignment = member.m_jit_vars->m_alignment;
if (cursor % member.m_jit_vars->m_alignment)
cursor += (member.m_jit_vars->m_alignment - (cursor % member.m_jit_vars->m_alignment));
member.m_jit_vars->m_offset = cursor;
cursor += member.m_jit_vars->m_size;
}
m_struct_size = cursor;
m_struct_laid_out = true;
return true;
}
bool ClangExpressionDeclMap::GetStructInfo
(
uint32_t &num_elements,
size_t &size,
off_t &alignment
)
{
if (!m_struct_laid_out)
return false;
num_elements = m_struct_members.Size();
size = m_struct_size;
alignment = m_struct_alignment;
return true;
}
bool
ClangExpressionDeclMap::GetStructElement
(
const clang::NamedDecl *&decl,
llvm::Value *&value,
off_t &offset,
ConstString &name,
uint32_t index
)
{
if (!m_struct_laid_out)
return false;
if (index >= m_struct_members.Size())
return false;
ClangExpressionVariable &member(m_struct_members.VariableAtIndex(index));
if (!member.m_parser_vars.get() ||
!member.m_jit_vars.get())
return false;
decl = member.m_parser_vars->m_named_decl;
value = member.m_parser_vars->m_llvm_value;
offset = member.m_jit_vars->m_offset;
name = member.m_name;
return true;
}
bool
ClangExpressionDeclMap::GetFunctionInfo
(
const clang::NamedDecl *decl,
llvm::Value**& value,
uint64_t &ptr
)
{
ClangExpressionVariable *entity = m_found_entities.GetVariable(decl);
if (!entity)
return false;
// We know m_parser_vars is valid since we searched for the variable by
// its NamedDecl
value = &entity->m_parser_vars->m_llvm_value;
ptr = entity->m_parser_vars->m_lldb_value->GetScalar().ULongLong();
return true;
}
bool
ClangExpressionDeclMap::GetFunctionAddress
(
const ConstString &name,
uint64_t &ptr
)
{
// Back out in all cases where we're not fully initialized
if (m_exe_ctx.frame == NULL)
return false;
SymbolContextList sc_list;
m_sym_ctx.FindFunctionsByName(name, false, sc_list);
if (!sc_list.GetSize())
return false;
SymbolContext sym_ctx;
sc_list.GetContextAtIndex(0, sym_ctx);
const Address *fun_address;
if (sym_ctx.function)
fun_address = &sym_ctx.function->GetAddressRange().GetBaseAddress();
else if (sym_ctx.symbol)
fun_address = &sym_ctx.symbol->GetAddressRangeRef().GetBaseAddress();
else
return false;
ptr = fun_address->GetLoadAddress (m_exe_ctx.target);
return true;
}
// Interface for CommandObjectExpression
bool
ClangExpressionDeclMap::Materialize
(
ExecutionContext *exe_ctx,
lldb::addr_t &struct_address,
Error &err
)
{
bool result = DoMaterialize(false, exe_ctx, NULL, err);
if (result)
struct_address = m_materialized_location;
return result;
}
Removed the hacky "#define this ___clang_this" handler for C++ classes. Replaced it with a less hacky approach: - If an expression is defined in the context of a method of class A, then that expression is wrapped as ___clang_class::___clang_expr(void*) { ... } instead of ___clang_expr(void*) { ... }. - ___clang_class is resolved as the type of the target of the "this" pointer in the method the expression is defined in. - When reporting the type of ___clang_class, a method with the signature ___clang_expr(void*) is added to that class, so that Clang doesn't complain about a method being defined without a corresponding declaration. - Whenever the expression gets called, "this" gets looked up, type-checked, and then passed in as the first argument. This required the following changes: - The ABIs were changed to support passing of the "this" pointer as part of trivial calls. - ThreadPlanCallFunction and ClangFunction were changed to support passing of an optional "this" pointer. - ClangUserExpression was extended to perform the wrapping described above. - ClangASTSource was changed to revert the changes required by the hack. - ClangExpressionParser, IRForTarget, and ClangExpressionDeclMap were changed to handle different manglings of ___clang_expr flexibly. This meant no longer searching for a function called ___clang_expr, but rather looking for a function whose name *contains* ___clang_expr. - ClangExpressionParser and ClangExpressionDeclMap now remember whether "this" is required, and know how to look it up as necessary. A few inheritance bugs remain, and I'm trying to resolve these. But it is now possible to use "this" as well as refer implicitly to member variables, when in the proper context. llvm-svn: 114384
2010-09-21 08:44:12 +08:00
bool
ClangExpressionDeclMap::GetObjectPointer
(
lldb::addr_t &object_ptr,
ExecutionContext *exe_ctx,
Error &err
)
Removed the hacky "#define this ___clang_this" handler for C++ classes. Replaced it with a less hacky approach: - If an expression is defined in the context of a method of class A, then that expression is wrapped as ___clang_class::___clang_expr(void*) { ... } instead of ___clang_expr(void*) { ... }. - ___clang_class is resolved as the type of the target of the "this" pointer in the method the expression is defined in. - When reporting the type of ___clang_class, a method with the signature ___clang_expr(void*) is added to that class, so that Clang doesn't complain about a method being defined without a corresponding declaration. - Whenever the expression gets called, "this" gets looked up, type-checked, and then passed in as the first argument. This required the following changes: - The ABIs were changed to support passing of the "this" pointer as part of trivial calls. - ThreadPlanCallFunction and ClangFunction were changed to support passing of an optional "this" pointer. - ClangUserExpression was extended to perform the wrapping described above. - ClangASTSource was changed to revert the changes required by the hack. - ClangExpressionParser, IRForTarget, and ClangExpressionDeclMap were changed to handle different manglings of ___clang_expr flexibly. This meant no longer searching for a function called ___clang_expr, but rather looking for a function whose name *contains* ___clang_expr. - ClangExpressionParser and ClangExpressionDeclMap now remember whether "this" is required, and know how to look it up as necessary. A few inheritance bugs remain, and I'm trying to resolve these. But it is now possible to use "this" as well as refer implicitly to member variables, when in the proper context. llvm-svn: 114384
2010-09-21 08:44:12 +08:00
{
if (!exe_ctx || !exe_ctx->frame || !exe_ctx->target || !exe_ctx->process)
{
err.SetErrorString("Couldn't load 'this' because the context is incomplete");
return false;
}
if (!m_object_pointer_type.GetOpaqueQualType())
{
err.SetErrorString("Couldn't load 'this' because its type is unknown");
return false;
}
static ConstString g_this_const_str ("this");
Variable *object_ptr_var = FindVariableInScope (*exe_ctx->frame, g_this_const_str, &m_object_pointer_type);
Removed the hacky "#define this ___clang_this" handler for C++ classes. Replaced it with a less hacky approach: - If an expression is defined in the context of a method of class A, then that expression is wrapped as ___clang_class::___clang_expr(void*) { ... } instead of ___clang_expr(void*) { ... }. - ___clang_class is resolved as the type of the target of the "this" pointer in the method the expression is defined in. - When reporting the type of ___clang_class, a method with the signature ___clang_expr(void*) is added to that class, so that Clang doesn't complain about a method being defined without a corresponding declaration. - Whenever the expression gets called, "this" gets looked up, type-checked, and then passed in as the first argument. This required the following changes: - The ABIs were changed to support passing of the "this" pointer as part of trivial calls. - ThreadPlanCallFunction and ClangFunction were changed to support passing of an optional "this" pointer. - ClangUserExpression was extended to perform the wrapping described above. - ClangASTSource was changed to revert the changes required by the hack. - ClangExpressionParser, IRForTarget, and ClangExpressionDeclMap were changed to handle different manglings of ___clang_expr flexibly. This meant no longer searching for a function called ___clang_expr, but rather looking for a function whose name *contains* ___clang_expr. - ClangExpressionParser and ClangExpressionDeclMap now remember whether "this" is required, and know how to look it up as necessary. A few inheritance bugs remain, and I'm trying to resolve these. But it is now possible to use "this" as well as refer implicitly to member variables, when in the proper context. llvm-svn: 114384
2010-09-21 08:44:12 +08:00
if (!object_ptr_var)
{
err.SetErrorString("Couldn't find 'this' with appropriate type in scope");
return false;
}
std::auto_ptr<lldb_private::Value> location_value(GetVariableValue(*exe_ctx,
object_ptr_var,
m_object_pointer_type.GetASTContext()));
if (!location_value.get())
{
err.SetErrorString("Couldn't get the location for 'this'");
return false;
}
if (location_value->GetValueType() == Value::eValueTypeLoadAddress)
{
lldb::addr_t value_addr = location_value->GetScalar().ULongLong();
uint32_t address_byte_size = exe_ctx->target->GetArchitecture().GetAddressByteSize();
lldb::ByteOrder address_byte_order = exe_ctx->process->GetByteOrder();
if (ClangASTType::GetClangTypeBitWidth(m_object_pointer_type.GetASTContext(), m_object_pointer_type.GetOpaqueQualType()) != address_byte_size * 8)
{
err.SetErrorStringWithFormat("'this' is not of an expected pointer size");
return false;
}
DataBufferHeap data;
data.SetByteSize(address_byte_size);
Error read_error;
if (exe_ctx->process->ReadMemory (value_addr, data.GetBytes(), address_byte_size, read_error) != address_byte_size)
{
err.SetErrorStringWithFormat("Coldn't read 'this' from the target: %s", read_error.AsCString());
return false;
}
DataExtractor extractor(data.GetBytes(), data.GetByteSize(), address_byte_order, address_byte_size);
uint32_t offset = 0;
object_ptr = extractor.GetPointer(&offset);
return true;
}
else
{
err.SetErrorString("'this' is not in memory; LLDB must be extended to handle registers");
return false;
}
}
bool
ClangExpressionDeclMap::Dematerialize
(
ExecutionContext *exe_ctx,
ClangExpressionVariable *&result,
Error &err
)
{
return DoMaterialize(true, exe_ctx, &result, err);
}
bool
ClangExpressionDeclMap::DumpMaterializedStruct
(
ExecutionContext *exe_ctx,
Stream &s,
Error &err
)
{
if (!m_struct_laid_out)
{
err.SetErrorString("Structure hasn't been laid out yet");
return false;
}
if (!exe_ctx)
{
err.SetErrorString("Received null execution context");
return false;
}
if (!exe_ctx->process)
{
err.SetErrorString("Couldn't find the process");
return false;
}
if (!exe_ctx->target)
{
err.SetErrorString("Couldn't find the target");
return false;
}
lldb::DataBufferSP data(new DataBufferHeap(m_struct_size, 0));
Error error;
if (exe_ctx->process->ReadMemory (m_materialized_location, data->GetBytes(), data->GetByteSize(), error) != data->GetByteSize())
{
err.SetErrorStringWithFormat ("Couldn't read struct from the target: %s", error.AsCString());
return false;
}
DataExtractor extractor(data, exe_ctx->process->GetByteOrder(), exe_ctx->target->GetArchitecture().GetAddressByteSize());
for (uint64_t member_index = 0, num_members = m_struct_members.Size();
member_index < num_members;
++member_index)
{
ClangExpressionVariable &member (m_struct_members.VariableAtIndex(member_index));
s.Printf("[%s]\n", member.m_name.GetCString());
if (!member.m_jit_vars.get())
return false;
extractor.Dump(&s, // stream
member.m_jit_vars->m_offset, // offset
lldb::eFormatBytesWithASCII, // format
1, // byte size of individual entries
member.m_jit_vars->m_size, // number of entries
16, // entries per line
m_materialized_location + member.m_jit_vars->m_offset, // address to print
0, // bit size (bitfields only; 0 means ignore)
0); // bit alignment (bitfields only; 0 means ignore)
s.PutChar('\n');
}
return true;
}
bool
ClangExpressionDeclMap::DoMaterialize
(
bool dematerialize,
ExecutionContext *exe_ctx,
ClangExpressionVariable **result,
Error &err
)
{
lldb::LogSP log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS));
if (!m_struct_laid_out)
{
err.SetErrorString("Structure hasn't been laid out yet");
return LLDB_INVALID_ADDRESS;
}
if (!exe_ctx)
{
err.SetErrorString("Received null execution context");
return LLDB_INVALID_ADDRESS;
}
if (!exe_ctx->frame)
{
err.SetErrorString("Received null execution frame");
return LLDB_INVALID_ADDRESS;
}
if (!m_struct_size)
{
if (log)
log->PutCString("Not bothering to allocate a struct because no arguments are needed");
m_allocated_area = NULL;
return true;
}
const SymbolContext &sym_ctx(exe_ctx->frame->GetSymbolContext(lldb::eSymbolContextEverything));
if (!dematerialize)
{
if (m_materialized_location)
{
exe_ctx->process->DeallocateMemory(m_materialized_location);
m_materialized_location = 0;
}
if (log)
log->PutCString("Allocating memory for materialized argument struct");
lldb::addr_t mem = exe_ctx->process->AllocateMemory(m_struct_alignment + m_struct_size,
lldb::ePermissionsReadable | lldb::ePermissionsWritable,
err);
if (mem == LLDB_INVALID_ADDRESS)
return false;
m_allocated_area = mem;
}
m_materialized_location = m_allocated_area;
if (m_materialized_location % m_struct_alignment)
m_materialized_location += (m_struct_alignment - (m_materialized_location % m_struct_alignment));
for (uint64_t member_index = 0, num_members = m_struct_members.Size();
member_index < num_members;
++member_index)
{
ClangExpressionVariable &member (m_struct_members.VariableAtIndex(member_index));
if (!member.m_parser_vars.get())
return false;
ClangExpressionVariable *entity = m_found_entities.GetVariable(member.m_parser_vars->m_named_decl);
ClangExpressionVariable *persistent_variable = m_persistent_vars->GetVariable(member.m_name);
if (entity)
{
if (!member.m_jit_vars.get())
return false;
if (!DoMaterializeOneVariable(dematerialize, *exe_ctx, sym_ctx, member.m_name, member.m_user_type, m_materialized_location + member.m_jit_vars->m_offset, err))
return false;
}
else if (persistent_variable)
{
if (member.m_name == m_result_name)
{
if (!dematerialize)
continue;
if (log)
log->PutCString("Found result member in the struct");
*result = &member;
}
if (log)
log->Printf("Searched for persistent variable %s and found %s", member.m_name.GetCString(), persistent_variable->m_name.GetCString());
if (!DoMaterializeOnePersistentVariable(dematerialize, *exe_ctx, persistent_variable->m_name, m_materialized_location + member.m_jit_vars->m_offset, err))
return false;
}
else
{
err.SetErrorStringWithFormat("Unexpected variable %s", member.m_name.GetCString());
return false;
}
}
return true;
}
bool
ClangExpressionDeclMap::DoMaterializeOnePersistentVariable
(
bool dematerialize,
ExecutionContext &exe_ctx,
const ConstString &name,
lldb::addr_t addr,
Error &err
)
{
ClangExpressionVariable *pvar(m_persistent_vars->GetVariable(name));
if (!pvar)
{
err.SetErrorStringWithFormat("Undefined persistent variable %s", name.GetCString());
return LLDB_INVALID_ADDRESS;
}
size_t pvar_size = pvar->Size();
if (!pvar->m_data_sp.get())
return false;
uint8_t *pvar_data = pvar->m_data_sp->GetBytes();
Error error;
if (dematerialize)
{
if (exe_ctx.process->ReadMemory (addr, pvar_data, pvar_size, error) != pvar_size)
{
err.SetErrorStringWithFormat ("Couldn't read a composite type from the target: %s", error.AsCString());
return false;
}
}
else
{
if (exe_ctx.process->WriteMemory (addr, pvar_data, pvar_size, error) != pvar_size)
{
err.SetErrorStringWithFormat ("Couldn't write a composite type to the target: %s", error.AsCString());
return false;
}
}
return true;
}
bool
ClangExpressionDeclMap::DoMaterializeOneVariable
(
bool dematerialize,
ExecutionContext &exe_ctx,
const SymbolContext &sym_ctx,
const ConstString &name,
TypeFromUser type,
lldb::addr_t addr,
Error &err
)
{
lldb::LogSP log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS));
if (!exe_ctx.frame || !exe_ctx.process)
return false;
Variable *var = FindVariableInScope (*exe_ctx.frame, name, &type);
if (!var)
{
err.SetErrorStringWithFormat("Couldn't find %s with appropriate type", name.GetCString());
return false;
}
if (log)
log->Printf("%s %s with type %p", (dematerialize ? "Dematerializing" : "Materializing"), name.GetCString(), type.GetOpaqueQualType());
std::auto_ptr<lldb_private::Value> location_value(GetVariableValue(exe_ctx,
var,
type.GetASTContext()));
if (!location_value.get())
{
err.SetErrorStringWithFormat("Couldn't get value for %s", name.GetCString());
return false;
}
// The size of the type contained in addr
size_t addr_bit_size = ClangASTType::GetClangTypeBitWidth(type.GetASTContext(), type.GetOpaqueQualType());
size_t addr_byte_size = addr_bit_size % 8 ? ((addr_bit_size + 8) / 8) : (addr_bit_size / 8);
Value::ValueType value_type = location_value->GetValueType();
switch (value_type)
{
default:
{
StreamString ss;
location_value->Dump(&ss);
err.SetErrorStringWithFormat("%s has a value of unhandled type: %s", name.GetCString(), ss.GetString().c_str());
return false;
}
break;
case Value::eValueTypeLoadAddress:
{
lldb::addr_t value_addr = location_value->GetScalar().ULongLong();
DataBufferHeap data;
data.SetByteSize(addr_byte_size);
lldb::addr_t src_addr;
lldb::addr_t dest_addr;
if (dematerialize)
{
src_addr = addr;
dest_addr = value_addr;
}
else
{
src_addr = value_addr;
dest_addr = addr;
}
Error error;
if (exe_ctx.process->ReadMemory (src_addr, data.GetBytes(), addr_byte_size, error) != addr_byte_size)
{
err.SetErrorStringWithFormat ("Couldn't read %s from the target: %s", name.GetCString(), error.AsCString());
return false;
}
if (exe_ctx.process->WriteMemory (dest_addr, data.GetBytes(), addr_byte_size, error) != addr_byte_size)
{
err.SetErrorStringWithFormat ("Couldn't write %s to the target: %s", name.GetCString(), error.AsCString());
return false;
}
if (log)
log->Printf("Copied from 0x%llx to 0x%llx", (uint64_t)src_addr, (uint64_t)addr);
}
break;
case Value::eValueTypeScalar:
{
if (location_value->GetContextType() != Value::eContextTypeRegisterInfo)
{
StreamString ss;
location_value->Dump(&ss);
err.SetErrorStringWithFormat("%s is a scalar of unhandled type: %s", name.GetCString(), ss.GetString().c_str());
return false;
}
lldb::RegisterInfo *register_info = location_value->GetRegisterInfo();
if (!register_info)
{
err.SetErrorStringWithFormat("Couldn't get the register information for %s", name.GetCString());
return false;
}
RegisterContext *register_context = exe_ctx.GetRegisterContext();
if (!register_context)
{
err.SetErrorStringWithFormat("Couldn't read register context to read %s from %s", name.GetCString(), register_info->name);
return false;
}
uint32_t register_number = register_info->kinds[lldb::eRegisterKindLLDB];
uint32_t register_byte_size = register_info->byte_size;
if (dematerialize)
{
// Moving from addr into a register
//
// Case 1: addr_byte_size and register_byte_size are the same
//
// |AABBCCDD| Address contents
// |AABBCCDD| Register contents
//
// Case 2: addr_byte_size is bigger than register_byte_size
//
// Error! (The register should always be big enough to hold the data)
//
// Case 3: register_byte_size is bigger than addr_byte_size
//
// |AABB| Address contents
// |AABB0000| Register contents [on little-endian hardware]
// |0000AABB| Register contents [on big-endian hardware]
if (addr_byte_size > register_byte_size)
{
err.SetErrorStringWithFormat("%s is too big to store in %s", name.GetCString(), register_info->name);
return false;
}
uint32_t register_offset;
switch (exe_ctx.process->GetByteOrder())
{
default:
err.SetErrorStringWithFormat("%s is stored with an unhandled byte order", name.GetCString());
return false;
case lldb::eByteOrderLittle:
register_offset = 0;
break;
case lldb::eByteOrderBig:
register_offset = register_byte_size - addr_byte_size;
break;
}
DataBufferHeap register_data (register_byte_size, 0);
Error error;
if (exe_ctx.process->ReadMemory (addr, register_data.GetBytes() + register_offset, addr_byte_size, error) != addr_byte_size)
{
err.SetErrorStringWithFormat ("Couldn't read %s from the target: %s", name.GetCString(), error.AsCString());
return false;
}
DataExtractor register_extractor (register_data.GetBytes(), register_byte_size, exe_ctx.process->GetByteOrder(), exe_ctx.process->GetAddressByteSize());
if (!register_context->WriteRegisterBytes(register_number, register_extractor, 0))
{
err.SetErrorStringWithFormat("Couldn't read %s from %s", name.GetCString(), register_info->name);
return false;
}
}
else
{
// Moving from a register into addr
//
// Case 1: addr_byte_size and register_byte_size are the same
//
// |AABBCCDD| Register contents
// |AABBCCDD| Address contents
//
// Case 2: addr_byte_size is bigger than register_byte_size
//
// Error! (The register should always be big enough to hold the data)
//
// Case 3: register_byte_size is bigger than addr_byte_size
//
// |AABBCCDD| Register contents
// |AABB| Address contents on little-endian hardware
// |CCDD| Address contents on big-endian hardware
if (addr_byte_size > register_byte_size)
{
err.SetErrorStringWithFormat("%s is too big to store in %s", name.GetCString(), register_info->name);
return false;
}
uint32_t register_offset;
switch (exe_ctx.process->GetByteOrder())
{
default:
err.SetErrorStringWithFormat("%s is stored with an unhandled byte order", name.GetCString());
return false;
case lldb::eByteOrderLittle:
register_offset = 0;
break;
case lldb::eByteOrderBig:
register_offset = register_byte_size - addr_byte_size;
break;
}
DataExtractor register_extractor;
if (!register_context->ReadRegisterBytes(register_number, register_extractor))
{
err.SetErrorStringWithFormat("Couldn't read %s from %s", name.GetCString(), register_info->name);
return false;
}
const void *register_data = register_extractor.GetData(&register_offset, addr_byte_size);
if (!register_data)
{
err.SetErrorStringWithFormat("Read but couldn't extract data for %s from %s", name.GetCString(), register_info->name);
return false;
}
Error error;
if (exe_ctx.process->WriteMemory (addr, register_data, addr_byte_size, error) != addr_byte_size)
{
err.SetErrorStringWithFormat ("Couldn't write %s to the target: %s", error.AsCString());
return false;
}
}
}
}
return true;
}
Variable *
ClangExpressionDeclMap::FindVariableInScope
(
StackFrame &frame,
const ConstString &name,
TypeFromUser *type
)
{
lldb::LogSP log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS));
VariableList *var_list = frame.GetVariableList(true);
Fixed an expression parsing issue where if you were stopped somewhere without debug information and you evaluated an expression, a crash would occur as a result of an unchecked pointer. Added the ability to get the expression path for a ValueObject. For a rectangle point child "x" the expression path would be something like: "rect.top_left.x". This will allow GUI and command lines to get ahold of the expression path for a value object without having to explicitly know about the hierarchy. This means the ValueObject base class now has a "ValueObject *m_parent;" member. All ValueObject subclasses now correctly track their lineage and are able to provide value expression paths as well. Added a new "--flat" option to the "frame variable" to allow for flat variable output. An example of the current and new outputs: (lldb) frame variable argc = 1 argv = 0x00007fff5fbffe80 pt = { x = 2 y = 3 } rect = { bottom_left = { x = 1 y = 2 } top_right = { x = 3 y = 4 } } (lldb) frame variable --flat argc = 1 argv = 0x00007fff5fbffe80 pt.x = 2 pt.y = 3 rect.bottom_left.x = 1 rect.bottom_left.y = 2 rect.top_right.x = 3 rect.top_right.y = 4 As you can see when there is a lot of hierarchy it can help flatten things out. Also if you want to use a member in an expression, you can copy the text from the "--flat" output and not have to piece it together manually. This can help when you want to use parts of the STL in expressions: (lldb) frame variable --flat argc = 1 argv = 0x00007fff5fbffea8 hello_world._M_dataplus._M_p = 0x0000000000000000 (lldb) expr hello_world._M_dataplus._M_p[0] == '\0' llvm-svn: 116532
2010-10-15 06:52:14 +08:00
if (!var_list)
return NULL;
lldb::VariableSP var_sp (var_list->FindVariable(name));
const bool append = true;
const uint32_t max_matches = 1;
if (!var_sp)
{
// Look for globals elsewhere in the module for the frame
ModuleSP module_sp (m_exe_ctx.frame->GetSymbolContext(eSymbolContextModule).module_sp);
if (module_sp)
{
VariableList module_globals;
if (module_sp->FindGlobalVariables (name, append, max_matches, module_globals))
var_sp = module_globals.GetVariableAtIndex (0);
}
}
if (!var_sp)
{
// Look for globals elsewhere in the program (all images)
TargetSP target_sp (m_exe_ctx.frame->GetSymbolContext(eSymbolContextTarget).target_sp);
if (target_sp)
{
VariableList program_globals;
if (target_sp->GetImages().FindGlobalVariables (name, append, max_matches, program_globals))
var_sp = program_globals.GetVariableAtIndex (0);
}
}
if (var_sp && type)
{
if (type->GetASTContext() == var_sp->GetType()->GetClangAST())
{
if (!ClangASTContext::AreTypesSame(type->GetASTContext(), type->GetOpaqueQualType(), var_sp->GetType()->GetClangType()))
return NULL;
}
else
{
if (log)
log->PutCString("Skipping a candidate variable because of different AST contexts");
return NULL;
}
}
return var_sp.get();
}
// Interface for ClangASTSource
void
ClangExpressionDeclMap::GetDecls (NameSearchContext &context, const ConstString &name)
{
lldb::LogSP log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS));
Removed the hacky "#define this ___clang_this" handler for C++ classes. Replaced it with a less hacky approach: - If an expression is defined in the context of a method of class A, then that expression is wrapped as ___clang_class::___clang_expr(void*) { ... } instead of ___clang_expr(void*) { ... }. - ___clang_class is resolved as the type of the target of the "this" pointer in the method the expression is defined in. - When reporting the type of ___clang_class, a method with the signature ___clang_expr(void*) is added to that class, so that Clang doesn't complain about a method being defined without a corresponding declaration. - Whenever the expression gets called, "this" gets looked up, type-checked, and then passed in as the first argument. This required the following changes: - The ABIs were changed to support passing of the "this" pointer as part of trivial calls. - ThreadPlanCallFunction and ClangFunction were changed to support passing of an optional "this" pointer. - ClangUserExpression was extended to perform the wrapping described above. - ClangASTSource was changed to revert the changes required by the hack. - ClangExpressionParser, IRForTarget, and ClangExpressionDeclMap were changed to handle different manglings of ___clang_expr flexibly. This meant no longer searching for a function called ___clang_expr, but rather looking for a function whose name *contains* ___clang_expr. - ClangExpressionParser and ClangExpressionDeclMap now remember whether "this" is required, and know how to look it up as necessary. A few inheritance bugs remain, and I'm trying to resolve these. But it is now possible to use "this" as well as refer implicitly to member variables, when in the proper context. llvm-svn: 114384
2010-09-21 08:44:12 +08:00
if (log)
log->Printf("Hunting for a definition for '%s'", name.GetCString());
// Back out in all cases where we're not fully initialized
if (m_exe_ctx.frame == NULL)
return;
Removed the hacky "#define this ___clang_this" handler for C++ classes. Replaced it with a less hacky approach: - If an expression is defined in the context of a method of class A, then that expression is wrapped as ___clang_class::___clang_expr(void*) { ... } instead of ___clang_expr(void*) { ... }. - ___clang_class is resolved as the type of the target of the "this" pointer in the method the expression is defined in. - When reporting the type of ___clang_class, a method with the signature ___clang_expr(void*) is added to that class, so that Clang doesn't complain about a method being defined without a corresponding declaration. - Whenever the expression gets called, "this" gets looked up, type-checked, and then passed in as the first argument. This required the following changes: - The ABIs were changed to support passing of the "this" pointer as part of trivial calls. - ThreadPlanCallFunction and ClangFunction were changed to support passing of an optional "this" pointer. - ClangUserExpression was extended to perform the wrapping described above. - ClangASTSource was changed to revert the changes required by the hack. - ClangExpressionParser, IRForTarget, and ClangExpressionDeclMap were changed to handle different manglings of ___clang_expr flexibly. This meant no longer searching for a function called ___clang_expr, but rather looking for a function whose name *contains* ___clang_expr. - ClangExpressionParser and ClangExpressionDeclMap now remember whether "this" is required, and know how to look it up as necessary. A few inheritance bugs remain, and I'm trying to resolve these. But it is now possible to use "this" as well as refer implicitly to member variables, when in the proper context. llvm-svn: 114384
2010-09-21 08:44:12 +08:00
SymbolContextList sc_list;
const char *name_unique_cstr = name.GetCString();
if (name_unique_cstr == NULL)
return;
// Only look for functions by name out in our symbols if the function
// doesn't start with our phony prefix of '$'
if (name_unique_cstr[0] != '$')
{
Variable *var = FindVariableInScope(*m_exe_ctx.frame, name);
// If we found a variable in scope, no need to pull up function names
if (var != NULL)
{
AddOneVariable(context, var);
}
else
{
m_sym_ctx.FindFunctionsByName (name, false, sc_list);
bool found_specific = false;
Symbol *generic_symbol = NULL;
Symbol *non_extern_symbol = NULL;
for (uint32_t index = 0, num_indices = sc_list.GetSize();
index < num_indices;
++index)
{
SymbolContext sym_ctx;
sc_list.GetContextAtIndex(index, sym_ctx);
if (sym_ctx.function)
{
// TODO only do this if it's a C function; C++ functions may be
// overloaded
if (!found_specific)
AddOneFunction(context, sym_ctx.function, NULL);
found_specific = true;
}
else if (sym_ctx.symbol)
{
if (sym_ctx.symbol->IsExternal())
generic_symbol = sym_ctx.symbol;
else
non_extern_symbol = sym_ctx.symbol;
}
}
if (!found_specific)
{
if (generic_symbol)
AddOneFunction(context, NULL, generic_symbol);
else if (non_extern_symbol)
AddOneFunction(context, NULL, non_extern_symbol);
}
ClangNamespaceDecl namespace_decl (m_sym_ctx.FindNamespace(name));
if (namespace_decl)
{
clang::NamespaceDecl *clang_namespace_decl = AddNamespace(context, namespace_decl);
if (clang_namespace_decl)
{
// TODO: is this how we get the decl lookups to be called for
// this namespace??
clang_namespace_decl->setHasExternalLexicalStorage();
}
}
}
}
else
{
static ConstString g_lldb_class_name ("$__lldb_class");
if (name == g_lldb_class_name)
{
// Clang is looking for the type of "this"
VariableList *vars = m_exe_ctx.frame->GetVariableList(false);
if (!vars)
return;
lldb::VariableSP this_var = vars->FindVariable(ConstString("this"));
if (!this_var)
return;
Type *this_type = this_var->GetType();
if (!this_type)
return;
if (log)
{
log->PutCString ("Type for \"this\" is: ");
StreamString strm;
this_type->Dump(&strm, true);
log->PutCString (strm.GetData());
}
TypeFromUser this_user_type(this_type->GetClangType(),
this_type->GetClangAST());
m_object_pointer_type = this_user_type;
void *pointer_target_type;
if (!ClangASTContext::IsPointerType(this_user_type.GetOpaqueQualType(),
&pointer_target_type))
return;
TypeFromUser class_user_type(pointer_target_type,
this_type->GetClangAST());
AddOneType(context, class_user_type, true);
return;
}
ClangExpressionVariable *pvar(m_persistent_vars->GetVariable(name));
if (pvar)
AddOneVariable(context, pvar);
}
// See information on gating of this operation next to the definition for
// m_lookedup_types.
if (m_lookedup_types.find(name_unique_cstr) == m_lookedup_types.end())
{
// 1 The name is added to m_lookedup_types.
m_lookedup_types.insert(std::pair<const char*, bool>(name_unique_cstr, true));
// 2 The type is looked up and added, potentially causing more type loookups.
lldb::TypeSP type_sp (m_sym_ctx.FindTypeByName (name));
if (type_sp)
{
if (log)
{
log->Printf ("Matching type found for \"%s\": ", name.GetCString());
StreamString strm;
type_sp->Dump(&strm, true);
log->PutCString (strm.GetData());
}
TypeFromUser user_type(type_sp->GetClangType(),
type_sp->GetClangAST());
AddOneType(context, user_type, false);
}
// 3 The name is removed from m_lookedup_types.
m_lookedup_types.erase(name_unique_cstr);
}
}
Value *
ClangExpressionDeclMap::GetVariableValue
(
ExecutionContext &exe_ctx,
Variable *var,
clang::ASTContext *parser_ast_context,
TypeFromUser *user_type,
TypeFromParser *parser_type
)
{
lldb::LogSP log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS));
Type *var_type = var->GetType();
if (!var_type)
{
if (log)
log->PutCString("Skipped a definition because it has no type");
return NULL;
}
void *var_opaque_type = var_type->GetClangType();
if (!var_opaque_type)
{
if (log)
log->PutCString("Skipped a definition because it has no Clang type");
return NULL;
}
TypeList *type_list = var_type->GetTypeList();
if (!type_list)
{
if (log)
log->PutCString("Skipped a definition because the type has no associated type list");
return NULL;
}
clang::ASTContext *exe_ast_ctx = type_list->GetClangASTContext().getASTContext();
if (!exe_ast_ctx)
{
if (log)
log->PutCString("There is no AST context for the current execution context");
return NULL;
}
DWARFExpression &var_location_expr = var->LocationExpression();
std::auto_ptr<Value> var_location(new Value);
Looking at some of the test suite failures in DWARF in .o files with the debug map showed that the location lists in the .o files needed some refactoring in order to work. The case that was failing was where a function that was in the "__TEXT.__textcoal_nt" in the .o file, and in the "__TEXT.__text" section in the main executable. This made symbol lookup fail due to the way we were finding a real address in the debug map which was by finding the section that the function was in in the .o file and trying to find this in the main executable. Now the section list supports finding a linked address in a section or any child sections. After fixing this, we ran into issue that were due to DWARF and how it represents locations lists. DWARF makes a list of address ranges and expressions that go along with those address ranges. The location addresses are expressed in terms of a compile unit address + offset. This works fine as long as nothing moves around. When stuff moves around and offsets change between the remapped compile unit base address and the new function address, then we can run into trouble. To deal with this, we now store supply a location list slide amount to any location list expressions that will allow us to make the location list addresses into zero based offsets from the object that owns the location list (always a function in our case). With these fixes we can now re-link random address ranges inside the debugger for use with our DWARF + debug map, incremental linking, and more. Another issue that arose when doing the DWARF in the .o files was that GCC 4.2 emits a ".debug_aranges" that only mentions functions that are externally visible. This makes .debug_aranges useless to us and we now generate a real address range lookup table in the DWARF parser at the same time as we index the name tables (that are needed because .debug_pubnames is just as useless). llvm-gcc doesn't generate a .debug_aranges section, though this could be fixed, we aren't going to rely upon it. Renamed a bunch of "UINT_MAX" to "UINT32_MAX". llvm-svn: 113829
2010-09-14 10:20:48 +08:00
lldb::addr_t loclist_base_load_addr = LLDB_INVALID_ADDRESS;
if (var_location_expr.IsLocationList())
{
SymbolContext var_sc;
var->CalculateSymbolContext (&var_sc);
loclist_base_load_addr = var_sc.function->GetAddressRange().GetBaseAddress().GetLoadAddress (exe_ctx.target);
Looking at some of the test suite failures in DWARF in .o files with the debug map showed that the location lists in the .o files needed some refactoring in order to work. The case that was failing was where a function that was in the "__TEXT.__textcoal_nt" in the .o file, and in the "__TEXT.__text" section in the main executable. This made symbol lookup fail due to the way we were finding a real address in the debug map which was by finding the section that the function was in in the .o file and trying to find this in the main executable. Now the section list supports finding a linked address in a section or any child sections. After fixing this, we ran into issue that were due to DWARF and how it represents locations lists. DWARF makes a list of address ranges and expressions that go along with those address ranges. The location addresses are expressed in terms of a compile unit address + offset. This works fine as long as nothing moves around. When stuff moves around and offsets change between the remapped compile unit base address and the new function address, then we can run into trouble. To deal with this, we now store supply a location list slide amount to any location list expressions that will allow us to make the location list addresses into zero based offsets from the object that owns the location list (always a function in our case). With these fixes we can now re-link random address ranges inside the debugger for use with our DWARF + debug map, incremental linking, and more. Another issue that arose when doing the DWARF in the .o files was that GCC 4.2 emits a ".debug_aranges" that only mentions functions that are externally visible. This makes .debug_aranges useless to us and we now generate a real address range lookup table in the DWARF parser at the same time as we index the name tables (that are needed because .debug_pubnames is just as useless). llvm-gcc doesn't generate a .debug_aranges section, though this could be fixed, we aren't going to rely upon it. Renamed a bunch of "UINT_MAX" to "UINT32_MAX". llvm-svn: 113829
2010-09-14 10:20:48 +08:00
}
Error err;
Looking at some of the test suite failures in DWARF in .o files with the debug map showed that the location lists in the .o files needed some refactoring in order to work. The case that was failing was where a function that was in the "__TEXT.__textcoal_nt" in the .o file, and in the "__TEXT.__text" section in the main executable. This made symbol lookup fail due to the way we were finding a real address in the debug map which was by finding the section that the function was in in the .o file and trying to find this in the main executable. Now the section list supports finding a linked address in a section or any child sections. After fixing this, we ran into issue that were due to DWARF and how it represents locations lists. DWARF makes a list of address ranges and expressions that go along with those address ranges. The location addresses are expressed in terms of a compile unit address + offset. This works fine as long as nothing moves around. When stuff moves around and offsets change between the remapped compile unit base address and the new function address, then we can run into trouble. To deal with this, we now store supply a location list slide amount to any location list expressions that will allow us to make the location list addresses into zero based offsets from the object that owns the location list (always a function in our case). With these fixes we can now re-link random address ranges inside the debugger for use with our DWARF + debug map, incremental linking, and more. Another issue that arose when doing the DWARF in the .o files was that GCC 4.2 emits a ".debug_aranges" that only mentions functions that are externally visible. This makes .debug_aranges useless to us and we now generate a real address range lookup table in the DWARF parser at the same time as we index the name tables (that are needed because .debug_pubnames is just as useless). llvm-gcc doesn't generate a .debug_aranges section, though this could be fixed, we aren't going to rely upon it. Renamed a bunch of "UINT_MAX" to "UINT32_MAX". llvm-svn: 113829
2010-09-14 10:20:48 +08:00
if (!var_location_expr.Evaluate(&exe_ctx, exe_ast_ctx, loclist_base_load_addr, NULL, *var_location.get(), &err))
{
if (log)
log->Printf("Error evaluating location: %s", err.AsCString());
return NULL;
}
clang::ASTContext *var_ast_context = type_list->GetClangASTContext().getASTContext();
void *type_to_use;
if (parser_ast_context)
{
type_to_use = ClangASTContext::CopyType(parser_ast_context, var_ast_context, var_opaque_type);
if (parser_type)
*parser_type = TypeFromParser(type_to_use, parser_ast_context);
}
else
type_to_use = var_opaque_type;
if (var_location.get()->GetContextType() == Value::eContextTypeInvalid)
var_location.get()->SetContext(Value::eContextTypeClangType, type_to_use);
if (var_location.get()->GetValueType() == Value::eValueTypeFileAddress)
{
SymbolContext var_sc;
var->CalculateSymbolContext(&var_sc);
if (!var_sc.module_sp)
return NULL;
ObjectFile *object_file = var_sc.module_sp->GetObjectFile();
if (!object_file)
return NULL;
Address so_addr(var_location->GetScalar().ULongLong(), object_file->GetSectionList());
lldb::addr_t load_addr = so_addr.GetLoadAddress(m_exe_ctx.target);
var_location->GetScalar() = load_addr;
var_location->SetValueType(Value::eValueTypeLoadAddress);
}
if (user_type)
*user_type = TypeFromUser(var_opaque_type, var_ast_context);
return var_location.release();
}
void
ClangExpressionDeclMap::AddOneVariable(NameSearchContext &context,
Removed the hacky "#define this ___clang_this" handler for C++ classes. Replaced it with a less hacky approach: - If an expression is defined in the context of a method of class A, then that expression is wrapped as ___clang_class::___clang_expr(void*) { ... } instead of ___clang_expr(void*) { ... }. - ___clang_class is resolved as the type of the target of the "this" pointer in the method the expression is defined in. - When reporting the type of ___clang_class, a method with the signature ___clang_expr(void*) is added to that class, so that Clang doesn't complain about a method being defined without a corresponding declaration. - Whenever the expression gets called, "this" gets looked up, type-checked, and then passed in as the first argument. This required the following changes: - The ABIs were changed to support passing of the "this" pointer as part of trivial calls. - ThreadPlanCallFunction and ClangFunction were changed to support passing of an optional "this" pointer. - ClangUserExpression was extended to perform the wrapping described above. - ClangASTSource was changed to revert the changes required by the hack. - ClangExpressionParser, IRForTarget, and ClangExpressionDeclMap were changed to handle different manglings of ___clang_expr flexibly. This meant no longer searching for a function called ___clang_expr, but rather looking for a function whose name *contains* ___clang_expr. - ClangExpressionParser and ClangExpressionDeclMap now remember whether "this" is required, and know how to look it up as necessary. A few inheritance bugs remain, and I'm trying to resolve these. But it is now possible to use "this" as well as refer implicitly to member variables, when in the proper context. llvm-svn: 114384
2010-09-21 08:44:12 +08:00
Variable* var)
{
lldb::LogSP log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS));
TypeFromUser ut;
TypeFromParser pt;
Value *var_location = GetVariableValue (m_exe_ctx,
var,
context.GetASTContext(),
&ut,
&pt);
Removed the hacky "#define this ___clang_this" handler for C++ classes. Replaced it with a less hacky approach: - If an expression is defined in the context of a method of class A, then that expression is wrapped as ___clang_class::___clang_expr(void*) { ... } instead of ___clang_expr(void*) { ... }. - ___clang_class is resolved as the type of the target of the "this" pointer in the method the expression is defined in. - When reporting the type of ___clang_class, a method with the signature ___clang_expr(void*) is added to that class, so that Clang doesn't complain about a method being defined without a corresponding declaration. - Whenever the expression gets called, "this" gets looked up, type-checked, and then passed in as the first argument. This required the following changes: - The ABIs were changed to support passing of the "this" pointer as part of trivial calls. - ThreadPlanCallFunction and ClangFunction were changed to support passing of an optional "this" pointer. - ClangUserExpression was extended to perform the wrapping described above. - ClangASTSource was changed to revert the changes required by the hack. - ClangExpressionParser, IRForTarget, and ClangExpressionDeclMap were changed to handle different manglings of ___clang_expr flexibly. This meant no longer searching for a function called ___clang_expr, but rather looking for a function whose name *contains* ___clang_expr. - ClangExpressionParser and ClangExpressionDeclMap now remember whether "this" is required, and know how to look it up as necessary. A few inheritance bugs remain, and I'm trying to resolve these. But it is now possible to use "this" as well as refer implicitly to member variables, when in the proper context. llvm-svn: 114384
2010-09-21 08:44:12 +08:00
NamedDecl *var_decl = context.AddVarDecl(pt.GetOpaqueQualType());
ClangExpressionVariable &entity(m_found_entities.VariableAtIndex(m_found_entities.CreateVariable()));
std::string decl_name(context.m_decl_name.getAsString());
entity.m_name.SetCString (decl_name.c_str());
entity.m_user_type = ut;
entity.EnableParserVars();
entity.m_parser_vars->m_parser_type = pt;
entity.m_parser_vars->m_named_decl = var_decl;
entity.m_parser_vars->m_llvm_value = NULL;
entity.m_parser_vars->m_lldb_value = var_location;
if (log)
{
std::string var_decl_print_string;
llvm::raw_string_ostream var_decl_print_stream(var_decl_print_string);
var_decl->print(var_decl_print_stream);
var_decl_print_stream.flush();
log->Printf("Found variable %s, returned %s", decl_name.c_str(), var_decl_print_string.c_str());
}
}
void
ClangExpressionDeclMap::AddOneVariable(NameSearchContext &context,
ClangExpressionVariable *pvar)
{
lldb::LogSP log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS));
TypeFromUser user_type = pvar->m_user_type;
TypeFromParser parser_type(ClangASTContext::CopyType(context.GetASTContext(),
user_type.GetASTContext(),
user_type.GetOpaqueQualType()),
context.GetASTContext());
NamedDecl *var_decl = context.AddVarDecl(parser_type.GetOpaqueQualType());
pvar->EnableParserVars();
pvar->m_parser_vars->m_parser_type = parser_type;
pvar->m_parser_vars->m_named_decl = var_decl;
pvar->m_parser_vars->m_llvm_value = NULL;
pvar->m_parser_vars->m_lldb_value = NULL;
if (log)
{
std::string var_decl_print_string;
llvm::raw_string_ostream var_decl_print_stream(var_decl_print_string);
var_decl->print(var_decl_print_stream);
var_decl_print_stream.flush();
log->Printf("Added pvar %s, returned %s", pvar->m_name.GetCString(), var_decl_print_string.c_str());
}
}
clang::NamespaceDecl *
ClangExpressionDeclMap::AddNamespace (NameSearchContext &context, const ClangNamespaceDecl &namespace_decl)
{
lldb::LogSP log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS));
clang::Decl *copied_decl = ClangASTContext::CopyDecl (context.GetASTContext(),
namespace_decl.GetASTContext(),
namespace_decl.GetNamespaceDecl());
return dyn_cast<clang::NamespaceDecl>(copied_decl);
}
void
ClangExpressionDeclMap::AddOneFunction(NameSearchContext &context,
Function* fun,
Symbol* symbol)
{
lldb::LogSP log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS));
NamedDecl *fun_decl;
std::auto_ptr<Value> fun_location(new Value);
const Address *fun_address;
// only valid for Functions, not for Symbols
void *fun_opaque_type = NULL;
clang::ASTContext *fun_ast_context = NULL;
if (fun)
{
Type *fun_type = fun->GetType();
if (!fun_type)
{
if (log)
log->PutCString("Skipped a function because it has no type");
return;
}
fun_opaque_type = fun_type->GetClangType();
if (!fun_opaque_type)
{
if (log)
log->PutCString("Skipped a function because it has no Clang type");
return;
}
fun_address = &fun->GetAddressRange().GetBaseAddress();
TypeList *type_list = fun_type->GetTypeList();
fun_ast_context = type_list->GetClangASTContext().getASTContext();
void *copied_type = ClangASTContext::CopyType(context.GetASTContext(), fun_ast_context, fun_opaque_type);
fun_decl = context.AddFunDecl(copied_type);
}
else if (symbol)
{
fun_address = &symbol->GetAddressRangeRef().GetBaseAddress();
fun_decl = context.AddGenericFunDecl();
}
else
{
if (log)
log->PutCString("AddOneFunction called with no function and no symbol");
return;
}
lldb::addr_t load_addr = fun_address->GetLoadAddress(m_exe_ctx.target);
fun_location->SetValueType(Value::eValueTypeLoadAddress);
fun_location->GetScalar() = load_addr;
ClangExpressionVariable &entity(m_found_entities.VariableAtIndex(m_found_entities.CreateVariable()));
std::string decl_name(context.m_decl_name.getAsString());
entity.m_name.SetCString(decl_name.c_str());
entity.m_user_type = TypeFromUser(fun_opaque_type, fun_ast_context);;
entity.EnableParserVars();
entity.m_parser_vars->m_named_decl = fun_decl;
entity.m_parser_vars->m_llvm_value = NULL;
entity.m_parser_vars->m_lldb_value = fun_location.release();
if (log)
{
std::string fun_decl_print_string;
llvm::raw_string_ostream fun_decl_print_stream(fun_decl_print_string);
fun_decl->print(fun_decl_print_stream);
fun_decl_print_stream.flush();
log->Printf("Found %s function %s, returned %s", (fun ? "specific" : "generic"), decl_name.c_str(), fun_decl_print_string.c_str());
}
}
void
ClangExpressionDeclMap::AddOneType(NameSearchContext &context,
Removed the hacky "#define this ___clang_this" handler for C++ classes. Replaced it with a less hacky approach: - If an expression is defined in the context of a method of class A, then that expression is wrapped as ___clang_class::___clang_expr(void*) { ... } instead of ___clang_expr(void*) { ... }. - ___clang_class is resolved as the type of the target of the "this" pointer in the method the expression is defined in. - When reporting the type of ___clang_class, a method with the signature ___clang_expr(void*) is added to that class, so that Clang doesn't complain about a method being defined without a corresponding declaration. - Whenever the expression gets called, "this" gets looked up, type-checked, and then passed in as the first argument. This required the following changes: - The ABIs were changed to support passing of the "this" pointer as part of trivial calls. - ThreadPlanCallFunction and ClangFunction were changed to support passing of an optional "this" pointer. - ClangUserExpression was extended to perform the wrapping described above. - ClangASTSource was changed to revert the changes required by the hack. - ClangExpressionParser, IRForTarget, and ClangExpressionDeclMap were changed to handle different manglings of ___clang_expr flexibly. This meant no longer searching for a function called ___clang_expr, but rather looking for a function whose name *contains* ___clang_expr. - ClangExpressionParser and ClangExpressionDeclMap now remember whether "this" is required, and know how to look it up as necessary. A few inheritance bugs remain, and I'm trying to resolve these. But it is now possible to use "this" as well as refer implicitly to member variables, when in the proper context. llvm-svn: 114384
2010-09-21 08:44:12 +08:00
TypeFromUser &ut,
bool add_method)
{
Removed the hacky "#define this ___clang_this" handler for C++ classes. Replaced it with a less hacky approach: - If an expression is defined in the context of a method of class A, then that expression is wrapped as ___clang_class::___clang_expr(void*) { ... } instead of ___clang_expr(void*) { ... }. - ___clang_class is resolved as the type of the target of the "this" pointer in the method the expression is defined in. - When reporting the type of ___clang_class, a method with the signature ___clang_expr(void*) is added to that class, so that Clang doesn't complain about a method being defined without a corresponding declaration. - Whenever the expression gets called, "this" gets looked up, type-checked, and then passed in as the first argument. This required the following changes: - The ABIs were changed to support passing of the "this" pointer as part of trivial calls. - ThreadPlanCallFunction and ClangFunction were changed to support passing of an optional "this" pointer. - ClangUserExpression was extended to perform the wrapping described above. - ClangASTSource was changed to revert the changes required by the hack. - ClangExpressionParser, IRForTarget, and ClangExpressionDeclMap were changed to handle different manglings of ___clang_expr flexibly. This meant no longer searching for a function called ___clang_expr, but rather looking for a function whose name *contains* ___clang_expr. - ClangExpressionParser and ClangExpressionDeclMap now remember whether "this" is required, and know how to look it up as necessary. A few inheritance bugs remain, and I'm trying to resolve these. But it is now possible to use "this" as well as refer implicitly to member variables, when in the proper context. llvm-svn: 114384
2010-09-21 08:44:12 +08:00
clang::ASTContext *parser_ast_context = context.GetASTContext();
clang::ASTContext *user_ast_context = ut.GetASTContext();
void *copied_type = ClangASTContext::CopyType(parser_ast_context, user_ast_context, ut.GetOpaqueQualType());
TypeFromParser parser_type(copied_type, parser_ast_context);
Removed the hacky "#define this ___clang_this" handler for C++ classes. Replaced it with a less hacky approach: - If an expression is defined in the context of a method of class A, then that expression is wrapped as ___clang_class::___clang_expr(void*) { ... } instead of ___clang_expr(void*) { ... }. - ___clang_class is resolved as the type of the target of the "this" pointer in the method the expression is defined in. - When reporting the type of ___clang_class, a method with the signature ___clang_expr(void*) is added to that class, so that Clang doesn't complain about a method being defined without a corresponding declaration. - Whenever the expression gets called, "this" gets looked up, type-checked, and then passed in as the first argument. This required the following changes: - The ABIs were changed to support passing of the "this" pointer as part of trivial calls. - ThreadPlanCallFunction and ClangFunction were changed to support passing of an optional "this" pointer. - ClangUserExpression was extended to perform the wrapping described above. - ClangASTSource was changed to revert the changes required by the hack. - ClangExpressionParser, IRForTarget, and ClangExpressionDeclMap were changed to handle different manglings of ___clang_expr flexibly. This meant no longer searching for a function called ___clang_expr, but rather looking for a function whose name *contains* ___clang_expr. - ClangExpressionParser and ClangExpressionDeclMap now remember whether "this" is required, and know how to look it up as necessary. A few inheritance bugs remain, and I'm trying to resolve these. But it is now possible to use "this" as well as refer implicitly to member variables, when in the proper context. llvm-svn: 114384
2010-09-21 08:44:12 +08:00
if (add_method && ClangASTContext::IsAggregateType(copied_type))
{
void *args[1];
args[0] = ClangASTContext::GetVoidPtrType(parser_ast_context, false);
void *method_type = ClangASTContext::CreateFunctionType (parser_ast_context,
ClangASTContext::GetBuiltInType_void(parser_ast_context),
args,
1,
false,
ClangASTContext::GetTypeQualifiers(copied_type));
const bool is_virtual = false;
const bool is_static = false;
const bool is_inline = false;
const bool is_explicit = false;
ClangASTContext::AddMethodToCXXRecordType (parser_ast_context,
copied_type,
"$__lldb_expr",
method_type,
lldb::eAccessPublic,
is_virtual,
is_static,
is_inline,
is_explicit);
Removed the hacky "#define this ___clang_this" handler for C++ classes. Replaced it with a less hacky approach: - If an expression is defined in the context of a method of class A, then that expression is wrapped as ___clang_class::___clang_expr(void*) { ... } instead of ___clang_expr(void*) { ... }. - ___clang_class is resolved as the type of the target of the "this" pointer in the method the expression is defined in. - When reporting the type of ___clang_class, a method with the signature ___clang_expr(void*) is added to that class, so that Clang doesn't complain about a method being defined without a corresponding declaration. - Whenever the expression gets called, "this" gets looked up, type-checked, and then passed in as the first argument. This required the following changes: - The ABIs were changed to support passing of the "this" pointer as part of trivial calls. - ThreadPlanCallFunction and ClangFunction were changed to support passing of an optional "this" pointer. - ClangUserExpression was extended to perform the wrapping described above. - ClangASTSource was changed to revert the changes required by the hack. - ClangExpressionParser, IRForTarget, and ClangExpressionDeclMap were changed to handle different manglings of ___clang_expr flexibly. This meant no longer searching for a function called ___clang_expr, but rather looking for a function whose name *contains* ___clang_expr. - ClangExpressionParser and ClangExpressionDeclMap now remember whether "this" is required, and know how to look it up as necessary. A few inheritance bugs remain, and I'm trying to resolve these. But it is now possible to use "this" as well as refer implicitly to member variables, when in the proper context. llvm-svn: 114384
2010-09-21 08:44:12 +08:00
}
context.AddTypeDecl(copied_type);
}