diff options
Diffstat (limited to 'lldb/source/Plugins/SymbolFile/DWARF/HashedNameToDIE.cpp')
-rw-r--r-- | lldb/source/Plugins/SymbolFile/DWARF/HashedNameToDIE.cpp | 136 |
1 files changed, 136 insertions, 0 deletions
diff --git a/lldb/source/Plugins/SymbolFile/DWARF/HashedNameToDIE.cpp b/lldb/source/Plugins/SymbolFile/DWARF/HashedNameToDIE.cpp new file mode 100644 index 00000000000..6eb249bfc34 --- /dev/null +++ b/lldb/source/Plugins/SymbolFile/DWARF/HashedNameToDIE.cpp @@ -0,0 +1,136 @@ +//===-- HashedNameToDIE.cpp -------------------------------------------*- C++ -*-===// +// +// The LLVM Compiler Infrastructure +// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// + +#include "HashedNameToDIE.h" +#include "lldb/Core/ConstString.h" +#include "lldb/Core/DataExtractor.h" +#include "lldb/Core/Stream.h" +#include "lldb/Core/StreamString.h" +#include "lldb/Core/RegularExpression.h" +#include "lldb/Symbol/ObjectFile.h" + +#include "DWARFCompileUnit.h" +#include "DWARFDebugInfo.h" +#include "DWARFDebugInfoEntry.h" +#include "SymbolFileDWARF.h" +using namespace lldb; +using namespace lldb_private; + +static uint32_t +dl_new_hash (const char *s) +{ + uint32_t h = 5381; + + for (unsigned char c = *s; c; c = *++s) + h = ((h << 5) + h) + c; + + return h; +} + +HashedNameToDIE::HashedNameToDIE (SymbolFileDWARF *dwarf, const DataExtractor &data) : + m_dwarf (dwarf), + m_data (data), + m_header () +{ + uint32_t offset = 0; + m_header.version = m_data.GetU16(&offset); + m_header.hash_type = m_data.GetU8(&offset); + m_header.hash_index_bitsize = m_data.GetU8(&offset); + m_header.num_buckets = m_data.GetU32(&offset); + m_header.num_hashes = m_data.GetU32(&offset); + m_header.die_offset_base = m_data.GetU32(&offset); +} + + +size_t +HashedNameToDIE::Find (const ConstString &name, DIEArray &die_ofsets) const +{ + const size_t initial_size = die_ofsets.size(); + const char *name_cstr = name.GetCString(); + if (name_cstr && name_cstr[0]) + { + // Hash the C string + const uint32_t name_hash = dl_new_hash (name_cstr); + + // Find the correct bucket for the using the hash value + const uint32_t bucket_idx = name_hash % m_header.num_buckets; + + // Calculate the offset for the bucket entry for the bucket index + uint32_t offset = GetOffsetForBucket (bucket_idx); + + // Extract the bucket entry. + const uint32_t bucket_entry = m_data.GetU32 (&offset); + if (bucket_entry) + { + // The bucket entry is non-zero which means it isn't empty. + // The bucket entry is made up of a hash index whose bit width + // is m_header.hash_index_bitsize, and a hash count whose value + // is the remaining bits in the 32 bit value. Below we extract + // the hash index and the hash count + const uint32_t hash_idx = bucket_entry & GetHashIndexMask(); + const uint32_t hash_count = bucket_entry >> m_header.hash_index_bitsize; + const uint32_t hash_end_idx = hash_idx + hash_count; + // Figure out the offset to the hash value by index + uint32_t hash_offset = GetOffsetForHash (hash_idx); + for (uint32_t idx = hash_idx; idx < hash_end_idx; ++idx) + { + // Extract the hash value and see if it matches our string + const uint32_t hash = m_data.GetU32 (&hash_offset); + if (hash == name_hash) + { + // The hash matches, but we still need to verify that the + // C string matches in case we have a hash collision. Figure + // out the offset for the data associated with this hash entry + offset = GetOffsetForOffset (hash_idx); + // Extract the first 32 bit value which is the .debug_str offset + // of the string we need + const uint32_t str_offset = m_data.GetU32 (&offset); + // Extract the C string and comapare it + const char *cstr_name = m_dwarf->get_debug_str_data().PeekCStr(str_offset); + if (strcmp(name_cstr, cstr_name) == 0) + { + // We have a match, now extract the DIE count + const uint32_t die_count = m_data.GetU32 (&offset); + // Now extract "die_count" DIE offsets and put them into the + // results + for (uint32_t die_idx = 0; die_idx < die_count; ++die_idx) + die_ofsets.push_back(m_data.GetU32 (&offset)); + } + } + } + } + } + return die_ofsets.size() - initial_size; +} + +size_t +HashedNameToDIE::Find (const RegularExpression& regex, DIEArray &die_ofsets) const +{ +// const size_t initial_info_array_size = info_array.size(); +// collection::const_iterator pos, end = m_collection.end(); +// for (pos = m_collection.begin(); pos != end; ++pos) +// { +// if (regex.Execute(pos->first)) +// info_array.push_back (pos->second); +// } +// return info_array.size() - initial_info_array_size; + return 0; +} + +void +HashedNameToDIE::Dump (Stream *s) +{ +// collection::const_iterator pos, end = m_collection.end(); +// for (pos = m_collection.begin(); pos != end; ++pos) +// { +// s->Printf("%p: 0x%8.8x 0x%8.8x \"%s\"\n", pos->first, pos->second.cu_idx, pos->second.die_idx, pos->first); +// } +} + + |