From ba9e5d6e3f8b744196c4dd42e035b4a03be35a03 Mon Sep 17 00:00:00 2001 From: davidben Date: Thu, 21 Jan 2016 17:41:41 -0800 Subject: [PATCH] Allow std::unordered_*. This is a reland of https://codereview.chromium.org/1502373009 with some fixes for components/metrics/leak_detector allocator type mismatches. Original issue's description: > Allow std::unordered_*. > > base::hash_* is, as a transition step, implemented in terms of > std::unordered_*. Later commits will convert existing uses. > > Also fix a host of IWYU problems that arose from this CL. > > (NOPRESUBMIT because the wstring presubmit check is overzealous > and complains about the reference to wstring in the comment.) > > Committed: https://crrev.com/3f37f7f1459e7b5a452c0e433493e0a6e9649ca7 > Cr-Commit-Position: refs/heads/master@{#370553} BUG=576864 TBR=derat@chromium.org,danakj@chromium.org,dalecurtis@chromium.org,jbauman@chromium.org,blundell@chromium.org NOPRESUBMIT=true CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel Review URL: https://codereview.chromium.org/1615713003 Cr-Commit-Position: refs/heads/master@{#370867} --- .../chromium/base/containers/hash_tables.h | 280 +++--------------- security/sandbox/chromium/base/hash.h | 82 +++++ security/sandbox/chromium/base/location.h | 4 +- .../sandbox/chromium/base/strings/string16.h | 17 ++ .../chromium/base/strings/string_piece.h | 6 +- 5 files changed, 141 insertions(+), 248 deletions(-) diff --git a/security/sandbox/chromium/base/containers/hash_tables.h b/security/sandbox/chromium/base/containers/hash_tables.h index c421dddf3e7c7..8da7b6726852d 100644 --- a/security/sandbox/chromium/base/containers/hash_tables.h +++ b/security/sandbox/chromium/base/containers/hash_tables.h @@ -1,281 +1,75 @@ // Copyright (c) 2011 The Chromium Authors. All rights reserved. // Use of this source code is governed by a BSD-style license that can be // found in the LICENSE file. -// - -// -// Deal with the differences between Microsoft and GNU implemenations -// of hash_map. Allows all platforms to use |base::hash_map| and -// |base::hash_set|. -// eg: -// base::hash_map my_map; -// base::hash_set my_set; -// -// NOTE: It is an explicit non-goal of this class to provide a generic hash -// function for pointers. If you want to hash a pointers to a particular class, -// please define the template specialization elsewhere (for example, in its -// header file) and keep it specific to just pointers to that class. This is -// because identity hashes are not desirable for all types that might show up -// in containers as pointers. #ifndef BASE_CONTAINERS_HASH_TABLES_H_ #define BASE_CONTAINERS_HASH_TABLES_H_ -#include -#include - -#include - -#include "base/strings/string16.h" -#include "build/build_config.h" - -#if defined(COMPILER_MSVC) +#include #include #include +#include -#define BASE_HASH_NAMESPACE std +#include "base/hash.h" -#elif defined(COMPILER_GCC) +// This header file is deprecated. Use the corresponding C++11 type +// instead. https://crbug.com/576864 +// Use a custom hasher instead. #define BASE_HASH_NAMESPACE base_hash -// This is a hack to disable the gcc 4.4 warning about hash_map and hash_set -// being deprecated. We can get rid of this when we upgrade to VS2008 and we -// can use and . -#ifdef __DEPRECATED -#define CHROME_OLD__DEPRECATED __DEPRECATED -#undef __DEPRECATED -#endif - -#include -#include -#define BASE_HASH_IMPL_NAMESPACE __gnu_cxx - -#include - -#ifdef CHROME_OLD__DEPRECATED -#define __DEPRECATED CHROME_OLD__DEPRECATED -#undef CHROME_OLD__DEPRECATED -#endif - namespace BASE_HASH_NAMESPACE { -// The pre-standard hash behaves like C++11's std::hash, except around pointers. -// const char* is specialized to hash the C string and hash functions for -// general T* are missing. Define a BASE_HASH_NAMESPACE::hash which aligns with -// the C++11 behavior. - +// A separate hasher which, by default, forwards to std::hash. This is so legacy +// uses of BASE_HASH_NAMESPACE with base::hash_map do not interfere with +// std::hash mid-transition. template struct hash { - std::size_t operator()(const T& value) const { - return BASE_HASH_IMPL_NAMESPACE::hash()(value); - } + std::size_t operator()(const T& value) const { return std::hash()(value); } }; -template -struct hash { - std::size_t operator()(T* value) const { - return BASE_HASH_IMPL_NAMESPACE::hash()( - reinterpret_cast(value)); +// Use base::IntPairHash from base/hash.h as a custom hasher instead. +template +struct hash> { + std::size_t operator()(std::pair value) const { + return base::HashInts(value.first, value.second); } }; -// The GNU C++ library provides identity hash functions for many integral types, -// but not for |long long|. This hash function will truncate if |size_t| is -// narrower than |long long|. This is probably good enough for what we will -// use it for. - -#define DEFINE_TRIVIAL_HASH(integral_type) \ - template<> \ - struct hash { \ - std::size_t operator()(integral_type value) const { \ - return static_cast(value); \ - } \ - } - -DEFINE_TRIVIAL_HASH(long long); -DEFINE_TRIVIAL_HASH(unsigned long long); - -#undef DEFINE_TRIVIAL_HASH - -// Implement string hash functions so that strings of various flavors can -// be used as keys in STL maps and sets. The hash algorithm comes from the -// GNU C++ library, in . It is duplicated here because GCC -// versions prior to 4.3.2 are unable to compile when RTTI -// is disabled, as it is in our build. - -#define DEFINE_STRING_HASH(string_type) \ - template<> \ - struct hash { \ - std::size_t operator()(const string_type& s) const { \ - std::size_t result = 0; \ - for (string_type::const_iterator i = s.begin(); i != s.end(); ++i) \ - result = (result * 131) + *i; \ - return result; \ - } \ - } - -DEFINE_STRING_HASH(std::string); -DEFINE_STRING_HASH(base::string16); - -#undef DEFINE_STRING_HASH - } // namespace BASE_HASH_NAMESPACE -#else // COMPILER -#error define BASE_HASH_NAMESPACE for your compiler -#endif // COMPILER - namespace base { -// On MSVC, use the C++11 containers. -#if defined(COMPILER_MSVC) - -template, - class Pred = std::equal_to, - class Alloc = std::allocator>> +// Use std::unordered_map instead. +template , + class Pred = std::equal_to, + class Alloc = std::allocator>> using hash_map = std::unordered_map; -template, - class Pred = std::equal_to, - class Alloc = std::allocator>> +// Use std::unordered_multimap instead. +template , + class Pred = std::equal_to, + class Alloc = std::allocator>> using hash_multimap = std::unordered_multimap; -template, - class Pred = std::equal_to, - class Alloc = std::allocator> +// Use std::unordered_multiset instead. +template , + class Pred = std::equal_to, + class Alloc = std::allocator> using hash_multiset = std::unordered_multiset; -template, - class Pred = std::equal_to, - class Alloc = std::allocator> +// Use std::unordered_set instead. +template , + class Pred = std::equal_to, + class Alloc = std::allocator> using hash_set = std::unordered_set; -#else // !COMPILER_MSVC - -// Otherwise, use the pre-standard ones, but override the default hash to match -// C++11. -template, - class Pred = std::equal_to, - class Alloc = std::allocator>> -using hash_map = BASE_HASH_IMPL_NAMESPACE::hash_map; - -template, - class Pred = std::equal_to, - class Alloc = std::allocator>> -using hash_multimap = - BASE_HASH_IMPL_NAMESPACE::hash_multimap; - -template, - class Pred = std::equal_to, - class Alloc = std::allocator> -using hash_multiset = - BASE_HASH_IMPL_NAMESPACE::hash_multiset; - -template, - class Pred = std::equal_to, - class Alloc = std::allocator> -using hash_set = BASE_HASH_IMPL_NAMESPACE::hash_set; - -#undef BASE_HASH_IMPL_NAMESPACE - -#endif // COMPILER_MSVC - -// Implement hashing for pairs of at-most 32 bit integer values. -// When size_t is 32 bits, we turn the 64-bit hash code into 32 bits by using -// multiply-add hashing. This algorithm, as described in -// Theorem 4.3.3 of the thesis "Über die Komplexität der Multiplikation in -// eingeschränkten Branchingprogrammmodellen" by Woelfel, is: -// -// h32(x32, y32) = (h64(x32, y32) * rand_odd64 + rand16 * 2^16) % 2^64 / 2^32 -// -// Contact danakj@chromium.org for any questions. -inline std::size_t HashInts32(uint32_t value1, uint32_t value2) { - uint64_t value1_64 = value1; - uint64_t hash64 = (value1_64 << 32) | value2; - - if (sizeof(std::size_t) >= sizeof(uint64_t)) - return static_cast(hash64); - - uint64_t odd_random = 481046412LL << 32 | 1025306955LL; - uint32_t shift_random = 10121U << 16; - - hash64 = hash64 * odd_random + shift_random; - std::size_t high_bits = static_cast( - hash64 >> (8 * (sizeof(uint64_t) - sizeof(std::size_t)))); - return high_bits; -} - -// Implement hashing for pairs of up-to 64-bit integer values. -// We use the compound integer hash method to produce a 64-bit hash code, by -// breaking the two 64-bit inputs into 4 32-bit values: -// http://opendatastructures.org/versions/edition-0.1d/ods-java/node33.html#SECTION00832000000000000000 -// Then we reduce our result to 32 bits if required, similar to above. -inline std::size_t HashInts64(uint64_t value1, uint64_t value2) { - uint32_t short_random1 = 842304669U; - uint32_t short_random2 = 619063811U; - uint32_t short_random3 = 937041849U; - uint32_t short_random4 = 3309708029U; - - uint32_t value1a = static_cast(value1 & 0xffffffff); - uint32_t value1b = static_cast((value1 >> 32) & 0xffffffff); - uint32_t value2a = static_cast(value2 & 0xffffffff); - uint32_t value2b = static_cast((value2 >> 32) & 0xffffffff); - - uint64_t product1 = static_cast(value1a) * short_random1; - uint64_t product2 = static_cast(value1b) * short_random2; - uint64_t product3 = static_cast(value2a) * short_random3; - uint64_t product4 = static_cast(value2b) * short_random4; - - uint64_t hash64 = product1 + product2 + product3 + product4; - - if (sizeof(std::size_t) >= sizeof(uint64_t)) - return static_cast(hash64); - - uint64_t odd_random = 1578233944LL << 32 | 194370989LL; - uint32_t shift_random = 20591U << 16; - - hash64 = hash64 * odd_random + shift_random; - std::size_t high_bits = static_cast( - hash64 >> (8 * (sizeof(uint64_t) - sizeof(std::size_t)))); - return high_bits; -} - -template -inline std::size_t HashPair(T1 value1, T2 value2) { - // This condition is expected to be compile-time evaluated and optimised away - // in release builds. - if (sizeof(T1) > sizeof(uint32_t) || (sizeof(T2) > sizeof(uint32_t))) - return HashInts64(value1, value2); - - return HashInts32(value1, value2); -} - } // namespace base -namespace BASE_HASH_NAMESPACE { - -// Implement methods for hashing a pair of integers, so they can be used as -// keys in STL containers. - -template -struct hash > { - std::size_t operator()(std::pair value) const { - return base::HashPair(value.first, value.second); - } -}; - -} // namespace BASE_HASH_NAMESPACE - -#undef DEFINE_PAIR_HASH_FUNCTION_START -#undef DEFINE_PAIR_HASH_FUNCTION_END - #endif // BASE_CONTAINERS_HASH_TABLES_H_ diff --git a/security/sandbox/chromium/base/hash.h b/security/sandbox/chromium/base/hash.h index ed8d9fd4cc9b9..97e251cdf1ed8 100644 --- a/security/sandbox/chromium/base/hash.h +++ b/security/sandbox/chromium/base/hash.h @@ -10,6 +10,7 @@ #include #include +#include #include "base/base_export.h" #include "base/logging.h" @@ -35,6 +36,87 @@ inline uint32_t Hash(const std::string& str) { return Hash(str.data(), str.size()); } +// Implement hashing for pairs of at-most 32 bit integer values. +// When size_t is 32 bits, we turn the 64-bit hash code into 32 bits by using +// multiply-add hashing. This algorithm, as described in +// Theorem 4.3.3 of the thesis "Über die Komplexität der Multiplikation in +// eingeschränkten Branchingprogrammmodellen" by Woelfel, is: +// +// h32(x32, y32) = (h64(x32, y32) * rand_odd64 + rand16 * 2^16) % 2^64 / 2^32 +// +// Contact danakj@chromium.org for any questions. +inline size_t HashInts32(uint32_t value1, uint32_t value2) { + uint64_t value1_64 = value1; + uint64_t hash64 = (value1_64 << 32) | value2; + + if (sizeof(size_t) >= sizeof(uint64_t)) + return static_cast(hash64); + + uint64_t odd_random = 481046412LL << 32 | 1025306955LL; + uint32_t shift_random = 10121U << 16; + + hash64 = hash64 * odd_random + shift_random; + size_t high_bits = + static_cast(hash64 >> (8 * (sizeof(uint64_t) - sizeof(size_t)))); + return high_bits; +} + +// Implement hashing for pairs of up-to 64-bit integer values. +// We use the compound integer hash method to produce a 64-bit hash code, by +// breaking the two 64-bit inputs into 4 32-bit values: +// http://opendatastructures.org/versions/edition-0.1d/ods-java/node33.html#SECTION00832000000000000000 +// Then we reduce our result to 32 bits if required, similar to above. +inline size_t HashInts64(uint64_t value1, uint64_t value2) { + uint32_t short_random1 = 842304669U; + uint32_t short_random2 = 619063811U; + uint32_t short_random3 = 937041849U; + uint32_t short_random4 = 3309708029U; + + uint32_t value1a = static_cast(value1 & 0xffffffff); + uint32_t value1b = static_cast((value1 >> 32) & 0xffffffff); + uint32_t value2a = static_cast(value2 & 0xffffffff); + uint32_t value2b = static_cast((value2 >> 32) & 0xffffffff); + + uint64_t product1 = static_cast(value1a) * short_random1; + uint64_t product2 = static_cast(value1b) * short_random2; + uint64_t product3 = static_cast(value2a) * short_random3; + uint64_t product4 = static_cast(value2b) * short_random4; + + uint64_t hash64 = product1 + product2 + product3 + product4; + + if (sizeof(size_t) >= sizeof(uint64_t)) + return static_cast(hash64); + + uint64_t odd_random = 1578233944LL << 32 | 194370989LL; + uint32_t shift_random = 20591U << 16; + + hash64 = hash64 * odd_random + shift_random; + size_t high_bits = + static_cast(hash64 >> (8 * (sizeof(uint64_t) - sizeof(size_t)))); + return high_bits; +} + +template +inline size_t HashInts(T1 value1, T2 value2) { + // This condition is expected to be compile-time evaluated and optimised away + // in release builds. + if (sizeof(T1) > sizeof(uint32_t) || (sizeof(T2) > sizeof(uint32_t))) + return HashInts64(value1, value2); + + return HashInts32(value1, value2); +} + +// A templated hasher for pairs of integer types. +template +struct IntPairHash; + +template +struct IntPairHash> { + size_t operator()(std::pair value) const { + return HashInts(value.first, value.second); + } +}; + } // namespace base #endif // BASE_HASH_H_ diff --git a/security/sandbox/chromium/base/location.h b/security/sandbox/chromium/base/location.h index d3bb23c63ea45..21e270c5a9fc5 100644 --- a/security/sandbox/chromium/base/location.h +++ b/security/sandbox/chromium/base/location.h @@ -11,7 +11,7 @@ #include #include "base/base_export.h" -#include "base/containers/hash_tables.h" +#include "base/hash.h" namespace tracked_objects { @@ -59,7 +59,7 @@ class BASE_EXPORT Location { // it comes from __FILE__, so no need to check the contents of the string. // See the definition of FROM_HERE in location.h, and how it is used // elsewhere. - return base::HashPair(reinterpret_cast(location.file_name()), + return base::HashInts(reinterpret_cast(location.file_name()), location.line_number()); } }; diff --git a/security/sandbox/chromium/base/strings/string16.h b/security/sandbox/chromium/base/strings/string16.h index e47669c1b565e..82dd0fab4f326 100644 --- a/security/sandbox/chromium/base/strings/string16.h +++ b/security/sandbox/chromium/base/strings/string16.h @@ -29,6 +29,8 @@ #include #include #include + +#include #include #include "base/base_export.h" @@ -182,6 +184,21 @@ BASE_EXPORT extern void PrintTo(const string16& str, std::ostream* out); extern template class BASE_EXPORT std::basic_string; +// Specialize std::hash for base::string16. Although the style guide forbids +// this in general, it is necessary for consistency with WCHAR_T_IS_UTF16 +// platforms, where base::string16 is a type alias for std::wstring. +namespace std { +template <> +struct hash { + std::size_t operator()(const base::string16& s) const { + std::size_t result = 0; + for (base::char16 c : s) + result = (result * 131) + c; + return result; + } +}; +} // namespace std + #endif // WCHAR_T_IS_UTF32 #endif // BASE_STRINGS_STRING16_H_ diff --git a/security/sandbox/chromium/base/strings/string_piece.h b/security/sandbox/chromium/base/strings/string_piece.h index 31e7596d119c2..92634b9362b7d 100644 --- a/security/sandbox/chromium/base/strings/string_piece.h +++ b/security/sandbox/chromium/base/strings/string_piece.h @@ -439,9 +439,9 @@ BASE_EXPORT std::ostream& operator<<(std::ostream& o, // We provide appropriate hash functions so StringPiece and StringPiece16 can // be used as keys in hash sets and maps. -// This hash function is copied from base/containers/hash_tables.h. We don't -// use the ones already defined for string and string16 directly because it -// would require the string constructors to be called, which we don't want. +// This hash function is copied from base/strings/string16.h. We don't use the +// ones already defined for string and string16 directly because it would +// require the string constructors to be called, which we don't want. #define HASH_STRING_PIECE(StringPieceType, string_piece) \ std::size_t result = 0; \ for (StringPieceType::const_iterator i = string_piece.begin(); \ -- GitLab