Abseil Common Libraries (C++) (grcp 依赖) https://abseil.io/
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

722 lines
28 KiB

Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
// Copyright 2017 The Abseil Authors.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// https://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
// This header file defines macros for declaring attributes for functions,
// types, and variables.
//
// These macros are used within Abseil and allow the compiler to optimize, where
// applicable, certain function calls.
//
// Most macros here are exposing GCC or Clang features, and are stubbed out for
// other compilers.
//
// GCC attributes documentation:
// https://gcc.gnu.org/onlinedocs/gcc-4.7.0/gcc/Function-Attributes.html
// https://gcc.gnu.org/onlinedocs/gcc-4.7.0/gcc/Variable-Attributes.html
// https://gcc.gnu.org/onlinedocs/gcc-4.7.0/gcc/Type-Attributes.html
//
// Most attributes in this file are already supported by GCC 4.7. However, some
// of them are not supported in older version of Clang. Thus, we check
// `__has_attribute()` first. If the check fails, we check if we are on GCC and
// assume the attribute exists on GCC (which is verified on GCC 4.7).
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
#ifndef ABSL_BASE_ATTRIBUTES_H_
#define ABSL_BASE_ATTRIBUTES_H_
#include "absl/base/config.h"
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
// ABSL_HAVE_ATTRIBUTE
//
// A function-like feature checking macro that is a wrapper around
// `__has_attribute`, which is defined by GCC 5+ and Clang and evaluates to a
// nonzero constant integer if the attribute is supported or 0 if not.
//
// It evaluates to zero if `__has_attribute` is not defined by the compiler.
//
// GCC: https://gcc.gnu.org/gcc-5/changes.html
// Clang: https://clang.llvm.org/docs/LanguageExtensions.html
#ifdef __has_attribute
#define ABSL_HAVE_ATTRIBUTE(x) __has_attribute(x)
#else
#define ABSL_HAVE_ATTRIBUTE(x) 0
#endif
// ABSL_HAVE_CPP_ATTRIBUTE
//
// A function-like feature checking macro that accepts C++11 style attributes.
// It's a wrapper around `__has_cpp_attribute`, defined by ISO C++ SD-6
// (https://en.cppreference.com/w/cpp/experimental/feature_test). If we don't
// find `__has_cpp_attribute`, will evaluate to 0.
#if defined(__cplusplus) && defined(__has_cpp_attribute)
// NOTE: requiring __cplusplus above should not be necessary, but
// works around https://bugs.llvm.org/show_bug.cgi?id=23435.
#define ABSL_HAVE_CPP_ATTRIBUTE(x) __has_cpp_attribute(x)
#else
#define ABSL_HAVE_CPP_ATTRIBUTE(x) 0
#endif
// -----------------------------------------------------------------------------
// Function Attributes
// -----------------------------------------------------------------------------
//
// GCC: https://gcc.gnu.org/onlinedocs/gcc/Function-Attributes.html
// Clang: https://clang.llvm.org/docs/AttributeReference.html
// ABSL_PRINTF_ATTRIBUTE
// ABSL_SCANF_ATTRIBUTE
//
// Tells the compiler to perform `printf` format string checking if the
// compiler supports it; see the 'format' attribute in
// <https://gcc.gnu.org/onlinedocs/gcc-4.7.0/gcc/Function-Attributes.html>.
//
// Note: As the GCC manual states, "[s]ince non-static C++ methods
// have an implicit 'this' argument, the arguments of such methods
// should be counted from two, not one."
#if ABSL_HAVE_ATTRIBUTE(format) || (defined(__GNUC__) && !defined(__clang__))
#define ABSL_PRINTF_ATTRIBUTE(string_index, first_to_check) \
__attribute__((__format__(__printf__, string_index, first_to_check)))
#define ABSL_SCANF_ATTRIBUTE(string_index, first_to_check) \
__attribute__((__format__(__scanf__, string_index, first_to_check)))
#else
#define ABSL_PRINTF_ATTRIBUTE(string_index, first_to_check)
#define ABSL_SCANF_ATTRIBUTE(string_index, first_to_check)
#endif
// ABSL_ATTRIBUTE_ALWAYS_INLINE
// ABSL_ATTRIBUTE_NOINLINE
//
// Forces functions to either inline or not inline. Introduced in gcc 3.1.
#if ABSL_HAVE_ATTRIBUTE(always_inline) || \
(defined(__GNUC__) && !defined(__clang__))
#define ABSL_ATTRIBUTE_ALWAYS_INLINE __attribute__((always_inline))
#define ABSL_HAVE_ATTRIBUTE_ALWAYS_INLINE 1
#else
#define ABSL_ATTRIBUTE_ALWAYS_INLINE
#endif
#if ABSL_HAVE_ATTRIBUTE(noinline) || (defined(__GNUC__) && !defined(__clang__))
#define ABSL_ATTRIBUTE_NOINLINE __attribute__((noinline))
#define ABSL_HAVE_ATTRIBUTE_NOINLINE 1
#else
#define ABSL_ATTRIBUTE_NOINLINE
#endif
// ABSL_ATTRIBUTE_NO_TAIL_CALL
//
// Prevents the compiler from optimizing away stack frames for functions which
// end in a call to another function.
#if ABSL_HAVE_ATTRIBUTE(disable_tail_calls)
#define ABSL_HAVE_ATTRIBUTE_NO_TAIL_CALL 1
#define ABSL_ATTRIBUTE_NO_TAIL_CALL __attribute__((disable_tail_calls))
#elif defined(__GNUC__) && !defined(__clang__) && !defined(__e2k__)
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
#define ABSL_HAVE_ATTRIBUTE_NO_TAIL_CALL 1
#define ABSL_ATTRIBUTE_NO_TAIL_CALL \
__attribute__((optimize("no-optimize-sibling-calls")))
#else
#define ABSL_ATTRIBUTE_NO_TAIL_CALL
#define ABSL_HAVE_ATTRIBUTE_NO_TAIL_CALL 0
#endif
// ABSL_ATTRIBUTE_WEAK
//
// Tags a function as weak for the purposes of compilation and linking.
// Weak attributes did not work properly in LLVM's Windows backend before
// 9.0.0, so disable them there. See https://bugs.llvm.org/show_bug.cgi?id=37598
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
// for further information.
// The MinGW compiler doesn't complain about the weak attribute until the link
// step, presumably because Windows doesn't use ELF binaries.
#if (ABSL_HAVE_ATTRIBUTE(weak) || \
(defined(__GNUC__) && !defined(__clang__))) && \
(!defined(_WIN32) || __clang_major__ < 9) && !defined(__MINGW32__)
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
#undef ABSL_ATTRIBUTE_WEAK
#define ABSL_ATTRIBUTE_WEAK __attribute__((weak))
#define ABSL_HAVE_ATTRIBUTE_WEAK 1
#else
#define ABSL_ATTRIBUTE_WEAK
#define ABSL_HAVE_ATTRIBUTE_WEAK 0
#endif
// ABSL_ATTRIBUTE_NONNULL
//
// Tells the compiler either (a) that a particular function parameter
// should be a non-null pointer, or (b) that all pointer arguments should
// be non-null.
//
// Note: As the GCC manual states, "[s]ince non-static C++ methods
// have an implicit 'this' argument, the arguments of such methods
// should be counted from two, not one."
//
// Args are indexed starting at 1.
//
// For non-static class member functions, the implicit `this` argument
// is arg 1, and the first explicit argument is arg 2. For static class member
// functions, there is no implicit `this`, and the first explicit argument is
// arg 1.
//
// Example:
//
// /* arg_a cannot be null, but arg_b can */
// void Function(void* arg_a, void* arg_b) ABSL_ATTRIBUTE_NONNULL(1);
//
// class C {
// /* arg_a cannot be null, but arg_b can */
// void Method(void* arg_a, void* arg_b) ABSL_ATTRIBUTE_NONNULL(2);
//
// /* arg_a cannot be null, but arg_b can */
// static void StaticMethod(void* arg_a, void* arg_b)
// ABSL_ATTRIBUTE_NONNULL(1);
// };
//
// If no arguments are provided, then all pointer arguments should be non-null.
//
// /* No pointer arguments may be null. */
// void Function(void* arg_a, void* arg_b, int arg_c) ABSL_ATTRIBUTE_NONNULL();
//
// NOTE: The GCC nonnull attribute actually accepts a list of arguments, but
// ABSL_ATTRIBUTE_NONNULL does not.
#if ABSL_HAVE_ATTRIBUTE(nonnull) || (defined(__GNUC__) && !defined(__clang__))
#define ABSL_ATTRIBUTE_NONNULL(arg_index) __attribute__((nonnull(arg_index)))
#else
#define ABSL_ATTRIBUTE_NONNULL(...)
#endif
// ABSL_ATTRIBUTE_NORETURN
//
// Tells the compiler that a given function never returns.
#if ABSL_HAVE_ATTRIBUTE(noreturn) || (defined(__GNUC__) && !defined(__clang__))
#define ABSL_ATTRIBUTE_NORETURN __attribute__((noreturn))
#elif defined(_MSC_VER)
#define ABSL_ATTRIBUTE_NORETURN __declspec(noreturn)
#else
#define ABSL_ATTRIBUTE_NORETURN
#endif
// ABSL_ATTRIBUTE_NO_SANITIZE_ADDRESS
//
// Tells the AddressSanitizer (or other memory testing tools) to ignore a given
// function. Useful for cases when a function reads random locations on stack,
// calls _exit from a cloned subprocess, deliberately accesses buffer
// out of bounds or does other scary things with memory.
// NOTE: GCC supports AddressSanitizer(asan) since 4.8.
// https://gcc.gnu.org/gcc-4.8/changes.html
#if ABSL_HAVE_ATTRIBUTE(no_sanitize_address)
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
#define ABSL_ATTRIBUTE_NO_SANITIZE_ADDRESS __attribute__((no_sanitize_address))
#else
#define ABSL_ATTRIBUTE_NO_SANITIZE_ADDRESS
#endif
// ABSL_ATTRIBUTE_NO_SANITIZE_MEMORY
//
// Tells the MemorySanitizer to relax the handling of a given function. All "Use
// of uninitialized value" warnings from such functions will be suppressed, and
// all values loaded from memory will be considered fully initialized. This
// attribute is similar to the ABSL_ATTRIBUTE_NO_SANITIZE_ADDRESS attribute
// above, but deals with initialized-ness rather than addressability issues.
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
// NOTE: MemorySanitizer(msan) is supported by Clang but not GCC.
#if ABSL_HAVE_ATTRIBUTE(no_sanitize_memory)
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
#define ABSL_ATTRIBUTE_NO_SANITIZE_MEMORY __attribute__((no_sanitize_memory))
#else
#define ABSL_ATTRIBUTE_NO_SANITIZE_MEMORY
#endif
// ABSL_ATTRIBUTE_NO_SANITIZE_THREAD
//
// Tells the ThreadSanitizer to not instrument a given function.
// NOTE: GCC supports ThreadSanitizer(tsan) since 4.8.
// https://gcc.gnu.org/gcc-4.8/changes.html
#if ABSL_HAVE_ATTRIBUTE(no_sanitize_thread)
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
#define ABSL_ATTRIBUTE_NO_SANITIZE_THREAD __attribute__((no_sanitize_thread))
#else
#define ABSL_ATTRIBUTE_NO_SANITIZE_THREAD
#endif
// ABSL_ATTRIBUTE_NO_SANITIZE_UNDEFINED
//
// Tells the UndefinedSanitizer to ignore a given function. Useful for cases
// where certain behavior (eg. division by zero) is being used intentionally.
// NOTE: GCC supports UndefinedBehaviorSanitizer(ubsan) since 4.9.
// https://gcc.gnu.org/gcc-4.9/changes.html
#if ABSL_HAVE_ATTRIBUTE(no_sanitize_undefined)
#define ABSL_ATTRIBUTE_NO_SANITIZE_UNDEFINED \
__attribute__((no_sanitize_undefined))
#elif ABSL_HAVE_ATTRIBUTE(no_sanitize)
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
#define ABSL_ATTRIBUTE_NO_SANITIZE_UNDEFINED \
__attribute__((no_sanitize("undefined")))
#else
#define ABSL_ATTRIBUTE_NO_SANITIZE_UNDEFINED
#endif
// ABSL_ATTRIBUTE_NO_SANITIZE_CFI
//
// Tells the ControlFlowIntegrity sanitizer to not instrument a given function.
// See https://clang.llvm.org/docs/ControlFlowIntegrity.html for details.
#if ABSL_HAVE_ATTRIBUTE(no_sanitize)
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
#define ABSL_ATTRIBUTE_NO_SANITIZE_CFI __attribute__((no_sanitize("cfi")))
#else
#define ABSL_ATTRIBUTE_NO_SANITIZE_CFI
#endif
// ABSL_ATTRIBUTE_NO_SANITIZE_SAFESTACK
//
// Tells the SafeStack to not instrument a given function.
// See https://clang.llvm.org/docs/SafeStack.html for details.
#if ABSL_HAVE_ATTRIBUTE(no_sanitize)
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
#define ABSL_ATTRIBUTE_NO_SANITIZE_SAFESTACK \
__attribute__((no_sanitize("safe-stack")))
#else
#define ABSL_ATTRIBUTE_NO_SANITIZE_SAFESTACK
#endif
// ABSL_ATTRIBUTE_RETURNS_NONNULL
//
// Tells the compiler that a particular function never returns a null pointer.
Export of internal Abseil changes -- f6fbb03bff276e72123e8590519079e87732ae62 by Abseil Team <absl-team@google.com>: Replace static absl::Mutex with SpinLock in absl::Cords to avoid static initializers by absl::Mutex destructors PiperOrigin-RevId: 370694199 -- 654b7d9edfdc24f226990b2b46cbf91451a1d92a by Martijn Vels <mvels@google.com>: Implement global data for CordzInfo in an ODR hardened way This change puts the global data into a global list structure, and stores a reference to the global list in the handle itself. This hardens the implementation against ODR violations where info pointers are crossing dynamic library boundaries which are privately loaded. PiperOrigin-RevId: 370673045 -- 712dba768e66ee2ba85d6010829c617cd2af6ba7 by Martijn Vels <mvels@google.com>: Intrument Cord::operator= for Cordz PiperOrigin-RevId: 370659149 -- c0b347a2289e151b72680269332e264b8fa989c0 by Matt Kulukundis <kfm@google.com>: Fix test guards for ABSL_ATTRIBUTE_RETURNS_NONNULL PiperOrigin-RevId: 370594807 -- c2bedaa3472ef223f907de2604f9b9b58852ec5f by Martijn Vels <mvels@google.com>: Add new Cordz instrumentation on GetAppendRegion. PiperOrigin-RevId: 370587761 -- 84fbfcc852697d509f6094482b86e84743a6b331 by Martijn Vels <mvels@google.com>: Add instrumentation on Cord::Apppend(string_view) PiperOrigin-RevId: 370576590 -- 9e077390b8ca2239e1cb7bfbe1d5a04f2fc11d30 by Abseil Team <absl-team@google.com>: Google-internal changes only. PiperOrigin-RevId: 370558424 -- fb53c149eb2364ea34e3a67235f873866618b8ac by Matt Kulukundis <kfm@google.com>: Update config.h macros with a few useful helpers to simplify version checking PiperOrigin-RevId: 370557684 -- abf8142e99b9ff7e15f6528a357f1005461950b0 by Martijn Vels <mvels@google.com>: clang-format cord PiperOrigin-RevId: 370549371 -- e555985eabe63fcf0e980e9c433dd84caffec191 by Martijn Vels <mvels@google.com>: Add MaybeUntrackCord() function This function is near identical to the old UntrackCord() but allows info to be null, moving the cord.is_profiled() branch into CordzInfo. PiperOrigin-RevId: 370528447 -- 3883538efe4601f7864bda70a50d868bb383c63b by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 370503186 -- a9514b65542fde1bc73584e6f3c1c4b3a05f215f by Derek Mauro <dmauro@google.com>: Add -Winvalid-constexpr to warning options for LLVM PiperOrigin-RevId: 370455171 -- d8a3966de2cf15a2dc28e17e49a3d27d205eca92 by Martijn Vels <mvels@google.com>: Add naive UniqueGenerator<T, kMaxValues, ...> to avoid flakes from dup random values. PiperOrigin-RevId: 370179772 -- 46d0caa1a12b68a5998d4f919e20f0f83b9286f8 by Martijn Vels <mvels@google.com>: Add new Cordz instrumentation on PrependTree. PiperOrigin-RevId: 370138969 GitOrigin-RevId: f6fbb03bff276e72123e8590519079e87732ae62 Change-Id: Ifa4c00a5c7b01198ee367a3253bea6b66612135e
4 years ago
#if ABSL_HAVE_ATTRIBUTE(returns_nonnull)
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
#define ABSL_ATTRIBUTE_RETURNS_NONNULL __attribute__((returns_nonnull))
#else
#define ABSL_ATTRIBUTE_RETURNS_NONNULL
#endif
// ABSL_HAVE_ATTRIBUTE_SECTION
//
// Indicates whether labeled sections are supported. Weak symbol support is
// a prerequisite. Labeled sections are not supported on Darwin/iOS.
#ifdef ABSL_HAVE_ATTRIBUTE_SECTION
#error ABSL_HAVE_ATTRIBUTE_SECTION cannot be directly set
#elif (ABSL_HAVE_ATTRIBUTE(section) || \
(defined(__GNUC__) && !defined(__clang__))) && \
!defined(__APPLE__) && ABSL_HAVE_ATTRIBUTE_WEAK
#define ABSL_HAVE_ATTRIBUTE_SECTION 1
// ABSL_ATTRIBUTE_SECTION
//
// Tells the compiler/linker to put a given function into a section and define
// `__start_ ## name` and `__stop_ ## name` symbols to bracket the section.
// This functionality is supported by GNU linker. Any function annotated with
// `ABSL_ATTRIBUTE_SECTION` must not be inlined, or it will be placed into
// whatever section its caller is placed into.
//
#ifndef ABSL_ATTRIBUTE_SECTION
#define ABSL_ATTRIBUTE_SECTION(name) \
__attribute__((section(#name))) __attribute__((noinline))
#endif
// ABSL_ATTRIBUTE_SECTION_VARIABLE
//
// Tells the compiler/linker to put a given variable into a section and define
// `__start_ ## name` and `__stop_ ## name` symbols to bracket the section.
// This functionality is supported by GNU linker.
#ifndef ABSL_ATTRIBUTE_SECTION_VARIABLE
#define ABSL_ATTRIBUTE_SECTION_VARIABLE(name) __attribute__((section(#name)))
#endif
// ABSL_DECLARE_ATTRIBUTE_SECTION_VARS
//
// A weak section declaration to be used as a global declaration
// for ABSL_ATTRIBUTE_SECTION_START|STOP(name) to compile and link
// even without functions with ABSL_ATTRIBUTE_SECTION(name).
// ABSL_DEFINE_ATTRIBUTE_SECTION should be in the exactly one file; it's
// a no-op on ELF but not on Mach-O.
//
#ifndef ABSL_DECLARE_ATTRIBUTE_SECTION_VARS
#define ABSL_DECLARE_ATTRIBUTE_SECTION_VARS(name) \
extern char __start_##name[] ABSL_ATTRIBUTE_WEAK; \
extern char __stop_##name[] ABSL_ATTRIBUTE_WEAK
#endif
#ifndef ABSL_DEFINE_ATTRIBUTE_SECTION_VARS
#define ABSL_INIT_ATTRIBUTE_SECTION_VARS(name)
#define ABSL_DEFINE_ATTRIBUTE_SECTION_VARS(name)
#endif
// ABSL_ATTRIBUTE_SECTION_START
//
// Returns `void*` pointers to start/end of a section of code with
// functions having ABSL_ATTRIBUTE_SECTION(name).
// Returns 0 if no such functions exist.
// One must ABSL_DECLARE_ATTRIBUTE_SECTION_VARS(name) for this to compile and
// link.
//
#define ABSL_ATTRIBUTE_SECTION_START(name) \
(reinterpret_cast<void *>(__start_##name))
#define ABSL_ATTRIBUTE_SECTION_STOP(name) \
(reinterpret_cast<void *>(__stop_##name))
#else // !ABSL_HAVE_ATTRIBUTE_SECTION
#define ABSL_HAVE_ATTRIBUTE_SECTION 0
// provide dummy definitions
#define ABSL_ATTRIBUTE_SECTION(name)
#define ABSL_ATTRIBUTE_SECTION_VARIABLE(name)
#define ABSL_INIT_ATTRIBUTE_SECTION_VARS(name)
#define ABSL_DEFINE_ATTRIBUTE_SECTION_VARS(name)
#define ABSL_DECLARE_ATTRIBUTE_SECTION_VARS(name)
#define ABSL_ATTRIBUTE_SECTION_START(name) (reinterpret_cast<void *>(0))
#define ABSL_ATTRIBUTE_SECTION_STOP(name) (reinterpret_cast<void *>(0))
#endif // ABSL_ATTRIBUTE_SECTION
// ABSL_ATTRIBUTE_STACK_ALIGN_FOR_OLD_LIBC
//
// Support for aligning the stack on 32-bit x86.
#if ABSL_HAVE_ATTRIBUTE(force_align_arg_pointer) || \
(defined(__GNUC__) && !defined(__clang__))
#if defined(__i386__)
#define ABSL_ATTRIBUTE_STACK_ALIGN_FOR_OLD_LIBC \
__attribute__((force_align_arg_pointer))
#define ABSL_REQUIRE_STACK_ALIGN_TRAMPOLINE (0)
#elif defined(__x86_64__)
#define ABSL_REQUIRE_STACK_ALIGN_TRAMPOLINE (1)
#define ABSL_ATTRIBUTE_STACK_ALIGN_FOR_OLD_LIBC
#else // !__i386__ && !__x86_64
#define ABSL_REQUIRE_STACK_ALIGN_TRAMPOLINE (0)
#define ABSL_ATTRIBUTE_STACK_ALIGN_FOR_OLD_LIBC
#endif // __i386__
#else
#define ABSL_ATTRIBUTE_STACK_ALIGN_FOR_OLD_LIBC
#define ABSL_REQUIRE_STACK_ALIGN_TRAMPOLINE (0)
#endif
// ABSL_MUST_USE_RESULT
//
// Tells the compiler to warn about unused results.
//
// When annotating a function, it must appear as the first part of the
// declaration or definition. The compiler will warn if the return value from
// such a function is unused:
//
// ABSL_MUST_USE_RESULT Sprocket* AllocateSprocket();
// AllocateSprocket(); // Triggers a warning.
//
// When annotating a class, it is equivalent to annotating every function which
// returns an instance.
//
// class ABSL_MUST_USE_RESULT Sprocket {};
// Sprocket(); // Triggers a warning.
//
// Sprocket MakeSprocket();
// MakeSprocket(); // Triggers a warning.
//
// Note that references and pointers are not instances:
//
// Sprocket* SprocketPointer();
// SprocketPointer(); // Does *not* trigger a warning.
//
// ABSL_MUST_USE_RESULT allows using cast-to-void to suppress the unused result
// warning. For that, warn_unused_result is used only for clang but not for gcc.
// https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66425
//
// Note: past advice was to place the macro after the argument list.
#if ABSL_HAVE_ATTRIBUTE(nodiscard)
#define ABSL_MUST_USE_RESULT [[nodiscard]]
#elif defined(__clang__) && ABSL_HAVE_ATTRIBUTE(warn_unused_result)
#define ABSL_MUST_USE_RESULT __attribute__((warn_unused_result))
#else
#define ABSL_MUST_USE_RESULT
#endif
// ABSL_ATTRIBUTE_HOT, ABSL_ATTRIBUTE_COLD
//
// Tells GCC that a function is hot or cold. GCC can use this information to
// improve static analysis, i.e. a conditional branch to a cold function
// is likely to be not-taken.
// This annotation is used for function declarations.
//
// Example:
//
// int foo() ABSL_ATTRIBUTE_HOT;
#if ABSL_HAVE_ATTRIBUTE(hot) || (defined(__GNUC__) && !defined(__clang__))
#define ABSL_ATTRIBUTE_HOT __attribute__((hot))
#else
#define ABSL_ATTRIBUTE_HOT
#endif
#if ABSL_HAVE_ATTRIBUTE(cold) || (defined(__GNUC__) && !defined(__clang__))
#define ABSL_ATTRIBUTE_COLD __attribute__((cold))
#else
#define ABSL_ATTRIBUTE_COLD
#endif
// ABSL_XRAY_ALWAYS_INSTRUMENT, ABSL_XRAY_NEVER_INSTRUMENT, ABSL_XRAY_LOG_ARGS
//
// We define the ABSL_XRAY_ALWAYS_INSTRUMENT and ABSL_XRAY_NEVER_INSTRUMENT
// macro used as an attribute to mark functions that must always or never be
// instrumented by XRay. Currently, this is only supported in Clang/LLVM.
//
// For reference on the LLVM XRay instrumentation, see
// http://llvm.org/docs/XRay.html.
//
// A function with the XRAY_ALWAYS_INSTRUMENT macro attribute in its declaration
// will always get the XRay instrumentation sleds. These sleds may introduce
// some binary size and runtime overhead and must be used sparingly.
//
// These attributes only take effect when the following conditions are met:
//
// * The file/target is built in at least C++11 mode, with a Clang compiler
// that supports XRay attributes.
// * The file/target is built with the -fxray-instrument flag set for the
// Clang/LLVM compiler.
// * The function is defined in the translation unit (the compiler honors the
// attribute in either the definition or the declaration, and must match).
//
// There are cases when, even when building with XRay instrumentation, users
// might want to control specifically which functions are instrumented for a
// particular build using special-case lists provided to the compiler. These
// special case lists are provided to Clang via the
// -fxray-always-instrument=... and -fxray-never-instrument=... flags. The
// attributes in source take precedence over these special-case lists.
//
// To disable the XRay attributes at build-time, users may define
// ABSL_NO_XRAY_ATTRIBUTES. Do NOT define ABSL_NO_XRAY_ATTRIBUTES on specific
// packages/targets, as this may lead to conflicting definitions of functions at
// link-time.
//
// XRay isn't currently supported on Android:
// https://github.com/android/ndk/issues/368
#if ABSL_HAVE_CPP_ATTRIBUTE(clang::xray_always_instrument) && \
!defined(ABSL_NO_XRAY_ATTRIBUTES) && !defined(__ANDROID__)
#define ABSL_XRAY_ALWAYS_INSTRUMENT [[clang::xray_always_instrument]]
#define ABSL_XRAY_NEVER_INSTRUMENT [[clang::xray_never_instrument]]
#if ABSL_HAVE_CPP_ATTRIBUTE(clang::xray_log_args)
#define ABSL_XRAY_LOG_ARGS(N) \
[[clang::xray_always_instrument, clang::xray_log_args(N)]]
#else
#define ABSL_XRAY_LOG_ARGS(N) [[clang::xray_always_instrument]]
#endif
#else
#define ABSL_XRAY_ALWAYS_INSTRUMENT
#define ABSL_XRAY_NEVER_INSTRUMENT
#define ABSL_XRAY_LOG_ARGS(N)
#endif
// ABSL_ATTRIBUTE_REINITIALIZES
//
// Indicates that a member function reinitializes the entire object to a known
// state, independent of the previous state of the object.
//
// The clang-tidy check bugprone-use-after-move allows member functions marked
// with this attribute to be called on objects that have been moved from;
// without the attribute, this would result in a use-after-move warning.
#if ABSL_HAVE_CPP_ATTRIBUTE(clang::reinitializes)
#define ABSL_ATTRIBUTE_REINITIALIZES [[clang::reinitializes]]
#else
#define ABSL_ATTRIBUTE_REINITIALIZES
#endif
// -----------------------------------------------------------------------------
// Variable Attributes
// -----------------------------------------------------------------------------
// ABSL_ATTRIBUTE_UNUSED
//
// Prevents the compiler from complaining about variables that appear unused.
//
// For code or headers that are assured to only build with C++17 and up, prefer
// just using the standard '[[maybe_unused]]' directly over this macro.
//
// Due to differences in positioning requirements between the old, compiler
// specific __attribute__ syntax and the now standard [[maybe_unused]], this
// macro does not attempt to take advantage of '[[maybe_unused]]'.
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
#if ABSL_HAVE_ATTRIBUTE(unused) || (defined(__GNUC__) && !defined(__clang__))
#undef ABSL_ATTRIBUTE_UNUSED
#define ABSL_ATTRIBUTE_UNUSED __attribute__((__unused__))
#else
#define ABSL_ATTRIBUTE_UNUSED
#endif
// ABSL_ATTRIBUTE_INITIAL_EXEC
//
// Tells the compiler to use "initial-exec" mode for a thread-local variable.
// See http://people.redhat.com/drepper/tls.pdf for the gory details.
#if ABSL_HAVE_ATTRIBUTE(tls_model) || (defined(__GNUC__) && !defined(__clang__))
#define ABSL_ATTRIBUTE_INITIAL_EXEC __attribute__((tls_model("initial-exec")))
#else
#define ABSL_ATTRIBUTE_INITIAL_EXEC
#endif
// ABSL_ATTRIBUTE_PACKED
//
// Instructs the compiler not to use natural alignment for a tagged data
// structure, but instead to reduce its alignment to 1. This attribute can
// either be applied to members of a structure or to a structure in its
// entirety. Applying this attribute (judiciously) to a structure in its
// entirety to optimize the memory footprint of very commonly-used structs is
// fine. Do not apply this attribute to a structure in its entirety if the
// purpose is to control the offsets of the members in the structure. Instead,
// apply this attribute only to structure members that need it.
//
// When applying ABSL_ATTRIBUTE_PACKED only to specific structure members the
// natural alignment of structure members not annotated is preserved. Aligned
// member accesses are faster than non-aligned member accesses even if the
// targeted microprocessor supports non-aligned accesses.
#if ABSL_HAVE_ATTRIBUTE(packed) || (defined(__GNUC__) && !defined(__clang__))
#define ABSL_ATTRIBUTE_PACKED __attribute__((__packed__))
#else
#define ABSL_ATTRIBUTE_PACKED
#endif
// ABSL_ATTRIBUTE_FUNC_ALIGN
//
// Tells the compiler to align the function start at least to certain
// alignment boundary
#if ABSL_HAVE_ATTRIBUTE(aligned) || (defined(__GNUC__) && !defined(__clang__))
#define ABSL_ATTRIBUTE_FUNC_ALIGN(bytes) __attribute__((aligned(bytes)))
#else
#define ABSL_ATTRIBUTE_FUNC_ALIGN(bytes)
#endif
// ABSL_FALLTHROUGH_INTENDED
//
// Annotates implicit fall-through between switch labels, allowing a case to
// indicate intentional fallthrough and turn off warnings about any lack of a
// `break` statement. The ABSL_FALLTHROUGH_INTENDED macro should be followed by
// a semicolon and can be used in most places where `break` can, provided that
// no statements exist between it and the next switch label.
//
// Example:
//
// switch (x) {
// case 40:
// case 41:
// if (truth_is_out_there) {
// ++x;
// ABSL_FALLTHROUGH_INTENDED; // Use instead of/along with annotations
// // in comments
// } else {
// return x;
// }
// case 42:
// ...
//
// Notes: When supported, GCC and Clang can issue a warning on switch labels
// with unannotated fallthrough using the warning `-Wimplicit-fallthrough`. See
// clang documentation on language extensions for details:
// https://clang.llvm.org/docs/AttributeReference.html#fallthrough-clang-fallthrough
//
// When used with unsupported compilers, the ABSL_FALLTHROUGH_INTENDED macro has
// no effect on diagnostics. In any case this macro has no effect on runtime
// behavior and performance of code.
#ifdef ABSL_FALLTHROUGH_INTENDED
#error "ABSL_FALLTHROUGH_INTENDED should not be defined."
#elif ABSL_HAVE_CPP_ATTRIBUTE(fallthrough)
#define ABSL_FALLTHROUGH_INTENDED [[fallthrough]]
#elif ABSL_HAVE_CPP_ATTRIBUTE(clang::fallthrough)
#define ABSL_FALLTHROUGH_INTENDED [[clang::fallthrough]]
#elif ABSL_HAVE_CPP_ATTRIBUTE(gnu::fallthrough)
#define ABSL_FALLTHROUGH_INTENDED [[gnu::fallthrough]]
#else
#define ABSL_FALLTHROUGH_INTENDED \
do { \
} while (0)
#endif
// ABSL_DEPRECATED()
//
// Marks a deprecated class, struct, enum, function, method and variable
// declarations. The macro argument is used as a custom diagnostic message (e.g.
// suggestion of a better alternative).
//
// Examples:
//
// class ABSL_DEPRECATED("Use Bar instead") Foo {...};
//
// ABSL_DEPRECATED("Use Baz() instead") void Bar() {...}
//
// template <typename T>
// ABSL_DEPRECATED("Use DoThat() instead")
// void DoThis();
//
// Every usage of a deprecated entity will trigger a warning when compiled with
// clang's `-Wdeprecated-declarations` option. This option is turned off by
// default, but the warnings will be reported by clang-tidy.
#if defined(__clang__) && defined(__cplusplus) && __cplusplus >= 201103L
#define ABSL_DEPRECATED(message) __attribute__((deprecated(message)))
#endif
#ifndef ABSL_DEPRECATED
#define ABSL_DEPRECATED(message)
#endif
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
// ABSL_CONST_INIT
//
// A variable declaration annotated with the `ABSL_CONST_INIT` attribute will
// not compile (on supported platforms) unless the variable has a constant
// initializer. This is useful for variables with static and thread storage
// duration, because it guarantees that they will not suffer from the so-called
// "static init order fiasco". Prefer to put this attribute on the most visible
// declaration of the variable, if there's more than one, because code that
// accesses the variable can then use the attribute for optimization.
//
// Example:
//
// class MyClass {
// public:
// ABSL_CONST_INIT static MyType my_var;
// };
//
// MyType MyClass::my_var = MakeMyType(...);
//
// Note that this attribute is redundant if the variable is declared constexpr.
#if ABSL_HAVE_CPP_ATTRIBUTE(clang::require_constant_initialization)
#define ABSL_CONST_INIT [[clang::require_constant_initialization]]
#else
#define ABSL_CONST_INIT
#endif // ABSL_HAVE_CPP_ATTRIBUTE(clang::require_constant_initialization)
// ABSL_ATTRIBUTE_PURE_FUNCTION
//
// ABSL_ATTRIBUTE_PURE_FUNCTION is used to annotate declarations of "pure"
// functions. A function is pure if its return value is only a function of its
// arguments. The pure attribute prohibits a function from modifying the state
// of the program that is observable by means other than inspecting the
// function's return value. Declaring such functions with the pure attribute
// allows the compiler to avoid emitting some calls in repeated invocations of
// the function with the same argument values.
//
// Example:
//
// ABSL_ATTRIBUTE_PURE_FUNCTION int64_t ToInt64Milliseconds(Duration d);
#if ABSL_HAVE_CPP_ATTRIBUTE(gnu::pure)
#define ABSL_ATTRIBUTE_PURE_FUNCTION [[gnu::pure]]
#elif ABSL_HAVE_ATTRIBUTE(pure)
#define ABSL_ATTRIBUTE_PURE_FUNCTION __attribute__((pure))
#else
#define ABSL_ATTRIBUTE_PURE_FUNCTION
#endif
Export of internal Abseil changes -- 6cfac39ea1266f01f195de5eb4c9a6fc9ea9b20a by Matt Kulukundis <kfm@google.com>: Fix a typo PiperOrigin-RevId: 366174890 -- 6ee8c58647aef171d394e59fa06f9bf8cd0306ec by Derek Mauro <dmauro@google.com>: Adds `ABSL_ATTRIBUTE_LIFETIME_BOUND` and applies it to the `const std::string&` constructor of `absl::string_view`. Compilers that support this attribute will emit a warning if the parameter does not have sufficient lifetime. PiperOrigin-RevId: 366027738 -- b944427d96e4b436b8fa0fe396c2a1118dbbbd13 by Abseil Team <absl-team@google.com>: Calls to `ResetToEmpty()` from `ClearSlow` use ~3% of the time in assignment. However, `ClearSlow()` is only used in contexts where `data_` is immediately reassigned. Rename `ClearSlow()` into `UnrefTree()` and remove `data_` resetting. PiperOrigin-RevId: 365977213 -- 7428b3147a5672c8bb55649efa3a1cfe19b52a8b by Abseil Team <absl-team@google.com>: Fix CordRepRing diabolical growth The 'Mutable' function in CordRepRing was over-eager in doubling capacity, which lead to 'ludicrous' growth in the diabolical test case as added to cord_test. This CL fixes the doubling for growing shared reps, and tempers CordRepRing growth for non shared capacity to 1.5 instead of 2, which is more inline with a conservative growth we also have in tree cord. After this change, CordRepRing no longer swamps the heap into the shadow realm, and is in effect reducing the memory used compared to the tree implementation. With a diabolical 5000 bytes growth pattern: Tree cord: 1523520 bytes Ring cord: 274232 bytes PiperOrigin-RevId: 365915857 -- f24d4aee48b03c5a7980664df903f947cbb198e8 by Andy Getzendanner <durandal@google.com>: Import of CCTZ from GitHub. PiperOrigin-RevId: 365873932 GitOrigin-RevId: 6cfac39ea1266f01f195de5eb4c9a6fc9ea9b20a Change-Id: I9737aa215ac732c9785a1d0032c77aba62330f12
4 years ago
// ABSL_ATTRIBUTE_LIFETIME_BOUND indicates that a resource owned by a function
// parameter or implicit object parameter is retained by the return value of the
// annotated function (or, for a parameter of a constructor, in the value of the
// constructed object). This attribute causes warnings to be produced if a
// temporary object does not live long enough.
//
// When applied to a reference parameter, the referenced object is assumed to be
// retained by the return value of the function. When applied to a non-reference
// parameter (for example, a pointer or a class type), all temporaries
// referenced by the parameter are assumed to be retained by the return value of
// the function.
//
// See also the upstream documentation:
// https://clang.llvm.org/docs/AttributeReference.html#lifetimebound
#if ABSL_HAVE_CPP_ATTRIBUTE(clang::lifetimebound)
#define ABSL_ATTRIBUTE_LIFETIME_BOUND [[clang::lifetimebound]]
#elif ABSL_HAVE_ATTRIBUTE(lifetimebound)
#define ABSL_ATTRIBUTE_LIFETIME_BOUND __attribute__((lifetimebound))
#else
#define ABSL_ATTRIBUTE_LIFETIME_BOUND
#endif
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
#endif // ABSL_BASE_ATTRIBUTES_H_