August 2021 - Boost-users - lists.preview.boost.org

Clarification on statechart post_event vs process_event within react()
by John McCabe 10 Jan '22

10 Jan '22

Hi, Apologies if this has already been answered/clarified, but my search of the internet to try to get a definitive answer, at least one I recognised as being definitive, was unsuccessful. Below I have posted some source code, based on the stopwatch example from the tutorial, but using an 'external' object to delegate the actual handling of event stuff to and using the state, rather than simple_state base class, and some other tweaks. This is a cut down version of something that someone I work with has devised, but I just wanted to get a clear answer on whether process_event() should be getting called in cases like this. I think it probably should not since, in the code below, you change state (with a state object's destructor being executed) whilst you're executing that state's react(EvStartStop) function. If someone can confirm this is the correct interpretation that would be very much appreciated. Any questions, feel free to ask (BTW - I'm compiling for C++17) John PS - In case it's of any use, the output from running this code is: ------ Entered Active Entered Stopped Stopped - payload2 m_charValue: a Exited ~Stopped <<<--- This is the bit I don't like the look of because... Entered Running Active - payload2 m_charValue: a <<<--- ... this happens from the forward_event() call payload1 value1: 20 payload1 value2: 200.2 Exited ~Running Entered Stopped Exiting Program Exited ~Stopped Exited ~Active ------ If I swap the //*** comments to the doProcess() lines instead of doPost(), the output seems more sensible ------ Entered Active Entered Stopped Stopped - payload2 m_charValue: a Exited ~Stopped Entered Running Active - payload2 m_charValue: a payload1 value1: 20 payload1 value2: 200.2 Exited ~Running Entered Stopped Exiting Program Exited ~Stopped Exited ~Active ------ ========== code =========== #include <boost/statechart/event.hpp> #include <boost/statechart/state_machine.hpp> #include <boost/statechart/state.hpp> #include <boost/statechart/transition.hpp> #include <boost/statechart/custom_reaction.hpp> #include <boost/mpl/list.hpp> #include <iostream> #include <string> namespace sc = boost::statechart; namespace mpl = boost::mpl; // Forward declare a handler 'class' that's going to have activity delegated to it struct Handler; // Normal example start/stop and reset events struct EvStartStop : sc::event<EvStartStop> {}; struct EvReset : sc::event<EvReset> {}; // A couple of events that carry data with them struct EvPayload1 : sc::event<EvPayload1> { EvPayload1(int value1, double value2) : m_value1 { value1 }, m_value2 { value2 } {}; int m_value1 { 0 }; double m_value2 { 0.0 }; }; struct EvPayload2 : sc::event<EvPayload2> { EvPayload2(const char& charValue) : m_charValue { charValue } {}; char m_charValue { false }; }; // Back to the example bits struct Active; struct StopWatch : sc::state_machine<StopWatch, Active> { typedef sc::transition<EvReset, Active> reactions; // The extra 'handler' stuff void setHandler(Handler* handler) { m_handler = handler; }; Handler* getHandler() const { return m_handler; }; Handler* m_handler { nullptr }; // Used to make post_event() accessible from an external class...seems unpleasant, but... using sc::state_machine<StopWatch, Active>::post_event; }; // The implementation of the Handler class; created with a reference to a StopWatch object struct Handler { Handler(StopWatch& stopWatch) : m_stopWatch { stopWatch } {}; void doPost() { m_stopWatch.post_event(EvStartStop()); } void doProcess() { m_stopWatch.process_event(EvStartStop()); } StopWatch& m_stopWatch; }; struct Stopped; struct Active : sc::state<Active, StopWatch, Stopped> { Active(my_context ctx) : my_base { ctx } { std::cout << "Entered Active" << std::endl; }; ~Active() { std::cout << "Exited ~Active" << std::endl; }; typedef mpl::list<sc::custom_reaction<EvPayload1>, sc::custom_reaction<EvPayload2>, sc::transition<EvReset, Active>> reactions; sc::result react(const EvPayload1& payload1) { std::cout << "payload1 Event NOT HANDLED" << std::endl; return discard_event(); } sc::result react(const EvPayload2& payload2) { std::cout << "Active - payload2 m_charValue: " << payload2.m_charValue << std::endl; return discard_event(); } }; struct Running : sc::state<Running, Active> { Running(my_context ctx) : my_base { ctx } { std::cout << "Entered Running" << std::endl; }; ~Running() { std::cout << "Exited ~Running" << std::endl; }; typedef mpl::list<sc::custom_reaction<EvPayload1>, sc::transition<EvStartStop, Stopped>> reactions; sc::result react(const EvPayload1& payload1) { std::cout << "payload1 value1: " << payload1.m_value1 << std::endl; std::cout << "payload1 value2: " << payload1.m_value2 << std::endl; //***context<StopWatch>().getHandler()->doPost(); context<StopWatch>().getHandler()->doProcess(); return discard_event(); } }; struct Stopped : sc::state<Stopped, Active> { Stopped(my_context ctx) : my_base { ctx } { std::cout << "Entered Stopped" << std::endl; }; ~Stopped() { std::cout << "Exited ~Stopped" << std::endl; }; typedef mpl::list<sc::custom_reaction<EvPayload2>, sc::transition<EvStartStop, Running>> reactions; sc::result react(const EvPayload2& payload2) { std::cout << "Stopped - payload2 m_charValue: " << payload2.m_charValue << std::endl; //***context<StopWatch>().getHandler()->doPost(); context<StopWatch>().getHandler()->doProcess(); return forward_event(); } }; int main() { StopWatch myWatch; Handler myHandler(myWatch); myWatch.setHandler(&myHandler); myWatch.initiate(); myWatch.process_event(EvPayload2 { 'a' }); myWatch.process_event(EvPayload1 { 20, 200.2 }); std::cout << "Exiting Program" << std::endl; return 0; }

1 1

[Boost.Asio] Segmentation fault when ::getaddrinfo() returns EAI_SYSTEM without system error code set (errno = 0) - is this a bug?
by Nils Frielinghaus 04 Oct '21

04 Oct '21

Hello Boost users, I experienced a segmentation fault inside Boost in the following scenario: o There was an illegal DNS entry o When resolving that illegal DNS entry on a Yocto-based embedded Linux distribution, ::getaddrinfo() returned with the error code EAI_SYSTEM but errno was set to 0 o The function translate_addrinfo_error(int) in boost/asio/detail/impl/socket_ops.ipp translates this into boost::system::error_code(0, boost::asio::error::get_system_category()) o This is then interpreted as a success by the calling function and the other output parameters of ::getaddrinfo() are accessed which resulted in the segmentation fault My question would be whether this should be considered a bug in the Boost ASIO library? One could argue that the system should not return EAI_SYSTEM with errno not being set properly. But one could also argue that returning EAI_SYSTEM indicates that something went wrong and that the output parameters are not to be accessed. Nils ------ Nils Frielinghaus

4 11

[Compute] automatically determine work group size
by Stephan Menzel 30 Aug '21

30 Aug '21

Hello everyone, I have recently made the switched an existing OpenCL application to boost compute. For reasons mostly. Now the transition went well for the most part but I struggle with one feature that appears to be missing. The determination of the local work group size. Originally the code was using clEnqueueNDRangeKernel() which has an option for the local work group size called local_work_size. When left empty, the system will determine that value itself at a small cost. Boost compute doesn't seem to have that. When I process an image for example: command_queue.enqueue_nd_range_kernel(kernel, bc::extents<2>{ 0, current_line }, bc::extents<2>{ width, height }, bc::extents<2>{ 128, 1 }); The last parameter is very hard for me to understand. I know roughly what it is and I understand it has to be dividable by 64, the total size (width * height) however has to be a multiple of it too, so there's a bunch of rules that apply and I would much rather leave that auto-determined as it was before. Leaving it bc::extents<2>{ 0, 0 } throws an assertion. So, essentially: Is there a way to leave it empty to be auto-defined? Cheers, Stephan

1 0

boost/rational.hpp and VS2019
by John Emmas 24 Aug '21

24 Aug '21

I'm guessing this might be solvable by updating to a newer version of boost (currently I'm running ver 1.71.0) but before updating, I wondered if there might be something obvious that 'm missing.... I'm building with Visual Studio 2019 and I get a whole bunch of errors simply by adding this line to an empty source file:- #include "boost/rational.hpp" It seems to be the same 6 errors but at various different lines - i.e. these errors, over & over again:- Visual Studio\2019\Community\VC\Tools\MSVC\14.29.30133\include\ammintrin.h(146,19): error C4430: missing type specifier - int assumed. Note: C++ does not support default-int Visual Studio\2019\Community\VC\Tools\MSVC\14.29.30133\include\ammintrin.h(146,8): error C2146: syntax error: missing ';' before identifier '_mm_macc_ps' Visual Studio\2019\Community\VC\Tools\MSVC\14.29.30133\include\ammintrin.h(147,20): error C4430: missing type specifier - int assumed. Note: C++ does not support default-int Visual Studio\2019\Community\VC\Tools\MSVC\14.29.30133\include\ammintrin.h(147,9): error C2146: syntax error: missing ';' before identifier '_mm_macc_pd' Visual Studio\2019\Community\VC\Tools\MSVC\14.29.30133\include\ammintrin.h(148,19): error C4430: missing type specifier - int assumed. Note: C++ does not support default-int Visual Studio\2019\Community\VC\Tools\MSVC\14.29.30133\include\ammintrin.h(148,1): error C2086: 'int __m128': redefinition I just wondered if this might ring a bell with someone? e.g. might there be some preprocesser #define that I should've defined? (or a #define that I shouldn't be defining...) Thanks for any suggestions, John

3 18

Boost's Stage\Lib *.a Files
by Alexander Tanseco 23 Aug '21

23 Aug '21

I am running a tutorial program and I found out that in the program's build options under Linker settings/Link Libraries I can put both or either of the ff and the program compiles & runs without errors. What is the difference between the two? libboost_filesystem-mgw8-mt-d-x64-1_77.dll.a OR libboost_filesystem-mgw8-mt-x64-1_77.dll.a

2 1

[multiprecission] Large divides layered on 128 bit divides
by Neill Clift 21 Aug '21

21 Aug '21

Hi, The architecture of cpp_int being build on 64 bit arithmetic using 128 bit double_limb_type is interesting. I have a question on the large divide (divide_unsigned_helper). It uses the upper portions of the large integers to get an estimation of the quotient. Subtracts out a multiple of that quotient and repeats. It does this in 128 bit values if available from the compiler: double_limb_type a = (static_cast<double_limb_type>(prem[r_order]) << CppInt1::limb_bits) | prem[r_order - 1]; double_limb_type b = py[y_order]; double_limb_type v = a / b; The compiler emulates this operation in the routine __udivmodti4 which itself uses an iterative approach. It seems to use a pretty basic shift and subtract algorithm mind you. As a general rule for multiprecision is it OK to layer the Knuth like algorithm D on top of each other this way. I have no idea myself but wonder if this is a known issue. I would have guessed that it made sense to do a 64 by 64 bit divide to guess the quotient and repeat. Thoughts? Thanks. Neill.

2 2

[multiprecission] some thoughts on add_unsigned.hpp
by Neill Clift 19 Aug '21

19 Aug '21

Hi, In my program I see a lot of time taken up by add/subtract_unsigned. I have a couple of thoughts on the code: Carry propagation outside the range of the smallest number can be improved: for (; i < x && carry; ++i) carry = ::boost::multiprecision::detail::addcarry_limb(carry, pa[i], 0, pr + i); Maybe this: for (; i < x && carry; ++i) carry = ::boost::multiprecision::detail::addcarry_limb(0, pa[i], 1, pr + i); This generates just an inc on my machine but unfortunately it doesn't eliminate the loads and stores. I think on average the carry chain must be small though. You have the same situation on subtract. add_unsigned and subtract_unsigned are inconsistent at the tail end after carry propagation: add_unsigned: if (i == x && carry) { // We overflowed, need to add one more limb: result.resize(x + 1, x + 1); if (result.size() > x) result.limbs()[x] = static_cast<limb_type>(1u); } else std::copy(pa + i, pa + x, pr + i); <--- if I == x we do the copy. We also do the copy even if the output and input overlap subtract_unsigned: // Any remaining digits are the same as those in pa: if ((x != i) && (pa != pr)) <-------- we do these checks here and avoid the copy std_constexpr::copy(pa + i, pa + x, pr + i); I am seeing a lot of time being used by that general copy above in add_unsigned when it can be avoided. This stuff is totally awesome and works well in my program. So thanks for working on this stuff. Neill.

2 1

Boost Installation in Codeblocks
by Alexander Tanseco 18 Aug '21

18 Aug '21

To Gavin Lambert, many thanks for your help. I finally got my program running without errors with the ff settings: Global variable Lib: c:\boost_1_77_0\stage\lib Program's build options: Linker settings/Link libraries: blank Search Directories/Compiler: $(#boost) Search Directories/Linker: $(#boost.lib).

1 0

Boost uses unoptimized subtraction because of typo in intel_intrinsics.hpp
by Neill Clift 18 Aug '21

18 Aug '21

Hi, In my program I am seeing my code use the slower (I assume) subtract_unsigned_constexpr rather than subtract_unsigned with he borrow chain handled intrinsics. This seems to be happening because of these lines in intel_intrinsics.hpp: #if defined(__clang__) && (__clang__ < 9) // We appear to crash the compiler if we try to use these intrinsics? #undef BOOST_MP_HAS_IMMINTRIN_H #endif __clang__ is not the version but just the value 1. I expect this should be the major version instead: #define __clang__ 1 #define __clang_major__ 12 #define __clang_minor__ 0 As an aside 128bit support does work in windows with clang. I have to force it on like this: #define __STRICT_ANSI__ #define BOOST_HAS_INT128 #define BOOST_MP_HAS_IMMINTRIN_H #include <boost/multiprecision/cpp_int.hpp> Thanks. Neill.

2 1

boostrap fails with intel-linux
by Alain O' Miniussi 18 Aug '21

18 Aug '21

Hi I have a C++11 compiler: [alainm@zztop boost_1_77_0]$ icpc --version icpc (ICC) 2021.3.0 20210609 Copyright (C) 1985-2021 Intel Corporation. All rights reserved. [alainm@zztop boost_1_77_0]$ I can use it to build b2: [alainm@zztop boost_1_77_0]$ icpc --version icpc (ICC) 2021.3.0 20210609 Copyright (C) 1985-2021 Intel Corporation. All rights reserved. [alainm@zztop boost_1_77_0]$ ./tools/build/src/engine/build.sh --cxx=icpc ... ### Using 'cxx' toolset. >>> Why cxx ? ... icpc (ICC) 2021.3.0 20210609 Copyright (C) 1985-2021 Intel Corporation. All rights reserved. ..... > icpc -DNDEBUG builtins.cpp class.cpp command.cpp compile.cpp constants.cpp cwd.cpp debug.cpp debugger.cpp execcmd.cpp execnt.cpp execunix.cpp filesys.cpp filent.cpp fileunix.cpp frames.cpp function.cpp glob.cpp hash.cpp hcache.cpp hdrmacro.cpp headers.cpp jam_strings.cpp jam.cpp jamgram.cpp lists.cpp make.cpp make1.cpp md5.cpp mem.cpp modules.cpp native.cpp object.cpp option.cpp output.cpp parse.cpp pathnt.cpp pathsys.cpp pathunix.cpp regexp.cpp rules.cpp scan.cpp search.cpp startup.cpp subst.cpp sysinfo.cpp timestamp.cpp variable.cpp w32_getreg.cpp modules/order.cpp modules/path.cpp modules/property-set.cpp modules/regex.cpp modules/sequence.cpp modules/set.cpp -o b2 > cp b2 bjam [alainm@zztop boost_1_77_0]$ But bootstrap.sh fails to see it: alainm@zztop boost_1_77_0]$ ./bootstrap.sh --with-toolset=intel-linux Building B2 engine.. A C++11 capable compiler is required for building the B2 engine. Toolset 'intel-linux' does not appear to support C++11. ... ** Note, the C++11 capable compiler is _only_ required for building the B2 ** engine. The B2 build system allows for using any C++ level and any other ** supported language and resource in your projects. ... You can specify the toolset as the argument, i.e.: ./build.sh [options] gcc Toolsets supported by this script are: acc, clang, como, gcc, intel-darwin, intel-linux, kcc, kylix, mipspro, pathscale, pgi, qcc, sun, sunpro, tru64cxx, vacpp ## And then a lot of comment regarding build.sh which we did not called ourselve so not very usefull Failed to build B2 build engine [alainm@zztop boost_1_77_0]$ Any idea of what went wrong ? Thanks ---- Alain Miniussi DSI, Pôles Calcul et Genie Log. Observatoire de la Côte d'Azur Tél. : +33492003009 (Mont-Gros) +33483618544 (Sophia Antipolis) +33609650665

1 0

Boost Installation in Codeblocks
by Alexander Tanseco 17 Aug '21

17 Aug '21

I successfully built Boost in my directory c:\boost_1_77_0. After which, I set up my global variable boost with base c:\boost_1_77_0. For the built-in field lib, I entered c:\boost_1_77_0\stage\lib (see attached screenshot). For the program's build options, under Linker settings/Link libraries: $(#boost.stage.lib) Under Search Directories/Compiler I entered $(#boost) and under Search Directories/Linker I entered $(#boost.stage.lib). When I built a typical Hello World program (see attached screenshot), I get an error message (see attached screenshot). Where did I go wrong? Somebody, please help.

2 1

[release]Boost 1.77.0
by Marshall Clow 14 Aug '21

14 Aug '21

Release 1.77.0 of the Boost C++ Libraries is now available. These open-source libraries work well with the C++ Standard Library, and are usable across a broad spectrum of applications. The Boost license encourages both commercial and non-commercial use. This release contains two new libraries and numerous enhancements and bug fixes for existing libraries. * Describe: A C++14 reflection library, from Peter Dimov. Provides macros for describing enumerators and struct/class members, and primitives for querying this information. * Lambda2: A C++14, dependency-free, single header lambda library, from Peter Dimov. For details, including download links, see <http://www.boost.org/users/news/version_1.77.0> You can also download directly from: <https://boostorg.jfrog.io/artifactory/main/release/1.77.0/> To install this release on your system, see <http://www.boost.org/doc/libs/release/more/getting_started/index.html> Thanks to everyone who participated in this release. Thanks, -- The Boost Release Team Marshall Clow, Michael Caisse, Glen Fernandes

2 1

Replace icpx with icpc
by Borchert, Christopher B ERDC-RDE-ITL-MS CIV 10 Aug '21

10 Aug '21

I'm trying to build Boost 1.76 from source on SLES 15 using Intel 19.1.3.304 (and GCC 10.2.0 underneath). Simply, how to I get b2 not to use icpx? With icpc it builds but icpx gives ./b2 toolset=intel-linux icpx: error while loading shared libraries: libimf.so: cannot open shared object file: No such file or directory I've tried everything and resorted to altering $PATH to remove icpx, which is how I know that's the problem. Thanks, Chris

3 5

[boost][multiprecision/mpc][math/quadrature] Why boost::multiprecision::exp gets stuck when evaluating complex-valued integral?
by Kirill Kudashkin 06 Aug '21

06 Aug '21

Dear all, I have encountered a problem while using Boost C++ libraries ( multiprecision/mpc.hpp, math/quadrature) which I cannot resolve. This problems appears when running the following piece of code || |#include <boost/math/constants/constants.hpp> #include <boost/math/quadrature/exp_sinh.hpp> #include <boost/multiprecision/mpc.hpp> #include <boost/multiprecision/mpfr.hpp> #include <boost/math/special_functions/fpclassify.hpp> namespace mpns = boost::multiprecision; typedef mpns::number<boost::multiprecision::mpc_complex_backend <50> > mpc_type ; typedef mpc_type::value_type mp_type ; int main() { using boost::math::constants::root_pi ; mpc_type z{mp_type(1),mp_type(1)} ; auto f = [&](mp_type x) { //actually I did not test if all these conditions are needed... if (boost::math::fpclassify<mp_type> (x) == FP_ZERO) { return mpc_type(mp_type(1)) ; }; if (boost::math::fpclassify<mp_type> (x) == FP_INFINITE) { return mpc_type(mp_type(0)) ; }; mp_type x2 = mpns::pow(x,2U) ; if (boost::math::fpclassify<mp_type> (x2) == FP_ZERO) { return mpc_type(mp_type(1)) ; }; if (boost::math::fpclassify<mp_type> (x2) == FP_INFINITE) { return mpc_type(mp_type(0)) ; }; mpc_type ex2 = mpns::exp(-z*x2) ; if (boost::math::fpclassify<mpc_type> (ex2) == FP_ZERO) { return mpc_type(mp_type(0)) ; }; return ex2 ; } ; mp_type termination = boost::math::tools::root_epsilon <mp_type> () ; mp_type error; mp_type L1; size_t max_halvings = 12; boost::math::quadrature::exp_sinh<mp_type> integrator(max_halvings) ; mpc_type res = integrator.integrate(f,termination,&error,&L1) ; mpc_type div = 2U*mpns::sqrt(z) ; mpc_type result = (mpc_type(root_pi<mp_type> ())/div) - res ; return 0; } | ||Namely, the problem appears when evaluating the exponent. The code is compiled without any problems, but hangs on computing | mpc_type ex2 = mpns::exp(-z*x2) ;| It was possible to go into the function while debugging. It crashed eventually, when the compiler reached "unable to resolve non-existing file.../src/init2.c". I let it run for a while checking the call stack. It appears that the programs is stuck atcalling ___gmpn_sub_n_coreisbr(). Or better, in the call stuck it appeared frequently. I did several tests. The real-valued version (with mpfr_backend <2000>) works without any problems. cpp_complex_backend <100> works fine, but I could not test higher precision yet, since it is much slower. I found a work-around which is not ideal. Namely, mpc_type ex2 = mpc_type(mpns::exp(x2)) ; mpc_type ezx2 = mpns::pow(ex2,-z) ; works fine with mpc_complex_backend <2000>. The lambda-function as it is declared in the code above works and yields correct numbers. The problem appears when it is called from the integrator. My system. I use openSUSE Tumbeleweed. I installed boost libraries as provided by |boost_1_76_0-gnu-mpich-hpc-devel.| Complier -> g++, cppStandard gnu+17. I check that I have all libs, e.g. libgmp, libmpc, libmpfr and etc. All headers are present (the programs is complied without any warnings). Thank you in advance and Best wishes, Kirill ||

2 4

[release] Boost 1.77.0 Release Candidate 1
by Marshall Clow 06 Aug '21

06 Aug '21

The first release candidates for the 1.77.0 release are now available at: <https://boostorg.jfrog.io/artifactory/main/release/1.77.0/source/> The SHA256 checksums are as follows: fc9f85fc030e233142908241af7a846e60630aa7388de9a5fafb1f3a26840854 boost_1_77_0_rc1.tar.bz2 f2dd9722b21450daf5ce8ef82c2d99c7ec304f913ee6c6b7f047d5639503dda1 boost_1_77_0_rc1.7z 5347464af5b14ac54bb945dc68f1dd7c56f0dad7262816b956138fc53bcc0131 boost_1_77_0_rc1.tar.gz d2886ceff60c35fc6dc9120e8faa960c1e9535f2d7ce447469eae9836110ea77 boost_1_77_0_rc1.zip As always, the release managers would appreciate it if you download the candidate of your choice and give building it a try. Please report both success and failure, and anything else that is noteworthy. -- The Release managers

4 4

Boost multiprecision as array index
by Claudio La Rosa 03 Aug '21

03 Aug '21

Hi to all, I unsuccessfully tried to use a boost multiprecision variable (object?) as index of an array. Int myArray[100]; uint128_t index = 23; int value = myArray[index]; This code don’t work. There is a way for to use a multiprecision as index of an array? Thanks. Claudio La Rosa

2 3

[Boost.Json] Loss of precision during serialization/parsing plain double ?
by Marc Viala 03 Aug '21

03 Aug '21

Hello Boost users, I'm starting to implement some serialization tools with Boost.Json but I've got some precision issues with double values. To demonstrate this "potential" issue, you will find here after a code snippet to reproduce it: #include <boost/json.hpp> BOOST_AUTO_TEST_CASE(bjson_double_serialize_ut) { namespace bjs = boost::json; static double ref_values[] = { 1217.2772861138403 , -161.68713249779881 , 267.04251495962637}; std::string s; { const bjs::value jv = { ref_values[0] , ref_values[1], ref_values[2], 0.0, 0.0, 0.0}; s = serialize(jv); } double values[3] = {}; { const auto jv{bjs::parse(s)}; values[0] = jv.as_array().at(0).as_double(); values[1] = jv.as_array().at(1).as_double(); values[2] = jv.as_array().at(2).as_double(); } static constexpr auto eps{std::numeric_limits<double>::epsilon()}; BOOST_TEST(std::abs(ref_values[0] - values[0]) <= eps); BOOST_TEST(std::abs(ref_values[1] - values[1]) <= eps); BOOST_TEST(std::abs(ref_values[2] - values[2]) <= eps); } When I compiling and running this code w/ Boost 1.75 and Microsoft Visual C++ version 14.2 (VS 2019) / Windows 10, I've got the following output: "bjson_double_serialize_ut": check std::abs(ref_values[0] - values[0]) <= eps has failed [2.2737367544323206e-13 > 2.2204460492503131e-16] "bjson_double_serialize_ut": check std::abs(ref_values[2] - values[2]) <= eps has failed [5.6843418860808015e-14 > 2.2204460492503131e-16] >From my point of view, I'm expecting to get a tolerance equal or inf. to my "double machine epsilon". Is this correct and how can I address it ? Marc -- Marc Viala

1 0