April 2019 - Boost - lists.boost.org

Re: [boost] [variant] Heads-up!
by Nikita Kniazev 29 Apr '19

29 Apr '19

There seems to be misunderstanding of what the thread is about. The short sum-up of the changes: Nothing changed for users that does not use recursive variant. The changes were applied mainly for recursive_wrapper, and affect recursive variant indirectly. The PR was proposed as a solution for two major recursive variant problems: * Move-only types support * Poor move performance (because of this one Spirit has its own variant) Most of the users are not affected by recursive_wrapper move constructor change. To hit it you have to visit (either directly, or with applying comparison operators) a recursive moved-out variant. > have to agree with Steven. Never empty guarantee is a crucial difference between Boost.Variant and std::variant. std::variant has no recursive variant support. In the common field nothing changed. See the comment above. > If the one negative is adding UB to valid existing code at the next point release, doesn't that outweigh any number of positive outcomes? I do not think there even a single user that visits moved-out variants. There is an assertation for such usage though. > Why wouldn't this be an opt-in strategy - this property of moves has been established for years now. Because it solves the bigger problems and for much more users that it potentially breaks.

2 2

[MP11] Loosing const-qualification on type result with GCC 4.8 only
by Mateusz Loskot 28 Apr '19

28 Apr '19

Hi, (initially posted to https://lists.boost.org/boost-users/2019/04/89778.php) I prepared a lenghty MWE which reproduces my problem with type manipulation using MP11 that for some reason looses const-qualification on a ::type result Here it is at https://godbolt.org/z/eCsakP and copied below. I hope the XXXXXX comments make sense and important static assertions in the main() at the bottom explain what is the expected result. ``` #include <boost/mp11.hpp> #include <cstdint> #include <type_traits> namespace mp11 = boost::mp11; template <typename BitField, int FirstBit, int NumBits, bool IsMutable> struct bit_range_reference { /* long definition, irrelevant */ }; template <typename BitField, typename FirstBit, typename NumBits> struct bit_range_reference_type { using type = bit_range_reference < BitField, FirstBit::value, NumBits::value, true > const; // XXXXXX Why this const gets lost with GCC 4.8 XXXXXX }; template <typename BitField, typename RangeBitSizes> class bit_range_reference_list { template <typename FirstBit, typename NumBits> using reference_type = typename bit_range_reference_type < BitField, FirstBit, NumBits >::type; // XXXXXX Should carry the const, shouldn't? GCC >4.8 do. // If RangeBitSizes is mp11::mp_list_c<int,7,7,2> // then first_bit_list will be mp11::mp_list_c<int,0,7,14,16> using first_bit_list = mp11::mp_fold_q < RangeBitSizes, mp11::mp_list<std::integral_constant<int, 0>>, mp11::mp_bind < mp11::mp_push_back, mp11::_1, mp11::mp_bind < mp11::mp_plus, mp11::mp_bind<mp11::mp_back, mp11::_1>, mp11::_2 > > >; static_assert(mp11::mp_at_c<first_bit_list, 0>::value == 0, "first bit must be 0"); public: using type = mp11::mp_transform < reference_type, mp11::mp_pop_back<first_bit_list>, RangeBitSizes >; }; using reference3_list = typename bit_range_reference_list < std::uint8_t, mp11::mp_list_c<int, 3> >::type; int main() { using r1_t = mp11::mp_at_c<reference3_list, 0>; using r2_t = bit_range_reference<std::uint8_t, 0, 3, true>; using r2c_t= bit_range_reference<std::uint8_t, 0, 3, true> const; //#if defined(__GNUC__) && (__GNUC__ == 4) // GCC 4.8 compiles, later versions of GCC fail to compile static_assert(std::is_same<r1_t, r2_t>::value, "GCC says r1_t is const-qualified"); //#else // GCC 4.9, 5, 6, 7, 8 versions compile, GCC 4.8 fail to compile static_assert(std::is_same<r1_t, r2c_t>::value, "GCC says r1_t is NOT const-qualified"); //#endif } ``` To summary, I expect this to yield true: std::is_same < mp11::mp_at_c<reference3_list, 0> bit_range_reference<std::uint8_t, 0, 3, true> const > I got this confirmed by all GCC versions but GCC 4.8. (clang and MSVC also seem to agree with the GCC >= 4.9) Is mp11::mp_transform loosing const-qualification or have I made a mistake somewhere? I'm using Boost.MP11 from Boost 1.70 and from current develop branch Best regards, -- Mateusz Loskot, http://mateusz.loskot.net

2 3

[endian] Request for comments
by Peter Dimov 28 Apr '19

28 Apr '19

There are two things that I suspect need to be fixed in Boost.Endian, but the changes are breaking. 1. endian_arithmetic<...> derives publicly from endian_buffer<...>. Intuitively, this inheritance relationship doesn't seem right, and it prevents endian_arithmetic from being a C++03 POD, due to having a base class. 2. the data() member function returns char const*. This type is strongly associated with null-terminated strings, and as such, using it for the underlying bytes of the endian object seems error-prone. I'd much prefer to return `unsigned char const*` from data() (although in C++17 `std::byte const*` might be even better.) I can't however estimate with any certainty whether the (mostly aesthetic) benefit of these changes would outweigh the users' annoyance from their code being broken for no strong reason. Comments?

4 3

CMake, modular Boost, and other stories
by Mike 27 Apr '19

27 Apr '19

Coming at this from a different angle: It might make sense to reserve find_package(Boost_foo) for libraries that directly support installation from cmake. Not sure if such a split is a good idea, but I wanted to put it up for consideration. Best Mike

3 3

PartIcpation in Boost
by Edward Diener 27 Apr '19

27 Apr '19

For personal reasons I will not be able to participate in Boost as I have in the past, which means I will no longer be able to be part of CMT. I will continue to maintain the libraries I created for Boost, tti and VMD, as well as maintaining the preprocessor library created by Paul Mensonides. This means that I will not have the time or the will to be involved with other libraries any more, except on the most general level of discussion. For what it is worth I think that Cromwell Enage should become the maintainer of the Parameter library. I have not, and still do not understand, why when talented C++ programmers are willing to contribute heavily to Boost they are not encouraged and supported more. But this is not my problem. For those who are interested, the experimental C++ standard preprocessor which had been added for VC++ in VS2017 via the '/experimental:preprocessor' switch is still supported in VS2019, although not documented anywhere in the official VS2019 docs. Ths VC++ experimental preprocessor is fully supported in Boost PP and Boost VMD. Unfortunately this preprocessor can not pass all the Boost PP tests, although it has gotten a little better in VS2019 from VS2017. I posted bug reports in Microsoft's bug reporting system regarding its bugs. I have not bothered to test it with the VMD tests, since if it can not pass all the PP tests it is not going to pass the VMD tests. Also while clang on Windows, whether targeting gcc or VC++, passes all the Boost PP tests it fails to pass the VMD tests with lots of errors, seemingly because it is running out of memory ( ? ). I have not bothered reporting this to clang, since my previous experience has shown me that they are clearly not that interested in their Windows implementation. Both gcc on both Linux and Windows, and clang on Linux, pass all the Boost PP and VMD tests. Of course you can try all this out for yourself if you are interested. Edward Diener

3 2

Re: [boost] CMake, modular Boost, and other stories
by Mike 27 Apr '19

27 Apr '19

> +1. Aside from the other benefits mentioned: What benefits exactly? I don't think anyone has mentioned any benefits of dropping the second form at all at all. This isn't an "either or" question - currently both versions are supported. So why exactly should one be dropped? We have talked a lot about Modularization, but little about the original question. Best Mike

1 0

CMake, modular Boost, and other stories
by Peter Dimov 27 Apr '19

27 Apr '19

It's been quiet lately, so it must be time for another CMake discussion. As you know, Boost 1.70 installs CMake configuration files on `b2 install`, and we've started to receive feedback, some of which more interesting than what doesn't work. At present, `b2 install --with-filesystem` for example installs a configuration file boost_filesystem-config.cmake, which enables CMake users to issue the command `find_package(boost_filesystem)` and then use Boost.Filesystem. This mimics what an ordinary, independent library that happened to be called "boost_filesystem" would do. It also installs an umbrella BoostConfig.cmake file, which enables the alternative use of `find_package(Boost COMPONENTS filesystem)`, which treats Filesystem as a component of the package Boost, rather than as a standalone library. This is compatible with the existing, pre-1.70 use of FindBoost.cmake to locate Boost from CMake. Brad King (from Kitware, who should know) argues that installing separate configuration files for all the libraries is wrong. He says that the preferred practice is to only support the second use, with Boost as the package, and the libraries as components, not accessible as packages on their own. (The discussion is here: https://gitlab.kitware.com/cmake/cmake/issues/19186) The reason I went with installing separate configuration files was that I expected that (a) this would be our preferred future approach when, or if, we switch to building with CMake and (b) even if not, this seemed like a step in the right, "modular", direction where libraries are treated as independent rather than parts of the monolithic package Boost. I'm having second thoughts though. It looks to me that Brad is right, and that the various library configuration files should be implementation details of a single Boost package (in the CMake sense) and not be user-visible. That is, only support find_package(Boost COMPONENTS filesystem) and drop find_package(boost_filesystem) Thoughts?

7 11

[build] b2 help moving disjoint_sets
by James E. King III 26 Apr '19

26 Apr '19

I have merged disjoint_sets into boostorg/graph locally and submitted a pull request for it (https://github.com/boostorg/graph/pull/169) The build jobs are failing however, due to disjoint_sets already existing and already having links: https://ci.appveyor.com/project/jzmaddock/graph/builds/24131163/job/uguk1li… b2 headers Performing configuration checks - default address-model : 32-bit - default architecture : x86 - symlinks supported : yes error: Cannot create link boost\pending\detail\disjoint_sets.hpp to libs\graph\include\boost\pending\detail\disjoint_sets.hpp. error: Link previously defined to another file, libs\disjoint_sets\include\boost\pending\detail\disjoint_sets.hpp. Command exited with code 1 This happened for me locally when I did the work, so I removed it from the superproject locally, and deleted disjoint_sets locally. Any thoughts on how I can tell the build to ignore disjoint_sets so it can install the headers from the new location and complete? This is a temporary situation until the superproject is updated, but the superproject cannot be updated until the code is merged into graph, and the builds are all failing so I cannot prove to the graph maintainers that the pull request is viable. The only way I can see around this is to rename the disjoint_sets headers from pending and into graph, and then fix up graph and graph_parallel, but my intention was to just merge the header into graph, not to change the location of delivered headers. That said, getting something out of pending is always a good thing. Any other ideas? Thanks, Jim

2 1

Making two Boost libraries interoperate
by Andrzej Krzemienski 26 Apr '19

26 Apr '19

Hi Everyone, Somewhat in connection with the discussion about modularizing Boost libraries, we had a certain conceptual problem recently about providing a customization code that allows Boost.Serialization library to serialize Boost.Optional. Such customization requires some code to be written (template specializations provided). Conceptually, such customization code does not belong to Boost.Serialization: we do not expect that Boost.Serialization provides specializations for every library that someone might ever want to serialize. It would also require the maintainer of Serialization to keep track of the implementation of Optional. However, it also does not belong to Boost.Optional: we do not expect that Boost.Optional provides customization points for every framework out there that the users may want to ever use. It would also require the maintainer of Optional to keep track of the implementation of Serialization. However, it is a natural expectation of a Boost user that if the library offers two components they should just smoothly inter-operate with one another. We could provide a small module Serialization-Optional-interop, but that would require many such components to exist and be maintained by someone. I would like to obtain some guidance on how to address such issue. The same problem must already be present between other Boost libraries. Maybe this has been already brought up, and we have a policy for solving this problem somewhere? Regards, &rzej;

10 23

Google Summer of Code slots
by David Bellot 25 Apr '19

25 Apr '19

Dear Boost community, Google has allocated 11 slots for our summer students this year. It's a great news. It means that 11 brilliant students will have the opportunity to work full-time with us and contribute to our libraries. The selected students will be announced soon. I'd like to thank all the mentors for their hard work in preparing the proposals and helping the students to candidate. And of course, a big thank you to Google for supporting open source every year. Best regards, David Bellot

1 0