- Boost - lists.boost.org

[ASIO] Clarification regarding socket release on Windows
by Gyorgy Szekely 04 Jul '18

04 Jul '18

Hi, The documentation (boost 1.67) of asio::basic_stream_socket::release() has the following remark: *This function is unsupported on Windows versions prior to Windows 8.1, and will fail with boost::asio::error::operation_not_supported on these platforms.* This is in accordance with the compiler warning I get when building my project (from asio basic_socket.hpp), and the runtime behavior: the native socket ownership is indeed retained by asio. For the IOCP backend! If I add the BOOST_ASIO_DISABLE_IOCP define to the build I still get the warning, but the release() appears to be working: the ownership is released and I can do whatever I want with the socket. My question: does the above comment apply only to the IOCP backend and I can use release() with the reactive socket service? Or it just happens to be working in my current setup but it's not guaranteed for others (or it leaks resources, etc...)? Regards, Gyorgy

1 0

[beast] warning
by Vinnie Falco 04 Jul '18

04 Jul '18

I'd like to merge to master, this tiny fix for a warning: <https://github.com/boostorg/beast/pull/1175> The OSX build failure is spurious and unrelated Thanks

1 0

Boost linked example
by degski 03 Jul '18

03 Jul '18

Could someone, please, point me to an example (in the Boost source, i.e. a file in the libs sub-folder) that would require (with certainty) a built library. Which library don't matter. I've tried some date-time examples and boost::random::random_device, but it works, while I hoped it wouldn't. The latter can also mean I have some strays, but then I know that. I would like to check vcpkg static-static linking for Boost. Thanks, degski -- *"If something cannot go on forever, it will stop" - Herbert Stein*

2 4

CMT Library and CI Status for Boost 1.68.0
by James E. King III 01 Jul '18

01 Jul '18

I have been working on adding CI to the CMT libraries (and a few others). This provides improved quality and metrics for each repository with builds for various linux and windows combinations, as well as code coverage and static code analysis with coverity scan. Here is the current status for work that will make it into 1.68.0: The CI framework I have been adding is here: https://github.com/jeking3/boost-ci CI has been enabled on the following repositories (and any identified defects have been resolved) and is passing: function, mpl, pool, property_map, ptr_container, rational, tokenizer In addition, Boost.Move and Boost.Signals2 are also now using this framework to improve deliverable quality. CI not yet enabled on (and reasons): Boost.Assign fails to compile <https://github.com/boostorg/assign/pull/16> with MSVC 2012 in Appveyor. Boost.ConceptCheck fails to compile <https://github.com/boostorg/concept_check/pull/13> on most compilers in Travis CI and Appveyor. (It looks like it's unusable.) Boost.DisjointSets - status unknown (no CI) - open question still unanswered: Can we merge this into container? It is only used by graph and graph_parallel. Boost.DynamicBitset fails to compile <https://github.com/boostorg/dynamic_bitset/pull/23> on all MSVC and on linux with c++03 mode. Boost.Interval (numeric/interval) fails to compile <https://github.com/boostorg/interval/pull/14> - most combinations fail and may go into an infinite loop in some cases. Boost.Logic (tribool) uncovered issue #8 <https://github.com/boostorg/logic/issues/8>. Boost.Signals - we discussed a removal notification in an upcoming release but no action has been taken by the CMT yet to make this happen so it may slip into the next release. Boost.Statechart needs some Jamfile help to get it to work with this CI infrastructure. I am most concerned about Boost.ConceptCheck and Boost.Interval. Both may be unusable at this point. If folks have any spare cycles and care to take a look, it would be appreciated. Thanks, Jim

3 3

[config] MSVC feature-test macros
by Stephan T. Lavavej 29 Jun '18

29 Jun '18

Hi Boost devs, FYI, VS 2017 15.8 Preview 3 has been released, with support for feature-test macros. If you want to take advantage of the feature-test macros, please notify me immediately about any problems so I can get them fixed for the production release. Notes: * __has_cpp_attribute is the one thing that isn't supported yet. * There are minor changes coming for the production release of 15.8, notably L suffixing and an increase in the value for __cpp_deduction_guides. * Library macros are defined by including any <meow> or <cmeow> header (but not <meow.h>). <ciso646> is the lightweight way to get all macros. This will be superseded by <version> in C++20, but that is not yet implemented. * The macros are always provided regardless of the /std compiler option. (Of course their definedness and values vary.) * We are defining __cpp_exceptions and __cpp_rtti, although those weren't picked up for C++20. * The only fundamentally new ability that these macros provide is the ability to sense whether thread-safe statics are supported; this was previously controlled by a compiler option, but not reflected in any MSVC-specific macro. * The macros respect all of our compiler options and library control macros - e.g. if std::byte's definition is suppressed (to avoid conflicts with the Windows SDK), then __cpp_lib_byte reflects that. Hope this helps, STL

6 7

[chrono] #13557 Boost 1.67.0 chrono library Windows build installs NTFS junction
by Stephan T. Lavavej 28 Jun '18

28 Jun '18

Hi Boost.Chrono devs, I have some additional information about https://svn.boost.org/trac10/ticket/13557 . (Also tracked at https://github.com/StephanTLavavej/mingw-distro/issues/58 , it affects both MSVC and MinGW.) As Joshua Morrison discovered, copying libs/chrono/stopwatches/include/boost/chrono/. to boost/chrono before building is a workaround that prevents the junction and hardlink from being created. Also, I mentioned this on Twitter and René Rivera said: "Those links are supposed to appear only for the git checkout of @Boost_Libraries and are created on demand when headers are reference during the header scanning. Creating those for a release is a bug in the release process." (https://twitter.com/grafikrobot/status/1011077139511369728 ) I'm still not sure what the correct fix is, but it seems that boost/chrono should contain the stopwatches files, if they're being referenced by other headers. Thanks, STL

2 2

[ Interest? ] [ out_ptr ] Tiny C++ Abstraction for C-Style Output Pointers
by ThePhD 28 Jun '18

28 Jun '18

Dear Boost Community, I was encouraged to submit this small library that I had created as a formal public investigation of the performance and API-interop benefits to the Boost Mailing List, to perhaps see if other people have come up with something similar to this at their own comapnies. out_ptr is a small abstraction for making types such as std::unique_ptr, std::shared_ptr, boost::shared_ptr, and the upcoming std::retain_ptr <http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2016/p0468r0.html> work work C-Style output functions as a parameter, addressing the need of wanting to be able to do "my_c_api_init( foo, bar, &my_smart_ptr );". It's design is meant to be extensible by both the library implementer and the user, enabling it to be used with other pointer types that do not originate from either `boost::` or `std::`. The library replaces the sought-after syntax of: my_c_api_init( foo, bar, &my_start_ptr ); with a free function that does not impose on the interface of the smart pointer: my_c_api_init( foo, bar, out_ptr( my_smart_ptr ) ); This makes it possible to use a temporary return value to give a reference to a pointer value that then performs a reset on the smart pointer. Is there any want for something like this? I know it has seen use at companies and Studios I have worked in and interacted with, including Microsoft and VMWare. std::retain_ptr's author also has a need for this to make interop easier with reference-counting C APIs and beyond, and I myself implemented something like it 5 years ago and have been using it ever since. Sincerely, JeanHeyd P.S.: this is also going to end up going in the Post-Rapperswil Standards C++ Mailing as well. The proposal is also contained in the repository: https://rawgit.com/ThePhD/out_ptr/master/papers/d1132.html

1 0

[release] 1.68.0 closed for major changes
by Marshall Clow 28 Jun '18

28 Jun '18

The master branch is now closed for new libraries and breaking changes, and will be closing for the beta in one week. Also, if you're ready, you can also add your libraries' changes to the release notes at: <https://github.com/boostorg/website/blob/master/feed/ history/boost_1_68_0.qbk> After editing the release notes, you should get a form to create a pull request, remember to submit it, otherwise I won't see your change. Thanks, -- The Boost Release Team

1 0

[Boost PP] MSVC PP
by degski 28 Jun '18

28 Jun '18

We seem to be getting there. Visual Studio 2017 version 15.8 Preview 3 Announcement <https://blogs.msdn.microsoft.com/visualstudio/2018/06/26/visual-studio-2017…> tl:dr: "*C++ Standards:* A new, experimental, token-based preprocessor that conforms to C++11 standards (including C99 preprocessor features), enabled with /experimental:preprocessor switch. This will be controlled with macro _ *MSVC*_TRADITIONAL, which will be defined to 1 when using the traditional preprocessor and 0 when using the new experimental standards conformant preprocessor." degski -- *"If something cannot go on forever, it will stop" - Herbert Stein*

4 4

[release]
by Zach Laine 27 Jun '18

27 Jun '18

I'm seeing a difference in the Yap Quickbook build in a local checkout of boost-with-submodules, develop branch, and the Circle CI develop build. The auto-index seems to be wrong in the Circle CI one: https://www.boost.org/doc/libs/develop/doc/html/yap.html Note that the stuff at the bottom of that page refers to content in other libs. The Reference section also points to the wrong stuff in the .hpp detail pages. When I originally moved Yap over to Boostorg, I did not initially update all the paths correctly for an integrated Boost build; I was doing a standalone doc build. I have since fixed the paths, and I see the right index and Reference sections when I build locally. I suspect that the Circle CI cache has some files lying around that are causing this problem, but I don't really know. Does anyone have any insight into this problem? Does anyone have admin rights to do a Circle CI rebuild-without-cache? Thanks, Zach

1 0