August 2016 - Boost - lists.boost.org

All geometry regression tests failing in develop and master
by Barend Gehrels 24 Aug '16

24 Aug '16

Hi, It seems that nearly all regression tests are failing for us in both develop and master: http://www.boost.org/development/tests/master/developer/geometry.html With the messages like: disjoint.obj : error LNK2001: unresolved external symbol "public: virtual void __cdecl boost::unit_test::output::junit_log_formatter::log_start( class std::basic_ostream<char,struct std::char_traits<char> > &,unsigned long)" (?log_start@junit_log_formatter@output@unit_test@boost@@UEAAXAEAV?$basic_ostream@DU?$char_traits@D@std@@@std@@K@Z) We use Boost.Test (header-only), is there something changed there? Do we need to change our Jam file? (of course we want to keep it header-only). The failures are new, before my one-week holiday it was still all OK. Regards, Barend

1 0

[Quickbook] Quickbook, SVG, graphviz dot
by Matt Calabrese 23 Aug '16

23 Aug '16

This is something of a stretch, but I'm hoping someone has a simple solution to my problem. I'm using Quickbook and in my docs I'm including an SVG that I generated from a graphviz dot file, which was also automatically generated from code that I'm documenting. Inside each of the nodes in the graph contained in the SVG, I would like to have source code snippets that are formatted in a way that is consistent with code snippets that are embedded in Quickbook files. Ideally, I'd like to even include Quickbook links in that source code, though I think that's likely a bit harder (my Quickbook code snippet files are usually written in a way that compiles in C++ but on the Quickbook end also replaces known functions with links to their corresponding docs -- I'd like to have that even in the nodes of the generated graph). I'm not certain where precisely to start as I'm rather far outside of my usual domain here. I pretty sure that it must be feasible in some way, as I know that SVG can embed links and use css, etc. but I'm not sure precisely how I should try to tackle this. Perhaps I can create some kind of post-process to run on the SVG that was generated from the dot file? Is there something simpler than this? Any ideas are appreciated. -- -Matt Calabrese

1 0

Re: [boost] double_ended - request formal review
by Mathias Gaunard 23 Aug '16

23 Aug '16

On 15 July 2016 at 16:51, Benedek Thaler <thalerbenedek(a)gmail.com> wrote: > Dear Boost, > > I'd like to request the review of the double_ended library. > > Code: https://github.com/erenon/double_ended > Documentation: http://erenon.hu/double_ended > > The library contains two classes. Excerpt from the documentation: > > * `devector` is a hybrid of the standard vector and deque containers, as > well as the small_vector of Boost.Container. It offers cheap (amortized > constant time) insertion at both the front and back ends, while also > providing the regular features of std::vector, in particular the contiguous > underlying memory. In contrast to the standard vector, however, devector > offers greater control over its internals. Like small_vector, it can store > elements internally without dynamically allocating memory. Not related to your request per-se, but I think this occasion is as good as any to discuss the library. devector is an interesting variant of vector, but it gives the illusion that is could be used as a queue. However given the implementation, it seems elements removed from either end will never be reclaimed. I suggest documenting that limitation or having a threshold above which memory is actually freed.

4 5

[CallableTraits] Seeking Review Manager
by Barrett Adair 23 Aug '16

23 Aug '16

Boost Developers, CallableTraits is a C++11/14/17 library for the inspection, synthesis, and decomposition of callable types. CallableTraits aims to be the "complete type manipulation facility for function types" mentioned in the final paragraph of C++17 proposal p0172. Library: https://github.com/badair/callable_traits Docs: http://badair.github.io/callable_traits/doc/html/index.html Boost Library Incubator: http://blincubator.com/bi_library/callabletraits-2/?gform_post_id=1574 Current Version: 1.0.1 I've received some great feedback on the mailing list. I consider the current version to be production-ready, which includes excellent cross-platform test coverage and thorough documentation filled with examples. Does anyone wish to be a review manager for this library? It is not a very large codebase, although the implementation is config-heavy and relatively noisy. The Library Submissions page on the Boost website indicates that a library manager should be found before requesting a formal review, but I noticed that there are many "scheduled reviews" that do not have a library manager. May I request a review as well? Thank you, Barrett

3 3

[release] Boost 1.62 closed for major changes.
by Rene Rivera 20 Aug '16

20 Aug '16

Reminder Boost 1.62, i.e. the master branches, are now closed for major changes. The only changes that should be going in at this time should be bug fixes for serious problems. Any documentation changes are fine to go in. And remember... Boost 1.62 will close for all changes a week from today. -- -- Rene Rivera -- Grafik - Don't Assume Anything -- Robot Dreams - http://robot-dreams.net -- rrivera/acm.org (msn) - grafikrobot/aim,yahoo,skype,efnet,gmail

4 8

[odeint ] Can fix for ticket #12034 be merged into master branch?
by Aparna Kumta 19 Aug '16

19 Aug '16

The fix for Boost ticket #12034 has been in the develop branch for about 6 months now. Could this please be merged into the master branch so it makes it in for Boost 1.62 release? Thanks, Aparna

1 0

[circular_buffer] Synchronize master with develop
by Andrey Semashev 18 Aug '16

18 Aug '16

Hi, I've created this PR to synchronize master with develop for Boost.CircularBuffer: https://github.com/boostorg/circular_buffer/pull/10 Rene has given the permission to merge in the "[boost] [release] Boost 1.62 closed for major changes." thread. I would appreciate if someone with permissions could merge the PR. Thanks.

1 0

Status on Travis
by Raffi Enficiaud 18 Aug '16

18 Aug '16

Hi dear list, Hi Rene, I would like to know more about Travis and its status. From what I understand, a lot of efforts have been made recently for making things running per commit on Travis and have a quick feedback. * Naturally the first question is: is the status given by Travis the "official" one? so we consider the status given by Travis an official source of information, especially wrt. to its competing regression dashboard. * What are the development plans for Travis? Right now I see 3 types of builds, and I do not fully understand what they cover (platform, compiler) nor their differences. * For boost.test, I am testing each feature branch before merging to develop, by 1. cloning boost superproject on develop 2. switching only boost.test to the feature branch This is for preventing the breakages on develop that we observed in the past since several libraries are depending on boost.test. Would it be possible to have that as well on Travis or any continuous CI? That would be per library then. Would it also be possible to have like a template .travis.yml coming from the boost.superproject which can partly be overiden by a library? (if we want to build things slightly differently) * The CI I am using is able to parse the junit files generated by the test framework, and summarizes the information of each build nicely (also tracks the tests history). Is there anything similar in Travis? Right now, apart from looking at the full logs (which is hard), is there any intermediate information than red/green overall status? * I can see many project under the boostorg of Travis https://travis-ci.org/boostorg/ . How do we do that? Thanks, Raffi

3 4

[Beast] Status: HTTP+WebSocket for Boost
by Vinnie Falco 17 Aug '16

17 Aug '16

I’m reporting on the status of Beast, a C++11 header-only library implementing HTTP and WebSocket which is submitted to the boost incubator. The official repository for Beast is at: https://github.com/vinniefalco/Beast It has been a few months since the last update and the number of users measured on GitHub has grown steadily. There is robust, active development with bug fixes and improvements. Unfortunately there hasn’t been much of in terms of attention from the Boost community, no feedback from the incubator was provided. If anyone would like to help out by checking out Beast in the incubator and leaving some comments that would be a tremendous help. Incubator link: http://blincubator.com/bi_library/beast-2/?gform_post_id=1579 The most exciting news is that Beast HTTP and WebSockets have been deployed in the production server environment for the company I work for, Ripple (http://ripple.com) On a side note, my co-worker David Schwartz is giving a CppCon 2016 Keynote speech on the relationship of C++ to public blockchain systems and cryptocurrencies: http://cppcon.org/schwartz-keynote-turner-plenary/ You can see the Beast integration in “rippled”, an open source server application that implements Ripple’s decentralized ledger: https://github.com/ripple/rippled/tree/develop/src/ripple/server/impl I will be doing a 15 minute lightning talk introducing CppCon 2016 conference attendees to Beast. The talk will have 8 minutes of presentation with the remainder of the time allocated for questions (if there are no questions I have additional materials which can be presented). At the conference I would love to meet anyone who is interested in the library or wants to have their questions answered - feel free to reach out to me through my email address vinnie.falco(a)gmail.com (also listed on my GitHub profile) if you’d like to get in contact. Thanks! -- Follow me on Github: https://github.com/vinniefalco/Beast

2 1

[contract] Contracts for constexpr functions
by Lorenzo Caminiti 16 Aug '16

16 Aug '16

Hello all, A couple of the proposals I read on adding Contract Programming to the C++ standard raise the question if contracts should be supported for constexpr functions or not. I tough about this question a bit... but I am still not sure of the answer. I wanted to discuss this topic with this mailing list, especially with respect to what Boost.Contract can currently do. My general impression is that preconditions, postconditions, and class invariants should useful for constexpr functions but noting that: * Probably old values will not be useful in constexpr functions because constexpr prevents its functions from having side-effects (so it seems there will no variable visible outside of the function that the function body can change, and those variables are usually the candidates for making old copies). * Subcontracting will never apply to constexpr functions because they cannot be virtual. For example: struct conststr { // No virtual function so no subcontracting. [[invariant: empty() == (size() == 0)]] // Useful class invariants. [[invariant: data() != nullptr]] template<unsigned N> constexpr conststr(char const (&s)[N]) [[requires: s != nullptr]] // Useful preconditions. : data_(s), size_(N - 1) {} constexpr char back() const [[ensures(result): result == data()[size() - 1]]] // Useful postconditions. // Old values never useful because constexpr has no side-effects? { return data_[size_ - 1]; } constexpr unsigned size() const { return size_; } constexpr bool empty() const { return size_ == 0; } constexpr char const* data() const { return data_; } private: char const* data_; unsigned size_; }; That said, if I understand it correctly, the following are not allowed from within constexpr functions: 1. (Literal) type local variables with non-trivial constexpr destructors (see N3597 and subsequent proposals). 2. Try-catch statements (see N3597 and subsequent proposals). 3. Lambda functions (see N4487). Reading a few proposals, it seems that supporting the above from constexpr functions would be technically possible, but difficult to implement (see N3597 and its subsequent proposals, N4487, etc.). I'm not really familiar with these constexpr restrictions and plans/discussions to relax them in future revisions of the standard... Boost.Contract implementation uses all the above 1-2-3 respectively to: a. Check class invariants and postconditions at function exit (using RAII). b. Catch assertion failures as well as any other exception that might occur while evaluating an asserted conditions and then call the appropriate contract failure handler. c. Conveniently program functors that check pre- and postconditions within the contracted function. Therefore Boost.Contract implementation does not currently support contracts for constexpr functions. --Lorenzo

2 1

'is_palindrome' and 'const char*'
by Зайцев Александр 14 Aug '16

14 Aug '16

Hello. I have some question about some case for 'is_palindrome' algo. 'is_palindrome' algorithm checks a sequence - if sequence is a palindromeor empty, function returns true, in other cases returns false. But in case e.g. "a" (is_palindrome("a")) result is false, because "a" = {'a', '\0'}. And maybe this behaviour is unexpected for users. So... what do you think about it? Should I implement special version for const char* or not? -- Best regards, Zaitsev Alexander

3 4

filesystem 1.61.0 and cygwin: Windows assumed, posix expected
by Frédéric Bron 12 Aug '16

12 Aug '16

Hi, On cygwin, the following program #include <iostream> #include <boost/filesystem.hpp> int main() { auto p = boost::filesystem::path{"foo"}; p /= "bar"; std::cout << p.string() << '\n'; return 0; } outputs: foo\bar but cygwin is a posix system and I expect: foo/bar I read in the doc that "User-defined BOOST_POSIX_API and BOOST_WINDOWS_API macros are no longer supported." So how can I have posix support on cygwin? Thanks, Frédéric

1 1

[graph] Sync develop with master?
by Murray Cumming 12 Aug '16

12 Aug '16

Now that Boost 1.61 is out, it would be great if the various fixes in boost graph's develop branch could be merged/cherry-picked/whatever to master, please. There seem to be a couple of years of fixes that have not appeared in a release. -- Murray Cumming murrayc(a)murrayc.com www.murrayc.com

5 21

[thread][circular_buffer] Unmerged changes for 1.62
by Andrey Semashev 12 Aug '16

12 Aug '16

Hi, There are a couple of my PRs hanging for quite some time that I'd like to be merged before 1.62 is released: https://github.com/boostorg/circular_buffer/pull/9 Changes the default implementation of circular_buffer in debug mode to the thread-safe by default. https://github.com/boostorg/thread/pull/92 Fixes compilation of Boost.Thread with MSVC-8. There is also https://github.com/boostorg/thread/pull/94, which is not mine but it looks like it fixes a genuine problem. Could someone have a look at them and merge? Thanks.

2 3

[interprocess] weirdest bug with offset_ptr equality comparison
by Joaquin M López Muñoz 11 Aug '16

11 Aug '16

The following is a distilled testcase after the ticket reported at https://svn.boost.org/trac/boost/ticket/12367 #include <boost/interprocess/allocators/allocator.hpp> #include <boost/interprocess/managed_shared_memory.hpp> template<typename Allocator> struct node_base { using allocator_type=Allocator; }; template<typename Allocator> struct node_impl { using value_type= typename Allocator::template rebind<node_impl>::other::value_type; }; template<typename Super> struct trampoline:node_impl<typename Super::allocator_type>{}; template<typename Super> struct node:Super,trampoline<Super>{}; int main() { using node_type= node< node< node_base< boost::interprocess::allocator< int, boost::interprocess::managed_shared_memory::segment_manager > > > >; using pointer=boost::interprocess::offset_ptr<node_type>; pointer p=nullptr; bool b=(p==p); // here's the problem } This works fine with GCC. When compiled with Visual Studio 2015, we get: 1>------ Build started: Project: borrame2015, Configuration: Debug x64 ------ 1> sandbox.cpp 1>d:\usr\joaquin\proyectos\boost-repo\trunk\boost\interprocess\detail\mpl.hpp(74): error C2594: 'argument': ambiguous conversions from 'node<node<node_base<boost::interprocess::allocator<int,boost::interprocess::segment_manager<CharType,MemoryAlgorithm,IndexType>>>>> ' to 'node_impl<boost::interprocess::allocator<int,boost::interprocess::segment_manager<CharType,MemoryAlgorithm,IndexType>>> ' 1> with 1> [ 1> CharType=char, 1> MemoryAlgorithm=boost::interprocess::rbtree_best_fit<boost::interprocess::mutex_family,boost::interprocess::offset_ptr<void,ptrdiff_t,size_t,0x00>,0x00>, 1> IndexType=boost::interprocess::iset_index 1> ] 1> d:\usr\joaquin\proyectos\borrame2015\sandbox.cpp(39): note: see reference to class template instantiation 'boost::interprocess::ipcdetail::is_convertible<node_type *,PointedType *>' being compiled 1> with 1> [ 1> PointedType=node_impl<boost::interprocess::allocator<int,boost::interprocess::segment_manager<char,boost::interprocess::rbtree_best_fit<boost::interprocess::mutex_family,boost::interprocess::offset_ptr<void,ptrdiff_t,size_t,0x00>,0x00>,boost::interprocess::iset_index>>> 1> ] ========== Build: 0 succeeded, 1 failed, 0 up-to-date, 0 skipped ========== Note that it is the line bool b=(p==p); // here's the problem that triggers the problem; if removed, compilation goes fine. Almost anything in the testcase program is needed for the problem to arise: - using node_type=node<node<...>> --> using node_type=node<...> (just one node level): compilation OK - remove using value_type=... in node_impl: compilation OK - rebind<node_mpl> --> rebind<char>: compilation OK - use std::allocator rather than boost::interprocess::allocator: compilation OK Ion, I'm totally puzzled by this and unable to navigate through the internals of Boost.Interprocess. Totally at your disposal if you need further info or assistance running tests etc. Thank you, Joaquín M López Muñoz

2 3

[contract] noexcept and throwing handlers
by Lorenzo Caminiti 10 Aug '16

10 Aug '16

Hello all, I would like to discuss contracts and exception specifications (e.g., noexcept) specifically with respect to what Boost.Contract does. On one hand, it seems silly to allow a contract failure handler to throw an exception to only translate such an exception into a call to terminate() for noexcept functions. However, noexcept can be considered part of the function contract (namely, the contract that says the function shall not throw) and contract failure handlers should not be allowed to break such contract, not even when they are allowed to throw. For example consider a noexcept function fclose() that is called from a destructor without a try-catch statement (correctly so because fclose is declared noexcept). If fclose() is now allowed to throw when its preconditions fail, that will cause the destructor ~x() to throw as well?! void fclose(file& f) noexcept [[requires: f.is_open()]] ... class x { ~x() noexcept { fclose(f_); // f_ might not be open because of some bug... ... } file f_; }; (A point similar to the above is also made in N4160, in contrast with what argued in N4110.) Functions should honor their exception specifications (noexcept, etc.) even when their contracts fail and contract failure handlers are configured to throw. That is what Boost.Contract does (given that contract checking code appears within the function definition so it is subject to the exception specifications that appear in the enclosing function declaration): void fclose(file& f) noexcept { // This will call terminate()... boost::contract::guard c = boost::contract::function() .precondition([&] { BOOST_CONTRACT_ASSERT(f.is_open()); // ...even if this throws. } ; ... } int main() { boost::contract::set_precondition_failure( [] (boost::contract::from where) { // Re-throw exceptions (assertion_failure, user-defined, etc.). if(where != boost::contract::from_destructor) throw; std::terminate(); // But destructors never throw. } ); ... } Also note that a contract framework that allows to install failure handlers that throw exceptions should also pass some sort of parameter to the handler function to indicate if the contract failure happened in a destructor or not. That is so the failure handlers can choose to never throw for destructors, not even if class invariants (or postconditions) fails when checked by destructors. --Lorenzo

3 4

[release] Continuous release archives.
by Rene Rivera 09 Aug '16

09 Aug '16

Starting with this release the archives for the prospective beta and release candidates are continuously built from the master branch and uploaded to SourceForge [1] (and to Bintray [2]). The release team would really appreciate if authors would download the archives and check the contents. In particular that the archives contain all the content you expect of your library, or libraries. And if you can also try building in your system that would stupendous. The more eyes we have on the final release products the smoother the process will be. And result in more time that the master branch is not blocked. Please report any problems sooner rather than later. So that we can try and address them before it's took late. Thank you, The Release Team [1] <https://sourceforge.net/projects/boost/files/boost/snapshots/master/> [2] <https://bintray.com/boostorg/master/snapshot#files> -- -- Rene Rivera -- Grafik - Don't Assume Anything -- Robot Dreams - http://robot-dreams.net -- rrivera/acm.org (msn) - grafikrobot/aim,yahoo,skype,efnet,gmail

5 7

Feature suggestions for Boost's program_options
by Eyal Rozenberg 09 Aug '16

09 Aug '16

Greetings Boost developers, [ Disclaimer: I'm new to this list - I joined by Vladimir Prus' suggestion that I raise my feature ideas here on the list. I've read the posting guidelines, and I haven't seen any discussion of program_options (at least not by Vladimir himself) for at least a year back in the list archives, so I think my post is legit. ] I've been using Boost's program_options library for several months now. It mostly does the job (that's a compliment...) - but there are some features whose lack is really bugging me. For some of them (especially Feature Idea 1) I actually work around by taking some options as unparsed remainders, sticking them in a string->string map and parsing them myself. Without further ado, and in decreasing order of importance to me: -------------------------------- Feature Idea 1: Support specifying multiple alternative long option names, using comma-separation. Right now we can specify, for example, "include-path,I" - one short form, one long form. Why not allow specifying "include-path,include-dir,headers-path,I" ? Only the first long option name would be printed as part of the usage information, to not complicate things overmuch, and merely tolerate alternative phrasing. Feature Idea 2: Have a (new?) parser which supports taking "semi-parsed command-line options" as input: A map from string keys to string values, or to optional<string> values to represent options without an argument as well. This is actually very close to environment parsing, except that the environment parser gets its "map" from, well, the environment variables. So the implementation might be simply exposing the existing functionality with a slightly richer interface. Feature Idea 3: A facility for pretty-printing parsed options to an std::ostream, in which the option description starts the line, and the option value (not the name used for parsing) is printed after it. There is some pretty-printing for the usage info to the console, so why not this? (Argument against this idea: Choosing a default is a very touchy subject. Counter-argument: ... but that's also somewhat true of printing usage information; and nobody has to use this default, it would be entirely optional.) Feature Idea 4: Support three rather than just two levels of describing options: name, for the command line; short description, for listing values (see Feature Idea 3; long description for the usage info. Feature Idea 5: Create a variant of update_with which doesn't just set the option value, but creates a link of sorts between the variable which the app would use and the variable_map entry. With this in place, further changes to the variable map automagically affect the app variable. -------------------------------- So, - What do you think about these suggestions, in principle? - If you find some of them relevant, can you help estimate the necessary development effort and possible negative implications/caveats to consider? - Assuming a feature idea is both relevant and feasible, is it customary to request implementation by the maintainer? Commit yourself to implementing it vis-a-vis statement of statement of partiality to accepting a patch? Some other practice? Thank you for getting through the long(ish) email and for your hard work, Eyal Rozenberg DB architecture group CWI Amsterdam

3 4

[release] Boost 1.62 closing for new libraries this week
by Rene Rivera 07 Aug '16

07 Aug '16

Hopefully everyone has had a good Summer and has gotten all the new libraries and breaking changes in.. As we are closing the release window for that this week. Please consult the calendar [1] for the full release schedule. And take this as a reminder to merge your changes that you want in the next release as soon as possible. [1] <http://www.boost.org/development/index.html> -- -- Rene Rivera -- Grafik - Don't Assume Anything -- Robot Dreams - http://robot-dreams.net -- rrivera/acm.org (msn) - grafikrobot/aim,yahoo,skype,efnet,gmail

1 0

[boost-process] Beta & Review
by Klemens Morgenstern 07 Aug '16

07 Aug '16

Hi, boost-process is now in beta, you can check it out here: https://github.com/klemens-morgenstern/boost-process and read the doc here: http://klemens-morgenstern.github.io/process/ It will probably be in review in the beginning of November, Antony Polukhin will be the review manager. I really hope I'll get some feedback before that, so I can work on the weak points and also get more detailed in the documentation. Thanks, Klemens

4 10

[boost-spirit-qi] Support for n-ary operators PR
by Frank Hein 07 Aug '16

07 Aug '16

Hello, some days ago I raised an PR (https://github.com/boostorg/spirit/pull/200) which adds support for a new class of parsers, in particular named n-ary operators. Joel de Guzman reviewed that. The code is ok. Because this addition touches the inner core of qi, the metagrammar, Joel asked me to drum for support in the community. So if you are in qi, may I kindly ask you to review this PR, the code, the discussion so far, and respond, if you want to support this PR or if you think, it's not needed. Thank you very much. I highly appreciate your feedback. The PR description is very detailled, so this mail has not to be. :) Regards, Frank

1 0

Review Schedule and Review Manager for PODs Flat Reflection (ex magic_get)
by Antony Polukhin 05 Aug '16

05 Aug '16

Hello, PODs Flat Reflection (or PFR) is a C++14 library that represents POD structures as tuples and provides tuple-like methods for PODs. Library has a set of predefined operators for PODs and useful functions for everyday use. Works with user-defined types without any macro or boilerplate code. Library: https://github.com/apolukhin/magic_get Docs: http://apolukhin.github.io/magic_get/ Boost Library Incubator: http://blincubator.com/bi_library/pfr-pod-flat-reflection/?gform_post_id=16… Could the PODs Flat Reflection library be put on review schedule? Anyone wishing to become a Review Manager for the library? -- Best regards, Antony Polukhin

6 9

[config] RFC PR 82
by Domagoj Saric 01 Aug '16

01 Aug '16

Hi everyone, I am finally getting back to (trying to) contributing some of my personal projects to Boost. For two of those, Err (https://github.com/psiha/err) and especially Functonoid (a C++11 generalization and rewrite of my previous Boost.Function related work) I need some lower level codegen and/or optimiser control functionality (i.e. portable macros wrapping toolset specific attributes and pragmas) that I've added to my personal fork of Boost.Config and which I've now submitted in the subject PR (https://github.com/boostorg/config/pull/82) I don't expect this PR to be accepted as is/'just like that' so I'm opening this thread where we can discuss which of those changes/macros are welcome, which need more work and which, for some reason, should not be part of Boost.Config (and which, in turn then, I have to move to some 'internal implementation headers' in libraries that will need them). To avoid 'spamming' (and save time;) I'll skip the explanation of the individual macros as I expect them to mostly be self-explanatory (if not from their name then from the minimal Boost.Config documentation additions that are part of th PR). ps. I'll be on the (off) road for the next three weeks so I don't know when I'll be able to respond until I get back... -- "What Huxley teaches is that in the age of advanced technology, spiritual devastation is more likely to come from an enemy with a smiling face than from one whose countenance exudes suspicion and hate." Neil Postman

8 31

constexpr, cmath functions and other functions and distributions
by Paul A. Bristow 01 Aug '16

01 Aug '16

After much head scratching (and some head banging) - against a string of "Computer Says No" inscrutable diagnostic messages. I believe that I have now produced a (partial) version of Boost.Math normal_distribution that can be a constexpr using GCC6.1.1. (diagnosis has got to improve if constexprizing existing code is to become popular? ) constexpr normal_distribution<double> n01(0., 1.); producing this code 960 .loc 1 47 0 961 000d 660FEFC0 pxor %xmm0, %xmm0 # tmp89 962 0011 F20F1145 movsd %xmm0, -32(%rbp) # tmp89, n01.m_mean 962 E0 963 0016 F20F1005 movsd .LC1(%rip), %xmm0 #, tmp90 963 80080000 964 001e F20F1145 movsd %xmm0, -24(%rbp) # tmp90, n01.m_sd 964 E8 and more interestingly, a PDF can be calculated at compile-time constexpr double p = pdf(n01, 0.5); 965 .loc 1 56 0 966 0023 F20F1005 movsd .LC2(%rip), %xmm0 #, tmp91 966 88080000 967 002b F20F1145 movsd %xmm0, -8(%rbp) # tmp91, p 967 F8 Perhaps the assembler whizzes can confirm that this looks like a compile-time computation? This relies on lots of C++14 and 17 constexpr features, and Most crucially it relies on being able to use several cmath functions as constexpr. For a couple of trivially simple examples: abs(x) and isfinite(x). constexpr double ax42 = abs(-42.); // Unspecified version of abs. constexpr double ax42g = ::abs(-42.); // C double version? constexpr double ax42b = __builtin_abs(-42.); // builtin GCC only constexpr double ax42s = std::abs(-42.); // std:: version Without most cmath functions it seems to me impossible to write any reasonable (so-called special or not) mathy function or statistical distribution that can be constexpr. At present, only GCC 6.1.1 authors have foreseen this need and provided these so-called built_in_* cmath function versions that are constexpr, see https://gcc.gnu.org/onlinedocs/gcc/Other-Builtins.html. (Aside on other compilers: · Clang 3.8 has some problem deep inside type_traits that I haven't yet investigated. · VS2015 update 3 still doesn't compile many essential C++17 features from my experiments so far, and doesn't provide any constexpr cmath functions. ) The many potential benefits of compiler-time computations are quite enticing to contemplate :-) * No run-time where possible * Errors caught at compile-time * Improved code readability * Reduce executable footprint and potentially most important * Avoid global initialization order issues * Avoid initialization thread races See · Scott Schurr, C++ Developer, Ripple Labs https://cppnow2015.sched.org/event/2oLK/constexpr-c-at-compile-time · Dietmar Khul Constant Fun https://www.youtube.com/watch?list=PL9hrFapz4dsM1B9bI8VmEE4JJlR0m-dvo <https://www.youtube.com/watch?list=PL9hrFapz4dsM1B9bI8VmEE4JJlR0m-dvo&v=T37…> &v=T37ElKPICvo&app=desktop Is there any support for making std::abs and friends constexpr where possible in future versions of the C++ Standard? Paul PS I can't conceive of any UDT using or providing floating-point without some function like frexp to get and set exponent and significand. I also note that making constexpr frexp and ldexp and other functions passing a pointer to a variable to be updated may be more difficult/impossible, and is very likely to require at least initialization of the to-be-updated variable before calling frexp. I've investigated overcoming this by writing a naïve new constexpr version of frexp returning a std::pair (significand and exponent) but, of course, it can never be a 'drop-in' replacement because it has a different signature, so lots of code changes are inevitable. --- Paul A. Bristow Prizet Farmhouse Kendal UK LA8 8AB +44 (0) 1539 561830

5 6