- Boost - lists.preview.boost.org

[Asio] Propose optimization patch for QNX OS.
by Alex Steshkin 06 Sep '21

06 Sep '21

Hello, By default on QNX OS the select_reactor class is used, which uses the POSIX select() to wait for I/O and write to a pipe to interrupt the wait. Neither select() nor pipes are efficient on QNX OS. We propose to introduce new ionotify_reactor class for QNX OS, which uses ionotify() and pulses instead of select() and pipes. If you are interested in this patch please let me know how to proceed with it: - what repo and branch should I use for this patch? - what is mandatory test set should be done for this patch? - what documentation should be updated to reflect this patch? Thanks, Alex Steshkin ---------------------------------------------------------------------- This transmission (including any attachments) may contain confidential information, privileged material (including material protected by the solicitor-client or other applicable privileges), or constitute non-public information. Any use of this information by anyone other than the intended recipient is prohibited. If you have received this transmission in error, please immediately reply to the sender and delete this information from your system. Use, dissemination, distribution, or reproduction of this transmission by unintended recipients is not authorized and may be unlawful.

2 1

Fwd: Reg, Boost lib dijkstra_shortest_paths for ECMP paths
by kathireswaran thanuskodi 06 Sep '21

06 Sep '21

Hi Team, I came across the boost lib to identify the shortest paths in an network using vertex/edge properties, and have using dijkstra_shortest_paths method with predecessor_map concept with that able to get the shortest path in the graph. Have a requirement like to know about multiple shortest paths also if present, for example below find the below network diagram with 4 edges, A ------- B | | | | D ------- C Edges -> (A,B), (A,D), (B,C), (C,D) with equal distance 1. With the predecessor_map as of now able to get path as A->B->C so that the parent of C is noted as B. In this example we have ECMP(equal cost multiple path) from src A to destination C, tried to search for existing lib to achieve fetching all the paths but unfortunately unable to identify the way. kindly guide how to get such kind of multiple path info with the existing library itself or not. If there exists how to get those info. *<code snip>* const int num_nodes = 4; enum nodes { A, B, C, D}; Edge edge_array[] = { Edge(A, B), Edge(B, C), Edge(A, D), Edge(C,D) }; int weights[] = { 1, 1, 1,1 }; int num_arcs = sizeof(edge_array) / sizeof(Edge); graph_traits<graph_t>::vertex_iterator i, iend; graph_t g(edge_array, edge_array + num_arcs, weights, num_nodes); property_map<graph_t, edge_weight_t>::type weightmap = get(edge_weight, g); // Manually intialize the vertex index and name maps property_map<graph_t, vertex_index_t>::type indexmap = get(vertex_index, g); property_map<graph_t, vertex_name_t>::type name = get(vertex_name, g); int c = 0; for (boost::tie(i, iend) = vertices(g); i != iend; ++i, ++c) { indexmap[*i] = c; name[*i] = 'A' + c; } vertex_descriptor s = vertex(A, g); property_map<graph_t, vertex_distance_t>::type d = get(vertex_distance, g); property_map<graph_t, vertex_predecessor_t>::type p = get(vertex_predecessor, g); dijkstra_shortest_paths(g, s, predecessor_map(p).distance_map(d)); std::cout << "distances and parents:" << std::endl; graph_traits < graph_t >::vertex_iterator vi, vend; for (boost::tie(vi, vend) = vertices(g); vi != vend; ++vi) { std::cout << "distance(" << name[*vi] << ") = " << d[*vi] << ", "; std::cout << "parent(" << name[*vi] << ") = " << name[p[*vi]] << std:: endl; } std::cout << std::endl; *<code snip>* *<output>* distances and parents: distance(A) = 0, parent(A) = A distance(B) = 1, parent(B) = A*distance(C) = 2, parent(C) = B <<<<<<<<<<<<<<<* distance(D) = 1, parent(D) = A Regards, Kathir

1 0

Documentation is not updating since June 10
by Nikita Kniazev 03 Sep '21

03 Sep '21

https://www.boost.org/doc/libs/develop/doc/html/accumulators.html Last revised: June 10, 2021 at 02:32:49 GMT https://www.boost.org/doc/libs/master/doc/html/accumulators.html Last revised: June 10, 2021 at 15:48:57 GMT I thought it was related to the regression matrix not being updated but that was fixed while documentation is still stale.

1 0

Requesting feedback on boost::algorithm::is_clamped
by Ivan Matek 29 Aug '21

29 Aug '21

Hi everyone, I have made a MR for boost::algorithm::is_clamped. Details are on github <https://github.com/boostorg/algorithm/pull/86>. MR is in draft state since doc changes are missing, but tests and code should™ be correct. If you have no time for review of code I would highly appreciate the feedback on the usability of this function. In my work I needed it a few times within last year, but I would like to get feedback from 5-10 other professional developers from diverse backgrounds.

2 3

Re: [boost] Regression tests not updating
by Nikita Kniazev 27 Aug '21

27 Aug '21

> What do you suggest as a solution? I agree that autodetection with no > way to override it is bad. (In my not yet finished CMake guidelines, I > advocate for only using autodetection to determine the default for > the configuration option, while still allowing the user to override it.) > But how does this translate to something actionable? How would you > have written the Nowide build/Jamfile? Sigh, if I had a good solution I would have presented it. An easiest and no-brainer way to tackle this is to have a per-library variable `is-boost-<library>-required/optional` on which library would have to omit checks (do not import Boost.Config, and do not pass checks that return `<build>no` to project/library requirements): ``` if is-boost-<library>-optional { import ../../config/checks/config : requires ; $(do-not-build-if) = [ requires ... ] ; } lib boost_<library> : ... : $(do-not-build-if) ; ``` However, in my opinion, a library must never add `<build>no` to its requirements by itself. Something external should decide if the build/checks failure is fatal or not: ``` rule can-be-built ( ... ) { import ../../config/checks/config : requires ; return [ requires ... ] ; } boost-install boost_<library> : @can-be-built ; ``` It would be perfect if the library could provide a meaningful explanation why it could not be built/describe not satisfied requirements, but that obviously requires a way more sophisticated system. > This is not a Mailman problem, this is a problem with how you are > replying or your mail client setup. Other users with gmail seem to not > have reply problems so I suspect it is not a gmail specific issue. A > quick look reveals that at least one concern is that your headers lack > the "in reply to". I am subscribed to the daily digest, I cannot reply to a message because I do not receive emails even if someone replied to my message or posted something to a thread I interested to be updated more frequently. > As a side-note: there are non-trivial issues with upgrading to mailman > v3; however, hopefully we will be able to upgrade things in the > not-to-far-off future. Thanks for working on that.

1 0

Regression tests not updating
by Joaquin M López Muñoz 27 Aug '21

27 Aug '21

Hi, just a heads up that regression tests are not updating since Aug 12: https://www.boost.org/development/tests/develop/developer/summary.html https://www.boost.org/development/tests/master/developer/summary.html Thank you, Joaquín M López Muñoz

3 3

Re: [boost] Unconditional conditional building is cursed"
by Nikita Kniazev 27 Aug '21

27 Aug '21

> But the problem has nothing to do with silently not building libraries. It's > caused by libraries assuming that Boost.Config is at ../config. Both have the same root cause -- no way to turn off these build checks. If the checks are not needed there is no need to import Boost.Config and assume its location too. P.S. I do not understand how to reply in the Boost mailing lists, please do not blame me if the message went to the wrong thread, I expressed my frustration about that previously, Boost needs to upgrade its Mailman to v3.

3 2

Unconditional conditional building is cursed
by Nikita Kniazev 26 Aug '21

26 Aug '21

Some libraries (Context, Contract, Fiber, Json, Math, Nowide) are built conditionally and use checks from the Config library to do that (`import ../../config/checks/config : requires ;`). Apparently, it happens to be a PITA for some consumers. Vcpkg had to hack around that and update the hack every Boost release, but Boost 1.76 revealed problems in it ( https://github.com/boostorg/build/issues/724) and Boost 1.77 is blocked by it (https://github.com/microsoft/vcpkg/pull/19556#issuecomment-904702798) I have a mixed feeling about silently not building libraries. I understand that for a lot of people it is reasonable that the default stage build will not build a library on an unsupported compiler. However, it is absolutely bad and a confusing thing to a user that the input is silently ignored when a library is requested using the `--with-<library>` flag. A way to turn off the conditional building of the libraries could also be in debug purposes alongside fixing `--with-<library>` behaviour and making possible for vcpkg to live without resorting to hacking the boost build system. I have no affiliation to vcpkg and I am not a maintainer of any mentioned libraries. It is purely my observation about how Boost is used.

2 1

Does boost have an assert that works nicely with constexpr evaluation, and if not should we add one?
by Ivan Matek 25 Aug '21

25 Aug '21

As you may know assert does not play nicely with constexpr evaluation. But std::is_constant_evaluated/if consteval gives us a way to check when the execution happens. https://www.godbolt.org/z/dfder8h16 Does something like this exist in boost, and if not would that be a nice addition? I can not figure out how to support line numbers, filename in a same way assert does, since assert will obviously just take the line info from where it is called. P.S. smart_assert is just a first name I thought of, I am fine with any name as long as it is not some lengthy monstrosity like is_constant_evaluated

4 7

Need cancel() on async_resolve() to callback the resolve handler immediately
by sandeep m.v 25 Aug '21

25 Aug '21

Hi All, I'm looking for a solution for below problem. We're using async_resolve() to perform DNS resolution using boost::asio::ip::tcp::resolver. In few instances probably during network connectivity issues, async_resolve is taking more time. When our application timeouts in 15seconds while waiting for DNS resolution, cancel() is called on resolver. But, that doesn't callback to resolve handle immediately. When I tried to search for this, it looks like, when a cancel() is called on resolver, that doesn't cancel ongoing actual resolution but may cancel if there are any other pending asynchronous action. It's taking 40seconds at times to call resolve handler even when failed to resolve host and even after calling cancel on it. My application waits until resolve handler is called after cancel() to proceed further processing. Is it possible to stop that async_resolve() and get callback to resolve handle immediately after cancel()? If that's impossible, is there a way that we can stop that doesn't callback resolve handler later? Is it possible to close underlying socket related to async_resolve from my application? Thank you in advance. Regards, Sandeep

4 8