July 2014 - Boost - lists.preview.boost.org

Re: [boost] Libraries failing across the board.
by Matthew Musto 11 Jul '14

11 Jul '14

David, Thank you, I understand. I guess I will have to wait for 1.57 or try to merge in the development branch to my local 1.56 version as soon as it is working. I realize we do not hold releases until all tests pass but I wonder if we should do something like that for one release a year as a long term support (LTS) version. It just makes relying upon boost very difficult from a end user perspective when major development projects often times rely upon multiple libraries. Would this be worth having a larger discussion about in its own thread? Thanks, -Matt From: David Bellot <david.bellot <at> gmail.com> Subject: Re: Libraries failing across the board. <http://news.gmane.org/find-root.php?message_id=CAOE6ZJGDf%5fwxRhp50KQyEbKE1…> Newsgroups: gmane.comp.lib.boost.devel <http://news.gmane.org/gmane.comp.lib.boost.devel> Date: 2014-07-11 11:36:24 GMT (2 hours and 7 minutes ago) yes, I know it's not ideal but we haven't time to do better this time [image: :-(] I will try to update the devel branch on git if it helps On Thu, Jul 10, 2014 at 12:22 AM, Matthew Musto <matthew.musto <at> gmail.com> wrote: > David, > > Just so I am clear, you are saying that 1.56 cannot be relied upon by users > who need ublas as there is no intention of making it work? > > If that is true, this release is effectively worthless to me or anyone else > who needs any other library in conjunction with ublas? For example, I rely > heavily upon threads, serialization, asio and numeric/ublas. 1.55 did not > compile on VS2013 and there was much hope for this release which now seems > dashed. > > -Matt > > > >Date: Tue, 8 Jul 2014 23:24:15 +0100 > >From: David Bellot <david.bellot <at> gmail.com> > >To: "boost <at> lists.boost.org" <boost <at> lists.boost.org> > >Subject: Re: [boost] Libraries failing across the board. > >Message-ID: > > <CAOE6ZJGEhHEcujgXw4N2r= > >DLsn3vhx9+qSc3eN=jdfPHNSUcJA <at> mail.gmail.com> > >Content-Type: text/plain; charset=UTF-8 > > > >for ublas we decided to skip this release and concentrate on 1.57 as we > >have more patches coming. > >- D > > > > > >On Tue, Jul 8, 2014 at 10:07 PM, Beman Dawes <bdawes <at> acm.org> wrote: > > > >> Marshall is just a few days away from being able to put out a beta > release > >> candidate. (And that is very good news!) > >> > >> But we still have too many libraries failing across the board on master: > >> > >> accumulators > >> interprocess > >> numeric/ublas > >> pool > >> proto > >> spirit/repository > >> spirit/test > >> tr1 > >> wave > >> > >> See > http://www.boost.org/development/tests/master/developer/summary.html > >> > >> What is the hold up with these? Does anyone need help? > >> > >> --Beman > > _______________________________________________ > Unsubscribe & other changes: > http://lists.boost.org/mailman/listinfo.cgi/boost >

1 0

Re: [boost] Interest in a new generic matrix library with specific 3D operations?
by Fabien Picarougne 11 Jul '14

11 Jul '14

Hello, I wanted to see my idea through to the end and I have updated my library to make the data container generic (https://bitbucket.org/fpicarougne/template-matrix-library) I got a little inspired from QVM but with a different implementation as you can see in the code. Now, the matrix library is independent from the storage, you can use predefined storages (see CMat.h for dense storage with SDenseStorage or SProxyArrayStorage in order to use a C array given by the user) but also built your own by derived (CRTP) from SStorage. For example, getting a column or a submatrix of a matrix is very easy. This make all the matrix operations independent from the storage type, and all the methods are usable if you have predefined basic operators (+, * ...). I admit that you can do quite the same thing with QVM, but I my opinion, factorization of methods is very easy with the architecture of my library because of inheritance. See test.cpp for a usage example. What do you think about this? Is it of interest to you? Fabien

1 0

Libraries failing across the board.
by Matthew Musto 11 Jul '14

11 Jul '14

David, Just so I am clear, you are saying that 1.56 cannot be relied upon by users who need ublas as there is no intention of making it work? If that is true, this release is effectively worthless to me or anyone else who needs any other library in conjunction with ublas? For example, I rely heavily upon threads, serialization, asio and numeric/ublas. 1.55 did not compile on VS2013 and there was much hope for this release which now seems dashed. -Matt >Date: Tue, 8 Jul 2014 23:24:15 +0100 >From: David Bellot <david.bellot(a)gmail.com> >To: "boost(a)lists.boost.org" <boost(a)lists.boost.org> >Subject: Re: [boost] Libraries failing across the board. >Message-ID: > <CAOE6ZJGEhHEcujgXw4N2r= >DLsn3vhx9+qSc3eN=jdfPHNSUcJA(a)mail.gmail.com> >Content-Type: text/plain; charset=UTF-8 > >for ublas we decided to skip this release and concentrate on 1.57 as we >have more patches coming. >- D > > >On Tue, Jul 8, 2014 at 10:07 PM, Beman Dawes <bdawes(a)acm.org> wrote: > >> Marshall is just a few days away from being able to put out a beta release >> candidate. (And that is very good news!) >> >> But we still have too many libraries failing across the board on master: >> >> accumulators >> interprocess >> numeric/ublas >> pool >> proto >> spirit/repository >> spirit/test >> tr1 >> wave >> >> See http://www.boost.org/development/tests/master/developer/summary.html >> >> What is the hold up with these? Does anyone need help? >> >> --Beman

2 1

Missing symbols when Compiling Boost on AIX 6.1
by Axel Ismirlian 10 Jul '14

10 Jul '14

Hello everyone, One of the errors that keeps popping up time and time again as I attempt to compile boost on AIX is the following: common.mkdir ../bin.v2/libs/asio/test/basic_datagram_socket_select.test common.mkdir ../bin.v2/libs/asio/test/basic_datagram_socket_select.test/gcc-4.8.1 common.mkdir ../bin.v2/libs/asio/test/basic_datagram_socket_select.test/gcc-4.8.1/debug common.mkdir ../bin.v2/libs/asio/test/basic_datagram_socket_select.test/gcc-4.8.1/debug/threading-multi gcc.compile.c+ + ../bin.v2/libs/asio/test/basic_datagram_socket_select.test/gcc-4.8.1/debug/threading-multi/basic_datagram_socket.o gcc.link ../bin.v2/libs/asio/test/basic_datagram_socket_select.test/gcc-4.8.1/debug/threading-multi/basic_datagram_socket_select ld: 0711-224 WARNING: Duplicate symbol: .init_unit_test_suite(int, char**) ld: 0711-345 Use the -bloadmap or -bnoquiet option to obtain more information. ld: 0711-317 ERROR: Undefined symbol: .test_main(int, char**) collect2: error: ld returned 8 exit status "g++" -o "../bin.v2/libs/asio/test/basic_datagram_socket_select.test/gcc-4.8.1/debug/threading-multi/basic_datagram_socket_select" "../bin.v2/libs/asio/test/basic_datagram_socket_select.test/gcc-4.8.1/debug/threading-multi/basic_datagram_socket.o" "../bin.v2/libs/test/build/gcc-4.8.1/debug/link-static/threading-multi/libboost_test_exec_monitor.a" "../bin.v2/libs/system/build/gcc-4.8.1/debug/threading-multi/libboost_system.so" "../bin.v2/libs/regex/build/gcc-4.8.1/debug/threading-multi/libboost_regex.so" "../bin.v2/libs/date_time/build/gcc-4.8.1/debug/threading-multi/libboost_date_time.so" -lrt -g -Wl,-brtl -Wl,-bnoipath -pthread ...failed gcc.link ../bin.v2/libs/asio/test/basic_datagram_socket_select.test/gcc-4.8.1/debug/threading-multi/basic_datagram_socket_select... So this particular error happens in every single asio test, but this kind of error also happens across a few other libraries. I tried to set the gcc flag that simply states that these type of linking errors should be ignored, but many tests apparently still reference the symbol on AIX. Additionally while debugging, I realized that the power linux platform also keeps the symbols test_main undefined, but apparently manages to throw them out in the executable. My main question is where in the building framework are those symbols being thrown out? Or is the issue something bigger? Thank you for reading. Sincerely, -Axel

1 0

GSoC: Boost.Pipeline -- scheduling of segments
by Benedek Thaler 10 Jul '14

10 Jul '14

Hi I'm working on the GSoC Pipeline project, which is based on the N3534 [1] proposal. Work in progress can be found on GitHub [2]. A simple example of using the pipeline: pipeline::from(input_container) | transformation1 | t2 | t3 | output_container; When running this pipeline, items will be read from `input_container`, processed by the transformations and written to the `output_container`. Each segment should be applied in a parallel manner, this is the point of the pipeline. However, scheduling of these works is not trivial. Quoting the proposal: "The current pipeline framework uses one thread for each stage of the pipeline. To limit the use of resources, it should be possible to run with fewer threads, using work-stealing or work-sharing techniques." We are wondering if we could improve on this. Lets assume a thread pool of a single thread and the example pipeline above. On run(), the ideal would be that the thread reads *some* items from `input_container`, applies the transformations on them and pushes the results to `output_container`. That is: spending some time on each transformation then yield and pick up the next one. However, doing this would imply the transformations are reentrant. This additional constraint must be considered carefully. This behavior is implemented in the `development` branch [2]. Aside the two solutions above, Vicente J. Botet Escriba coined in the following idea: Let the threads work on a single transformation until the queue gets closed (no more input), then move to the next one. This is easy to implement and scales to as many threads as many segments are present. On the other hand, it kills the performance of the "online usecase": pipeline::from(read_message_from_socket) | process_message | send_message; It's not deterministic when the first segment will end, the pipeline will hang and won't provide any output. Also, in a slightly less strict scenario, when the end of input is specified, the latency could be just too high. To summarize, the following options are on the table: 1. Dedicate a thread to each segment (what to do with a fixed size threadpool?) 2. Constrain the transformations to be reentrant. 3. Run each transformations until there is input to be processed, from beginning to the end. We are kindly asking the recipients to share their ideas or opinions, ideally with a matching usecase. Thanks, Benedek [1]: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3534.html [2]: https://github.com/erenon/pipeline/tree/development

8 35

[iterator] Move iterators to a separate namespace
by Andrey Semashev 10 Jul '14

10 Jul '14

Hi, I've created this pull request a few days ago: https://github.com/boostorg/iterator/pull/5 It basically moves all iterators to boost::iterators namespace (while still importing the symbols to boost namespace for backward compatibility). This change is also intended to fix the counting_iterator_test test on older compilers, which was broken by my previous pull request (there appeared a name clash with Boost.TypeTraits). Could it be merged?

1 1

permission to merge odeint to trunk
by Karsten Ahnert 10 Jul '14

10 Jul '14

Can I still merge the develop branch of odeint into the master? The develop branch compiles fine and the tests are running for some time now. I also guess that no other boost libraries depend on odeint, so it should be relatively safe to merge it. Btw. there were no announcements that the master branch closes for the release, so I thought there is still some time to merge develop into master. But it seems that the release of Boost 1.56 has gained momentum and I don't want to silently merge into master.

2 2

[mpl] Can I merge pull requests for 'develop' into 'master' for Boost 1.56
by Edward Diener 09 Jul '14

09 Jul '14

Eric Niebler has issued a couple of pull requests for MPL on 'develop' involving VC++ and the has_xxx functionality which I have merged to 'develop' after testing them locally and successfully with various versions of VC++ from VC8 to VC12. The earlier already cycled through regression tests successfully but the latest was merged just recently. The latest involves a change for the NVidia CUDA compiler, for which I see no regression testing. I have not tested this change with the CUDA compiler ( is there even a Boost Build toolset for this compiler ? ) but I implictly trust Eric's judgment. Can/should these changes be merged to 'master' for 1.56 ?

3 7

Best home for cxx11_char_types.hpp?
by Beman Dawes 09 Jul '14

09 Jul '14

I'm working on an update of Boost.Filesystem that adds support for the C++11 char16_t, char23_t, u16string, and u32string types. For compilers that don't yet support these types, a header cxx11_char_types.hpp (see attached) provides emulation that is good enough for filesystem, and is probably useful for other libraries and for user code too. We discussed this header several years ago, and I've been using it since then. Where should this header live? It can certainly go in libs/filesystem/include/boost/filesystem, but that doesn't really expose it to general Boost users. Is there a better place, but without introducing unwanted dependencies? --Beman

2 2

[type_erasure] Force any to have typedef
by Benedek Thaler 09 Jul '14

09 Jul '14

Hi, Assume I have the following function template: template <typename Container> typename Container::value_type get_front(const Container& c) { return c.front(); } Is there a way I can force type_erasure::any<> to have a `value_type` typedef? I can create a concept which requires the underlying type to have it, but I'd like to make `any` to have such a typedef. Thanks, Benedek

2 2