May 2010 - Boost - lists.boost.org

[Boost.Thread] Patch for Intel on Darwin
by Belcourt, Kenneth 17 May '10

17 May '10

Hi, The thread library is failing to build with Intel on Darwin with this error. "/opt/intel/Compiler/11.1/067/bin/intel64/icpc" -xc++ -w1 -O0 -g - inline-level=0 -vec-report0 -DBOOST_ALL_NO_LIB=1 - DBOOST_THREAD_BUILD_DLL=1 -DBOOST_THREAD_POSIX -I".." -c -o "/Volumes/ Scratch/kbelco/boost/results/boost/bin.v2/libs/thread/build/intel- darwin-11.1/debug/threading-multi/pthread/thread.o" "../libs/thread/ src/pthread/thread.cpp" ../libs/thread/src/pthread/thread.cpp(385): error: identifier "get_nprocs" is undefined return get_nprocs(); ^ The fix is to test for Apple and Intel before the GNU test. Index: libs/thread/src/pthread/thread.cpp =================================================================== --- libs/thread/src/pthread/thread.cpp (revision 61969) +++ libs/thread/src/pthread/thread.cpp (working copy) @@ -381,6 +381,10 @@ { #if defined(PTW32_VERSION) || defined(__hpux) return pthread_num_processors_np(); +#elif defined(__APPLE__) && defined (__INTEL_COMPILER) + int count; + size_t size=sizeof(count); + return sysctlbyname("hw.ncpu",&count,&size,NULL,0)?0:count; #elif defined(_GNU_SOURCE) return get_nprocs(); #elif defined(__APPLE__) || defined(__FreeBSD__) This has been tested on Darwin and Linux. Okay to commit? -- Noel

2 1

[website] [1.43.0] broken link
by Mostafa 17 May '10

17 May '10

FYI, The following webpage: http://www.boost.org/doc/libs/1_43_0/libs/range/doc/html/range/reference.ht… has a broken link at the "Overview" section, ie, http://www.boost.org/doc/libs/1_43_0/libs/range/doc/html/range/reference/ov… -Mostafa

3 3

Re: [boost] New moderators request unclear? (was: A Remedy for the Review Manager Starvation)
by Thomas Klimpel 17 May '10

17 May '10

Manfred Doudar wrote: > And now, I think a little more about Joachim's mail, and the stated > reception his ideas had at BoostCon (no, I wasn't there - but take > Joachim on his word); and now also recall Beman's email of a while > back, asking for persons to come forward and nominate themselves to > moderate the list - because: Oh, maybe I completely misunderstood Beman's request then. I'm not a native speaker, but I even looked up "nominate" at <http://dict.leo.org>. It seems to mean something like "to elect somebody". (I had looked it up, because it was unclear to me whether "nomination" includes asking the person for permission first.) But as you put it, it seems to mean something like "to volunteer". > "[existing moderators'] day job responsibilities have changed, or > other distractions are preventing them from giving list moderation > the attention it deserves." > > ..this makes me think that there is a call of sorts for "new-blood", > because some boosters are moving on, in some respects (not > necessarily entirely, but other responsibilities keep them away, and > they may not be able to devote as much time as they once did, be it > moderation or otherwise). I wasn't sure here whether Beman was explicitly asking for new blood, or just formulating his request in a way that it was clear that new blood was also allowed. But if somebody would nominate Eric Niebler, Daniel James, Rene Rivera, Hartmut Kaiser, Vladimir Prus, Robert Ramey, Gennadiy Rozental, Joel de Guzman or any other of the usual suspect, I'm pretty sure that Beman wouldn't object. So if I understood Beman's request correctly, he is still waiting for additional nominations. If I understood him incorrectly, and he is just waiting for volunteers, I think he should add a clarification to his request that is also understandable for non-native speakers (that are unaware of the connotations of the word "nomination", if these connotations should really exist). Regards, Thomas

3 2

[test] New location of fraction_tolerance()
by Jeremiah Willcock 17 May '10

17 May '10

There used to be a function named fraction_tolerance() in boost::test_tools (in <boost/test/floating_point_comparison.hpp>) that marked that a floating-point comparison uses a fraction as its tolerance value. That function seems to have disappeared in one of the recent checkins. What is the intended replacement? There is no boost::math::fpc::fraction_tolerance which is what appears from the new naming pattern to be the correct name. Is there an equivalent to that function? -- Jeremiah Willcock

2 1

[contract] static assertions
by Lorenzo Caminiti 16 May '10

16 May '10

Hello all, Shall Boost.Contract static static assertions? 1) N1962 removed static assertions previously added by revisions N1866 and N1773 of the same proposal -- why? 2) In my experience, I never used static assertions within the contracts. However, I am thinking to provide them because they might be useful to better document contract interfaces with assertions that can be efficiently evaluated at compile-time. For example: template<typename T> struct x { CONTRACT_CLASS( (x) ) public: CONTRACT_FUNCTION( (public) (void) (f)( (T)(val) ) (precondition)( (static)(sizeof(T) <= sizeof(long)) // static assertion (checked at compile-time) (val != 0) // run-time assertion ) ({ ... }) }; This code will use one of the BOOST_MPL_ASSERT() to assert the static condition in the precondition function generated by the CONTRACT_FUNCTION() macro. I think this is similar to asserting BOOST_MPL_ASSERT((sizeof(T) <= sizeof(long))) in the function body or at class scope. However, having the static assertion in the contracts has the following advantages: (+) The static assertion compilation and check can be selectively enabled/disabled when preconditions/postcondition/invariants are turned on/off using CONTRACT_CHECK_PRECONDITON/etc. (+) It more clearly documents that the (static) requirement `sizeof(T) <= sizeof(long)` is added by the function f (instead of the class or some other member function). -- Lorenzo

3 4

[fiber] new version
by Oliver Kowalke 16 May '10

16 May '10

I've uploaded a new version of boost.fiber (0.4.2) to boost-vault. The context/stack swapping is implemented in assembler for i386 and x86_64 (see documentation). Because of the limitations of boost.build you have to compile bjam with python support (see docu of boost.fiber). Please note that I plan to refactor boost.fiber into boost.continuation and boost.microthread. Oliver

3 4

[Boost'Con] Thanks for very fruitful week
by joel falcou 16 May '10

16 May '10

Hi all, again this year Boost'con was a success. I wanted to thanks everbody involved for your dedication to make this event such a wonderful place to be. See you next year -- ___________________________________________ Joel Falcou - Assistant Professor PARALL Team - LRI - Universite Paris Sud XI Tel : (+33)1 69 15 66 35

3 2

[type_traits] is_base_of<> GCC error for v1.42.0
by Lorenzo Caminiti 16 May '10

16 May '10

Hello all, The following code gives the GCC compiler error below for Boost 1.42.0 while it compiles just fine for Boost 1.34.1. Is this a bug in Boost.TypeTraits for 1.42.0? // File: b00.cpp #include <boost/type_traits.hpp> #include <iostream> struct y {}; struct x: y { static const bool v = boost::is_base_of<y, x>::value; }; int main() { std::cout << x::v << std::endl; return 0; } $ g++ -Wall -Werror b00.cpp /usr/include/boost/type_traits/is_base_and_derived.hpp: In instantiation of ‘boost::detail::is_base_and_derived_impl2<y, x>’: /usr/include/boost/type_traits/is_base_and_derived.hpp:219: instantiated from ‘const bool boost::detail::is_base_and_derived_impl<y, x>::value’ /usr/include/boost/type_traits/is_base_of.hpp:29: instantiated from ‘const bool boost::detail::is_base_of_imp<y, x>::value’ /usr/include/boost/type_traits/is_base_of.hpp:35: instantiated from ‘boost::is_base_of<y, x>’ b00.cpp:8: instantiated from here /usr/include/boost/type_traits/is_base_and_derived.hpp:145: error: invalid application of ‘sizeof’ to incomplete type ‘x’ /usr/include/boost/type_traits/is_base_and_derived.hpp:145: error: invalid application of ‘sizeof’ to incomplete type ‘boost::STATIC_ASSERTION_FAILURE<false>’ $ g++ --version g++ (GCC) 4.2.4 (Ubuntu 4.2.4-1ubuntu4) Copyright (C) 2007 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. Thank you. -- Lorenzo

3 3

[website] Dates on pages are always the current date
by Jeremiah Willcock 15 May '10

15 May '10

One issue I noticed with the Boost documentation pages is that the rendered versions always have today's date on them, regardless of which Boost version the docs are for. This means that searches for Boost libraries on Google often find pages for old versions in preference to new versions; I think that changing the "Date" header to the date of the correct Boost release would make Google prefer more up-to-date docs. Is this a change that is easy to make? -- Jeremiah Willcock

2 1

[metagraph] mpl_graph and MSM
by Gordon Woodhull 15 May '10

15 May '10

Friends, I'm pleased to announce that I'll soon be submitting mpl_graph for review. This is an implementation of the BGL concepts for compile time metaprogramming. Currently it supports incidence list, adjacency list, depth first search, and breadth first search. It is available in the sandbox under metagraph. I'll write some documentation soon; for now consult BGL's concepts, which are almost the same, just s/()/ typename <>::type/ ;-) It's pretty small - 825 lines of code so far. Standing on the shoulders of MPL... I've also provided a minimal example of running depth first search on the canonical MSM CD Player example. See sandbox/metagraph/libs/ metagraph/example/msm_adaptor.cpp Christophe, thanks for prodding me to get this out the door in your talk here at BoostCon! It sounds like we should write a connected components algorithm to get the features you are looking for, which will be easy with the right dfs visitor. Running DFS does cost in compile time. I haven't yet attempted to profile or optimize the template instantiations, and would appreciate any advice. On this example, it increases the compile time from 3.775 to 5.634 seconds on gcc 4.4. Since the metadata graph structure is lazy (you pay approximately for each graph concept that you exercise), construction of the graph is only 0.2s out of the 1.85s. Interestingly, running BFS in addition to DFS on the same graph adds less than 0.2s -- this makes sense because the big expense should be generating the mpl::maps which support vertex and edge lookup. Here is the test I ran. (I took the median time from each DFS level.) for((dfs=0;dfs<6;++dfs)); do echo; echo DFS=$dfs, 5 runs; for((i=0;i<5;++i)); do time g++ -D DO_DFS=$dfs -I ../../.. -I ~/src/MSMv2.10 -I ~/src/boost- trunk msm_adaptor.cpp -o msm_adaptor; done ; done The good news is that it takes less than 25 lines of code to turn the transition table into a graph and run DFS on it. Thanks to everyone for a truly awesome BoostCon! Gordon P.S. mpl_graph is part of a larger effort to generate patterns of objects from graph metadata. (Metagraph to me means "graph of graphs".) Would people rather I submitted mpl_graph separately, or keep it as a sublibrary of metagraph? (The full metagraph is still months or years away.) P.P.S. Gurtovoy/Abrahams, is the name OK? It's meant to emphasize that it's a simple extension of MPL into graphs.

2 2