November 2010 - Boost - lists.boost.org

[type traits] is_virtual_base_of causes C4250 (class inherits member via dominance) with VS 2005
by Martin B. 06 Nov '10

06 Nov '10

Hi! I didn't find any references to this, so I thought I'd let you know that boost::is_virtual_base_of<Base, Derived> will cause (harmless) C4250 when evaluating to true. (VC++ 2005) I hit this when using the BOOST_CLASS_EXPORT macro from Boost.Serialization, as this uses is_virtual_base_of internally. cheers, Martin ps: For code see: http://stackoverflow.com/q/4091166/321013

3 2

[icl/itl] Interval Container Library ready for inclusion
by Joachim Faulhaber 06 Nov '10

06 Nov '10

Dear list! After the review of the Interval Template Library in February and its acceptance for inclusion, I worked on a number of suggestions and proposals from the reviewers, in order to improve the design and quality of the library. The library is now renamed to Boost.Icl: Interval Container Library. I have completed a release icl-4.0.0, where those suggestions are implemented. Apart from typos and flaws in the docs that might be there after renaming, this will be the version for inclusion into the next boost release (1.46.0). The now renamed library has a new path in the sandbox: https://svn.boost.org/svn/boost/sandbox/icl/ The sources are also in the vault http://www.boostpro.com/vault/index.php?&direction=0&order=&directory=Conta… and on sourceforge (extended version) http://sourceforge.net/projects/itl/ The BoostBook docs are online available here: http://www.joachim-faulhaber.de/boost_icl/doc/libs/icl/doc/html/index.html Many of the proposals from the reviewers were agreed upon during the discussion on the library. In particular I took into account suggestions by Jeff Flinn (JF), Luke Simonson (LS), Barend Gehrels (BG), Vicente Botet (VB), Thomas Klimpel (TK), Paul A. Bristow (PB), Robert Stewart (RS), John Reid (JR) and Markus Werle (MW). Some suggestions by Phil Endecott, Jerry Jeremiah, Brian Wood, Barend Gehrels, Luke Simonson and John Reid are not or not yet implemented for different reasons but they may be considered for the future evolution of the library. * The library has been renamed from Interval Template Library (ITL) to Interval Container Library (ICL) Its namespace is now icl instead of itl. (Renaming was suggested by MW). * Introduction of dynamically bounded and statically bounded intervals (JF,JR,VB) + The 'universal' interval class itl::interval has been replaced by + dynamically bounded discrete_interval and continuous_interval http://www.herold-faulhaber.de/boost_icl/doc/libs/icl/doc/html/boost_icl/ex… + statically bounded right_open_interval, left_open_interval closed_interval and open_interval and a uniform set of namespace global functions or interval_concept for them. Statically bounded intervals don't carry information about their interval bounds and use less memory. (JF,JR) http://www.herold-faulhaber.de/boost_icl/doc/libs/icl/doc/html/boost_icl/ex… + The interval class templates are now quite small. The functions are namespace global functions that make up an interval concept (JF,LS) + Added functions 'distance' and 'inner_complement' for intervals.(JR) http://www.herold-faulhaber.de/boost_icl/doc/libs/icl/doc/html/boost_icl/fu… * User defined interval types are now customizable (JF) + An interval_traits and dynamic_interval_traits template describes the generic properties of the interval concept ... + and allows to customize the icl library for user defined interval types. http://www.herold-faulhaber.de/boost_icl/doc/libs/icl/doc/html/boost_icl/ex… * Extraction of namespace global functions. (LS,JF,RS) + Member functions have been extracted form class templates as long as they are not needed to maintain type invariants. + The sets of namespace global functions build concepts icl::element_set, icl::element_map, icl::interval_set, icl::interval_map + Some functions have been renamed to conform existing standards (ISO/OGC). (BG) + std::set and other stl conformant set types are models of the icl's element_set concept. Class template icl::set has been removed. + Member functions for element iterations are extracted from the interval container class templates. (LS) * Names, referred to as 'colorful terminology' have been renamed (TK,VB,PB) + neutron -> identity_element + unon -> unit_element + ... and also their occurrence in composite identifiers If you find bugs in the library or typos or other flaws in the docs, please let me know so I can fix them before inclusion of the library into the next boost release. Enjoy, Joachim

5 4

bjam in lib/<x>/test is equivalent to run.py
by Jim Bell 06 Nov '10

06 Nov '10

Compelled as I am to audit everything, I had to compare running bjam in the lib/<x>/test directory with run.py. I'm confident that they do produce (essentially) the same compile, link, and execute commands. Notes, from a regression-runner's POV: * They use different directories: run.py uses {boost-root}/../results/bin.v2/..., and bjam uses {boost-root}/bin.v2/... * run.py sets up bjam with a lot of flags, like --debug-configuration [shouldn't matter for most people] -d2 [important for many, described below] -lxxx [that's lower-case L, with a time-out in seconds: potentially important] preserve-test-targets=off [not sure if this matters or not] * run.py runs from the boost/status directory, I think. Might matter for some tests. Not sure about straight bjam. To make your own script to rebuild/rerun a particular test, run bjam with -d+2 (i.e., -d2) and direct the output to a file. Then copy the compile & link commands out of that file. The -lxxx command (e.g., -l120) matters if a test could hang (e.g., <https://svn.boost.org/trac/boost/ticket/4766>).

3 5

Re: [boost] [Boost-users] [icl/itl] Interval Container Library ready for inclusion
by Joachim Faulhaber 05 Nov '10

05 Nov '10

2010/11/5 Jeff Flinn <TriumphSprint2000(a)hotmail.com>: > Hartmut Kaiser wrote: >>> >>> After the review of the Interval Template Library in February and its >>> acceptance for inclusion, I worked on a number of suggestions and >>> proposals from the reviewers, in order to improve the design and quality >>> of the library. The library is now renamed to Boost.Icl: Interval >>> Container Library. >>> >>> I have completed a release icl-4.0.0, where those suggestions are >>> implemented. Apart from typos and flaws in the docs that might be there >>> after renaming, this will be the version for inclusion into the next >>> boost >>> release (1.46.0). >>> >>> The now renamed library has a new path in the sandbox: >>> https://svn.boost.org/svn/boost/sandbox/icl/ >>> >>> The sources are also in the vault >>> >>> http://www.boostpro.com/vault/index.php?&direction=0&order=&directory=Cont >>> ainers >>> >>> and on sourceforge (extended version) >>> http://sourceforge.net/projects/itl/ >>> >>> The BoostBook docs are online available here: >>> >>> http://www.joachim-faulhaber.de/boost_icl/doc/libs/icl/doc/html/index.html >> >> Joachim, that's great news. Many people are waiting for this! > > Awesome! I'm looking forward to working with the final result. > >> Why don't you just go ahead and move everything into Boost trunk. That >> does >> not disrupt any release procedures, but gives you the benefit of running >> the >> tests on a daily basis on a broad range of platforms. Additionally it >> simplifies things for people working from Boost SVN trunk wanting to use >> your library. > > Yep, that would be great. Thank you all for your welcoming remarks on the new version of the Interval Container Library. Yes, I am prepared to move this version of the ICL into the Boost trunk, but I don't have write permission yet. How do I get it? Are there any special policies to follow when working with the Boost trunk? Best regards, Joachim

2 1

[Boost.config] User.hpp for new PGI 10.9 compiler
by K. Noel Belcourt 05 Nov '10

05 Nov '10

Hi, Can a Boost.config developer take a look at the attached user.hpp generated by the configure script for the pgi-10.9 toolset? Are there any obvious changes necessary to boost/config/compiler/pgi.hpp? Thanks. -- Noel

2 5

Best method of checking for ==, <, streamability
by Christopher Jefferson 05 Nov '10

05 Nov '10

A user recently posted on boost-users asking for operator== support in boost::any. I have added this support to my private version, which extends spirit's implementation of any. I am interested in cleaning this up and submitting it to replace boost::any, if everyone is happy with this. The major issue is that my implementation requires type support operator==, operator< , hash and operator<< / operator>>. Are there any techniques in boost for detecting the presence of these operators which are sufficiently reliable they could be used in boost::any? Apologises for this question, but I have lost track of how well such checks can be implemented. Thanks, Christopher Jefferson

5 4

How about a call for bug-fix volunteers?
by Jim Bell 05 Nov '10

05 Nov '10

Library authors shouldn't have to bear the burden of bug reports for their libraries alone. Even maintaining their regression tests across all platforms isn't realistic (and isn't being done). So what if we try to recruit a couple helpers for each library? I see two avenues where we can use some volunteer help to improve boost's quality. 1. Dandelion pickers: look at failed/yellow regression tests and try to fix them. Volunteers could pick a platform they have access to, and try to work across all libraries. You might be surprised at how easy many are to fix. 2. Ticket triage: pick a library and go through all open tickets. Change severity, resolve, request more information, try to map to an existing regression test, interact with the submitter in the comments. (I think Trac needs a "requesting more information" state.) Even a brief 'triage' on each would be helpful, and help users feel they're being heard. (Trac needs to track this, too.) Write new regressions from legitimate issues. Recruiting pitch: it's a great way to improve your skills--troubleshooting, C++, platform-specific expertise, communication. Not so glamorous, but important and great experience. Perhaps we could assign a little of each to each SoC person. Maybe put the word out on Slashdot. See previous topic threads (http://lists.boost.org/Archives/boost/2010/10/index.php) "Report # 29 (aka 1.45 blockers) is too narrow" "[1.45] Beta final schedule" "[1.45] Thread issues"

14 18

Named subexpressions using \k<name>
by Chris Dragon 04 Nov '10

04 Nov '10

I have a user who says they have a bunch of expressions written for the Oniguruma regex library that use \k<name> syntax. According to http://www.boost.org/doc/libs/1_44_0/libs/regex/doc/html/boost_regex/syntax… Boost.regex should also support that syntax, but it does not seem to work. For example: boost::regex_replace( std::string("xyyz"), boost::regex(".*(?<first>x)(?<second>y).*", boost::regex_constants::perl), std::string("$+{second}\\k<first>"), boost::match_default | boost::format_all | boost::match_extra); results in "yk<first>" This result shows that $+{second} evaluated to "y" but \k<first> evaluated to "k<first>". I thought maybe the \k<name> syntax was only supported as a backreference within the first regex instead of the replace string, but this doesn't work either: boost::regex_replace( std::string("xyyz"), boost::regex(".*(?<first>x)(?<second>y)\\k<second>.*", boost::regex_constants::perl), std::string("$+{second}"), boost::match_default | boost::format_all | boost::match_extra); The result is "xyyz" meaning nothing was replaced (the regex did not match). However, if modify the regex to use \\g{name} syntax, it works: boost::regex_replace( std::string("xyyz"), boost::regex(".*(?<first>x)(?<second>y)\\g{second}.*", boost::regex_constants::perl), std::string("$+{second}"), boost::match_default | boost::format_all | boost::match_extra); The result is "y". Am I doing something wrong or have I found a bug? I'm using the latest stable 1_44_0 release compiled from source in VC9 and I also tried the pre-compiled library (http://sourceforge.net/projects/boost/files/boost-binaries/1.44.0/libboost_…) The OS is Windows 7. Thanks for the help!

2 2

[Boost.Timer] Why timer doesn't have a specific windows version using QueryPerformanceCounter() and QueryPerformanceFrequency()?
by dlaugt＠free.fr 04 Nov '10

04 Nov '10

Hi, Boost.Timer is based on the C function clock(). On windows, this function gives a precision of 1 millisecond only as CLOCKS_PER_SEC is equal to 1000. Why Boost.Timer doesn't use for windows QueryPerformanceCounter() and QueryPerformanceFrequency()? These functions give a precision around 2 microseconds. Daniel.

3 2

[build][doc] Testing section and run rule
by Mateusz Loskot 04 Nov '10

04 Nov '10

Hi, I'm looking for reference about run rule and the only thing I can find is this short reference: http://www.boost.org/doc/tools/build/doc/html/bbv2/reference/rules.html pointing to the Testing section which is currently blank: http://www.boost.org/doc/tools/build/doc/html/bbv2/tutorial/testing.html Would there be any chance to get this part updated? I'm looking for details about the run rule arguments: rule run ( sources + : args * : input-files * : requirements * : target-name ? : default-build * ) I'd appreciate if someone could explain what is the differences: sources vs input-files or args vs input-files. Best regards, -- Mateusz Loskot, http://mateusz.loskot.net Charter Member of OSGeo, http://osgeo.org Member of ACCU, http://accu.org

2 2