June 2005 - Boost-users - lists.boost.org

Re: [Boost-users] Serialization problem in a shared library
by Tapio Vuorinen (JO/LMF) 20 Jun '05

20 Jun '05

Hi, On Sat, 2005-06-18 at 06:24, Robert Ramey wrote: >I believe that the next releae of boost - version 1.33 - will address this > problem. > Thanks for the answer. If the CVS already contains the required bugfixes for the problem could you please list the files which I need to merge into our copy of Boost to get the important bugfixes, the shared-library fixes, so that I do not break too much the existing code? Do the bugfixes depend on other Boost libraries or is it only the serialization library I need to fix? By the way, when the next version (1.33.0) will be released? Thanks! Tapio Vuorinen

1 0

Using filesystem.path on OS/390
by Tom Quarendon 19 Jun '05

19 Jun '05

In the filesystem rationale documentation there are references to the need and desire to represent paths on OS/390. However I can find no documentation about how this is achieved. Does the filesystem path class support OS/390 paths such as the syntax for PDSE members or DDnames? If so are there some examples? Many thanks. Tom Quarendon.

2 1

Spirit inserting values into map
by Leon Mergen 19 Jun '05

19 Jun '05

Hello, First of all, I love spirit. I've started using it yesterday and I cannot believe how easy it is. At the moment, I'm trying to parse digits that are supposed to go into a std::map < unsigned int, bool >. On a certain part, if a digits matches, the value /always/ needs to be true. Now, I'm trying to figure out how to statically set the std::map < unsigned int, bool>::value_type that Spirit allows you to assign values to; especially, how can I define that for a certain match, the ->second part should be true, instead of assigning it based on a match of the grammar ? I wasn't able to find this in the manual. I hope I'm making sense... :) Thanks in advance for any responses. Regards, Leon Mergen

1 0

Boost.Iostreams: patch for putback
by Takeshi Mouri 19 Jun '05

19 Jun '05

Hello, I found a trivial mistake in boost/iostreams/read.hpp of current CVS HEAD version. So, I report it. Thanks Takeshi Mouri *** read.hpp.orig Mon May 30 16:30:27 2005 --- read.hpp Sun Jun 19 23:42:40 2005 *************** *** 158,164 **** { // gcc 2.95 needs namespace qualification for char_traits. typedef typename char_type_of<T>::type char_type; typedef iostreams::char_traits<char_type> traits_type; ! return !traits_type::is_eof(t.sputbackc()); } }; --- 158,164 ---- { // gcc 2.95 needs namespace qualification for char_traits. typedef typename char_type_of<T>::type char_type; typedef iostreams::char_traits<char_type> traits_type; ! return !traits_type::is_eof(t.sputbackc(c)); } };

2 1

lexical_cast update + Boost.Test question
by Terje Slettebø 19 Jun '05

19 Jun '05

Hi. lexical_cast has now been updated in the CVS, with the following, as has been reported as bugs or suggested improvements: - Fixed so it doesn't trim any whitespace when converting (it used to throw on " 123" -> int, but not on "123 " -> int, which was inconsistent). - Fixed reversed bad_lexical_cast constructor arguments. - lexical_cast now uses pass-by-const reference. For compilers lacking support for partial specialisation of class templates (like MSVC 6), it falls back on pass-by-value. Sorry for the late update. It's been quite busy at work lately, so it's been hard to concentrate on non-work projects, but I'm now more on top of things. I'll read up on any lexical_cast-related postings, and can make subsequent updates, should it be desired. I'll also check up the results after the next regression test run, to see if there are any remaining issues. The changes are of course done in cooperation with Kevlin. It has been tested on Intel C++ 7.1, MSVC 2005 Final Beta, g++ 3.4 and Borland C++ Builder X, and all tests pass there. I get errors (from Boost.Test) when trying to run the test on MSVC 6, but testing lexical_cast by itself seems to work. Might Boost.Test have a problem with MSVC 6? By the way, I don't see the changes using the Boost ViewCVS, but it's listed e.g. with "log". Are the ViewCVS results cached, somehow, or may the commit have ended up on some other place than the trunk? I did make sure to check out the trunk, before commit. Regards, Terje

3 3

Serialization and BGL
by Jean-Noël Rivasseau 19 Jun '05

19 Jun '05

Hello, I was wondering if one could use Boost serialization library with the Boost Graph Library. Specifically, if one my class holds a graph (for example, an adjacency_list class), is it possible to serialize it? From what I've read from the Boost serialization documentation, this doesn't seem possible as it eventually requires access to the class members; in the case of the BGL, the implementation of the classes is probably so complex that I wouldn't go through that route; plus, they may very well be private. I was wondering if there is any way to do serialization via the public interface of the BGL (ie, add_edge, etc...) ? If I understood the Serialization library correctly, everything works via the operator & which acts as >> and <<. If I define a << operator for the BGL classes, that implement the export / import operations, needed, would it be possible to use the BGL with serialization? Any help from knowledgeable people would be appreciated (I now have a quite good knowledge of the BGL but I am new to the serialization library). Jean-Noël

4 4

boost.build : need help/bug specifying struct-alignment ?
by Seb Martel 18 Jun '05

18 Jun '05

Hi, I am trying to build boost.threads 1.32 statically with struct alignment set to 1. I dug up from the documentation that this should do the trick: bjam -sTOOLS=vc-7_1 "-sBUILD=debug <struct-alignment>1" -a -d+2 however, I don't see the alignment switch '/Zp1' on the command line : "cl" /Zm800 -nologo /EHsc -c -DBOOST_THREAD_LIB_NAME=boost_thread -DBOOST_THREAD_BUILD_LIB=1 /Z7 /Od /Ob0 /EHsc /GR /MDd /Op Zc:forScope /Zc:wchar_t -I"..\..\..\bin\boost\libs\thread\build" -I"C:\Program Files\Microsoft Visual Studio .NET 2003\VC7\include" I"G:\boost_1_32_0" -Fo"..\..\..\bin\boost\libs\thread\build\libboost_thread.lib\vc-7_1\debug\threading-multi\tss.obj" -Tp"..\..\..\libs\thread\build\..\src\tss.cpp" Is there something I am missing ? -seb

2 2

[tokenizer/string_algo] splitting after a sequence
by Yuval Ronen 18 Jun '05

18 Jun '05

Hi. I need to tokenize a string according to a delimiter sequence, rather than the common "any-of-characters" algorithm. What I mean is that if I'll use the sequence "ab" to split the string "1ab2ab3", I'll get { "1", "2", "3" }, but if I'll use the same "ab" sequence to split a slightly different "1ab2ba3" string, I'll get { "1", "2ba3" }. I've looked at the tokenizer lib and string_algo's split method, but couldn't find any evidence that any of them already contain such a feature. Is there such a thing in Boost? Thanks, Yuval

3 2

smart pointer declaration question
by Jack Nguyen 18 Jun '05

18 Jun '05

I currently have template <typename T> class Vector { private: int Size; boost::shared_array<T> mpD; void VecAlloc( int size) { boost::shared_array<T> mpD_temp(new T[size]); mpD=mpD_temp; } public: Vector() { } Vector(int size) :Size(size) { VecAlloc(size); } } My question is how do i get rid of boost::shared_array<T> mpD_temp(new T[size]); ???

4 7

Generating a unique symbol
by Martin Slater 18 Jun '05

18 Jun '05

Hi there, Is there anything in boost that will generate a guarenteed unique symbol, possibly based on file and line? I've had a look through the preprocessor library but cannot see anything. Strict portabilty is not an issue as my only requirement is for vc7.1. thanks Martin

2 2