June 2008 - Boost - lists.preview.boost.org

[serialization] Boost 1.35.0 version breaks backward compatibility.
by Jurko Gospodnetić 12 Jun '08

12 Jun '08

Hi all. I am not sure if this was the desired or even acceptable effect but Boost 1.35.0 version of the serialization library introduces an optimization in vector.hpp when serializing a std::vector of 'trivially constructible' types (e.g. int :-)) that breaks backward compatibility when serializing into a text archive. Example for a std::vector<int> holding values { 2, 4 }: Old implementation: '2 0 2 4' New implementation: '2 2 4' We found a workaround to make boost use the old serialization implementation by wrapping our 'default constructible' objects inside a a simple class with a non trivial constructor. None of this is documented in http://www.boost.org/doc/libs/1_35_0/libs/serialization/doc/release.html#di…. ----------------------- Some related questions: ----------------------- * Does anyone know if Boost serialization library is intended to remain backward compatible across different Boost releases or at least provide for explicitly turning on such backward compatibility? * What to do when one wants to use new & improved serialization algorithms but needs to occasionally support older ones in the same application? I know our class serialization can be conditioned based on some 'version' parameter but what to do about 'standard' classes, e.g. std::vector<int>, whose serialization is implemented by the Boost serialization library itself? For example, a new 'network application' release supports a new & improved serialization algorithm but still wants to support connecting to old 'network applications' that do not know how to 'speak the new language'. Many thanks for any explanation regarding this topic. On a related but separate issue - The optimization is conditioned on whether the type is 'default constructible' which seems incorrect. Default constructibility test as currently implemented does actually just test whether the constructor is trivial (using a compiler specific intrinsic) but that is not what the name 'has_default_constructor()' suggests. Hope this helps. Best regards, Jurko Gospodnetić

2 1

[thread] condvar wait exception
by Emil Dotchevski 12 Jun '08

12 Jun '08

Hello, I'm reading in the documentation that barrier::wait may throw thread_resource_error. Looking at the source, it uses condition_variable::wait, which in turn (I think) throws nothing on windows, but may throw boost::condition_error under pthread. So the documentation appears to be wrong. But also, I am puzzled why would wait() throw anything but boost::thread_interrupted? Let's say I have two threads call wait() on a barrier, and one of them throws an error. What am I supposed to do with that exception? How can I recover from such a failure? I am not an expert but I think that if the exception is used to report a bug in user code, it would be better to assert() instead of throwing an exception. One final note, Boost Thread doesn't use boost::throw_exception to throw exceptions. This means that the exception_ptr functionality in Boost Exception is crippled for exceptions emitted by Boost Thread. This is unfortunate because transporting exception objects between threads is a key motivation for boost::exception_ptr; see http://svn.boost.org/svn/boost/trunk/libs/exception/doc/cloning.html. Emil Dotchevski Reverge Studios, Inc. http://www.revergestudios.com/reblog/index.php?n=ReCode

2 1

erfc id not found
by Gary.Stevens＠deq.idaho.gov 12 Jun '08

12 Jun '08

I have been trying to compile code that uses the erfc function. I can get it to build successfully using DevC++4.9.9.2, but cannot using Win-Visual C++2005 Express (C90 standard) which I have been using to construct a form for input-output. I have downloaded the boost libraries and have changed the configuration properties to include the boost directory and changed to Not Using Precompiled Headers. I continue to receive the following: 'erfc': identifier not found. I have tried adding the libraies for DevC++4.9.9.2 but receive the same error. I am new to C++ and this a small simple program but am spending a significant amount of time getting the erfc issue resolved. Any help would be very appreciated. Thank you. Gary

2 4

[ptr_container] Failures for ptr_set and ptr_unordered_set on Intel, Acc, and Cxx
by Markus Schöpflin 12 Jun '08

12 Jun '08

Hello, currently these tests fail with "../boost/ptr_container/ptr_set_adapter.hpp", line 362: error #2309: more than one instance of constructor "boost::ptr_container_detail::ptr_set_adapter_base<Key, VoidPtrSet, CloneAllocator, Ordered>::ptr_set_adapter_base [with Key=Base, VoidPtrSet=std::set<void *, boost::void_ptr_indirect_fun<std::less<Base>, Base, Base>, std::allocator<void *>>, CloneAllocator=boost::heap_clone_allocator, Ordered=true]" matches the argument list: function template "boost::ptr_container_detail::ptr_set_adapter_base<Key, VoidPtrSet, CloneAllocator, Ordered>::ptr_set_adapter_base(const Compare &, const Allocator &) [with Key=Base, VoidPtrSet=std::set<void *, boost::void_ptr_indirect_fun<std::less<Base>, Base, Base>, std::allocator<void *>>, CloneAllocator=boost::heap_clone_allocator, Ordered=true]" function template "boost::ptr_container_detail::ptr_set_adapter_base<Key, VoidPtrSet, CloneAllocator, Ordered>::ptr_set_adapter_base(InputIterator, InputIterator) [with Key=Base, VoidPtrSet=std::set<void *, boost::void_ptr_indirect_fun<std::less<Base>, Base, Base>, std::allocator<void *>>, CloneAllocator=boost::heap_clone_allocator, Ordered=true]" argument types are: (boost::void_ptr_iterator<__rw::__rw_tree_iter<void *, ptrdiff_t, void *const *, void *const &, __rw::__rw_rb_tree_node<std::allocator<void *>, void *, void *, __rw::__ident<void *, void *>>>, Base>, boost::void_ptr_iterator<__rw::__rw_tree_iter<void *, ptrdiff_t, void *const *, void *const &, __rw::__rw_rb_tree_node<std::allocator<void *>, void *, void *, __rw::__ident<void *, void *>>>, Base>) : base_type( first, last ) The compilers has difficulties choosing between: template< class Compare, class Allocator > ptr_set_adapter_base( const Compare& comp, const Allocator& a ) and template< class InputIterator > ptr_set_adapter_base( InputIterator first, InputIterator last ) Interestingly, if I just remove the following constructors, these tests pass for me on Cxx, and no other failures are introduced. template< class Compare, class Allocator > ptr_set_adapter_base( const Compare& comp, const Allocator& a ) template< class Compare, class Allocator > associative_ptr_container( const Compare& comp, const Allocator& a ) So are these constructors needed at all? Markus

2 2

[thread_safe_signals] reviving boost::trackable?
by Frank Mori Hess 12 Jun '08

12 Jun '08

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 It's occurred to me that thread-safe automatic connection management for signal/slot connections made in an object's constructor to one of its member functions could be accomplished by using a boost::trackable class that looks like class trackable: public enable_shared_from_this<trackable> { //... }; Above, I'm only requiring the "classic" enable_shared_from_this functionality, no special support for shared_from_this in constructors, etc. You would derive from trackable, visit_each would discover it inside bind objects, and its destructor would disconnect its associated connections, similar to how the current signals library works. It would do a bit more though. When the signal is invoked, it would: 1) check that the connection has not been disconnected. If not, it calls shared_from_this() to obtain a shared_ptr which can be used for connection tracking in the usual thread_safe_signals way. Calling shared_from_this can be made safe since the trackable destructor (which disconnects) will be called before the enable_shared_from_this destructor. 2) if the shared_ptr from shared_from_this is empty, that means the trackable object is either in the process of being destroyed, or hasn't been passed to an external shared_ptr owner yet. In either case, the slot is skipped by the signal invocation. However, there is a downside to this and other schemes where the signal/slot connection is made in the constructor but blocked until the object is passed to an external shared_ptr. In a common use case, an observer will connect to an observed object's signal and then manually query the observed object once to get its initial state. It can then rely on the signal to keep in sync with any changes the observed object undergoes. If the signal is blocked until the observer is owned by an external shared_ptr, it raises the possibility of the observer getting out of sync with the observed, since a signal could occur and be missed after the manual query of the observed object's initial state. So, doing all the connection setup in a postconstructor still seems preferable. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.6 (GNU/Linux) iD8DBQFIUS7c5vihyNWuA4URAhPIAKC40cK1kgX9o7aPYILBaHZu3jwZhQCfYFjw o/drv0EOsHqvEt+7exy/PDo= =Lj6m -----END PGP SIGNATURE-----

1 0

Re: [boost] [circular_buffer] why is there no reserve(size_type)?
by Jan Gaspar 12 Jun '08

12 Jun '08

> no, but there are overlaps in functionality/use-cases and not having > reserve makes it harder to write generic code that accepts either type. list and deqeue functionality/use-cases overlap with vector too but they do not have reserve as well ... But if you really need it you can implement it by yourself, don't forget to document it properly and add unit tests. Ideally if you can send me a patch. Jan Send instant messages to your online friends http://uk.messenger.yahoo.com

1 0

Re: [boost] [circular_buffer] why is there no reserve(size_type)?
by Jan Gaspar 12 Jun '08

12 Jun '08

Sorry I don't have time to do this. Anyway IMO implementing reserve would give an impression that the capacity is not fixed and increases as you put elements into the circular buffer. And at last circular buffer is not a vector. ----- Original Message ---- From: Thorsten Ottosen <thorsten.ottosen(a)dezide.com> To: boost(a)lists.boost.org Sent: Wednesday, 11 June, 2008 9:49:29 PM Subject: Re: [boost] [circular_buffer] why is there no reserve(size_type)? Jan Gaspar skrev: > because: > 1) it's behaviour is slightly different from the standard - you can increase but also decrease the capaciry of the container; decreasing of the capacity is not supported by reserve() > 2) the word capacity is used all over the documentation of the circular buffer. I think set_capacity expresses better what the method does. So the function is slightly different from the one in std::vector :-) Having a reserve() won't hurt ... in fact, it will make it easier to write code that works for both vector and circular_buffer. -Thorsten _______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost __________________________________________________________ Sent from Yahoo! Mail. A Smarter Email http://uk.docs.yahoo.com/nowyoucan.html

2 1

interest in thread-pool library
by Oliver Kowalke 12 Jun '08

12 Jun '08

Hi, I don't know why but my previous email from yesterday did not occur on the mailing list? I've uploaded two versions of my threadpool library - maybe you are interressted in. It provides thread management, queueing and scheduling policies as well chaining of tasks and lazy processing of submited tasks (== processed only if requested). version 1: submit returens a future (Braddock future library) whic his cancelable (remove pending task from queue) version 2: submit returns a task object which provides a interrupt memebr function (task can be interrupted if its is running and cooperative) Both libraries are available at Boost.Vault / Concurrent Programming. so long, Oliver -- Psssst! Schon vom neuen GMX MultiMessenger gehört? Der kann`s mit allen: http://www.gmx.net/de/go/multimessenger

1 0

Boost build mailing list moderation?
by Emil Dotchevski 12 Jun '08

12 Jun '08

Recently I subscribed and made 2 posts to the boost build mailing list, and got the standard reply that they need to be reviewed by the moderators and that they will either be posted or I'll get a notification that there is a problem. I got no notification but I don't see my questions appearing in the boost build mailing list. My 2 posts have the following subject lines: - Using boost build on linux - Testing problem with Boost Build Any ideas? Emil Dotchevski Reverge Studios, Inc. http://www.revergestudios.com/reblog/index.php?n=ReCode

2 1

Boost exceptions under Windows
by Zack S 11 Jun '08

11 Jun '08

Hi folks, Why doesn't the Boost code, in the cases where it needs to throw an exception under Windows, offer the option to display a dialog window? Something like a #define'able option could be useful. Zack

10 17

Re: [boost] [circular_buffer] why is there no reserve(size_type)?
by Jan Gaspar 11 Jun '08

11 Jun '08

because: 1) it's behaviour is slightly different from the standard - you can increase but also decrease the capaciry of the container; decreasing of the capacity is not supported by reserve() 2) the word capacity is used all over the documentation of the circular buffer. I think set_capacity expresses better what the method does. Jan ----- Original Message ---- From: Steven Watanabe <watanabesj(a)gmail.com> To: boost(a)lists.boost.org Sent: Wednesday, 11 June, 2008 3:49:21 PM Subject: Re: [boost] [circular_buffer] why is there no reserve(size_type)? AMDG Jan Gaspar wrote: > it is called set_capacity(capacity_type) > Why is this different from the standard library name? In Christ, Steven Watanabe _______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost __________________________________________________________ Sent from Yahoo! Mail. A Smarter Email http://uk.docs.yahoo.com/nowyoucan.html

2 1

Building boost 1.35 with boost.serialization 1.36
by Chris Hamilton 11 Jun '08

11 Jun '08

I require a few of the new features introduced in Boost.Serialization 1.36, namely the support of serialization across multiple shared libraries and multi-threaded support. The Serialization library itself doesn't build cleanly, but adding a missing #include in one of the headers solves that for me. However, I now get errors when building the MPI library: gcc.compile.c++ bin.v2/libs/mpi/build/gcc-3.4.6/release/threading-multi/packed_iarchive.o In file included from libs/mpi/src/packed_iarchive.cpp:11: ./boost/archive/impl/archive_pointer_oserializer.ipp:27: error: redefinition of `class boost::archive::detail::<unnamed>::serializer_map<Archive>' ./boost/archive/impl/archive_pointer_iserializer.ipp:27: error: previous definition of `class boost::archive::detail::<unnamed>::serializer_map<Archive>' "g++" -ftemplate-depth-128 -O3 -finline-functions -Wno-inline -Wall -fPIC -pthread -DBOOST_ALL_NO_LIB=1 -DBOOST_MPI_DYN_LINK=1 -DBOOST_MPI_SOURCE=1 -DNDEBUG -I"." -I"/opt/lam/gnu/include" -c -o "bin.v2/libs/mpi/build/gcc-3.4.6/release/threading-multi/packed_iarchive.o" "libs/mpi/src/packed_iarchive.cpp" ...failed gcc.compile.c++ bin.v2/libs/mpi/build/gcc-3.4.6/release/threading-multi/packed_iarchive.o... ...removing bin.v2/libs/mpi/build/gcc-3.4.6/release/threading-multi/packed_iarchive.o gcc.compile.c++ bin.v2/libs/mpi/build/gcc-3.4.6/release/threading-multi/packed_oarchive.o gcc.compile.c++ bin.v2/libs/mpi/build/gcc-3.4.6/release/threading-multi/packed_skeleton_iarchive.o In file included from libs/mpi/src/packed_skeleton_iarchive.cpp:11: ./boost/archive/impl/archive_pointer_oserializer.ipp:27: error: redefinition of `class boost::archive::detail::<unnamed>::serializer_map<Archive>' ./boost/archive/impl/archive_pointer_iserializer.ipp:27: error: previous definition of `class boost::archive::detail::<unnamed>::serializer_map<Archive>' "g++" -ftemplate-depth-128 -O3 -finline-functions -Wno-inline -Wall -fPIC -pthread -DBOOST_ALL_NO_LIB=1 -DBOOST_MPI_DYN_LINK=1 -DBOOST_MPI_SOURCE=1 -DNDEBUG -I"." -I"/opt/lam/gnu/include" -c -o "bin.v2/libs/mpi/build/gcc-3.4.6/release/threading-multi/packed_skeleton_iarchive.o" "libs/mpi/src/packed_skeleton_iarchive.cpp" ...failed gcc.compile.c++ bin.v2/libs/mpi/build/gcc-3.4.6/release/threading-multi/packed_skeleton_iarchive.o... ...removing bin.v2/libs/mpi/build/gcc-3.4.6/release/threading-multi/packed_skeleton_iarchive.o gcc.compile.c++ bin.v2/libs/mpi/build/gcc-3.4.6/release/threading-multi/packed_skeleton_oarchive.o ...skipped <pbin.v2/libs/mpi/build/gcc-3.4.6/release/threading-multi>libboost_mpi-gcc34-mt-1_35.so.1.35.0 for lack of <pbin.v2/libs/mpi/build/gcc-3.4.6/release/threading-multi>packed_iarchive.o... Any idea on how to solve these? It seems that Boost.MPI is rather 'chummy' with the internals of Boost.Serialization, and that things have changed rather significantly there in 1.36. Regards, Chris Hamilton

2 1

[serialization] ambiguous symbol if include "binary_iarchive.hpp"
by Wang Yun 11 Jun '08

11 Jun '08

Hi, I meet a compile error of boost::function only when I add this line #include <boost/archive/binary_iarchive.hpp> I'm using VC2005, boost 1.35.0 error message: error C2872: 'function' : ambiguous symbol 1> could be 'e:\libraries\boost_1_35_0\boost\function\function_base.hpp(106) : boost::function' 1> or 'boost::detail::function' 1> or 'e:\libraries\boost_1_35_0\boost\function\function_base.hpp(106) : boost::function' // ======= the code =============================== #include "stdafx.h" #include <boost/function.hpp> using boost::function; #include <boost/archive/binary_iarchive.hpp> //////////////// when add this line, error will happen int _tmain(int argc, _TCHAR* argv[]) { function<void()> fn; return 0; } //================================================== Thanks, Wang Yun

3 2

[python] can I stop iteration without exception being thrown?
by Gennadiy Rozental 11 Jun '08

11 Jun '08

Hi, I an using vector_indexing_suite for both standard containers export and for our in-house ones. I've noticed that boost::python::error_already_set is bwing thrown every time iteration in python over container instances is about to finish. The exception is ignored somewhere inside boost::python but it's still quite inconvenient for debugging: I want to catch actual error, but can't skip 100s of fake ones. Is there any reason why it was designed this way? Is there any way to avoid it? Gennadiy

6 14

Re: [boost] [circular_buffer] why is there no reserve(size_type)?
by Jan Gaspar 11 Jun '08

11 Jun '08

it is called set_capacity(capacity_type) ----- Original Message ---- From: Thorsten Ottosen <thorsten.ottosen(a)dezide.com> To: "boost(a)lists.boost.org" <boost(a)lists.boost.org> Sent: Wednesday, 11 June, 2008 2:05:14 PM Subject: [boost] [circular_buffer] why is there no reserve(size_type)? Hi Jan, The headline pretty much says it all. -Thorsten _______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost __________________________________________________________ Sent from Yahoo! Mail. A Smarter Email http://uk.docs.yahoo.com/nowyoucan.html

2 1

[circular_buffer] why is there no reserve(size_type)?
by Thorsten Ottosen 11 Jun '08

11 Jun '08

Hi Jan, The headline pretty much says it all. -Thorsten

1 0

Bug in iterator_range when NDEBUG not defined but boost asserts are enabled
by John Reid 11 Jun '08

11 Jun '08

iterator_range.hpp defines is_singular() if NDEBUG is not defined but it calls is_singular through the BOOST_ASSERT macro. Unfortunately it is possible to enable BOOST_ASSERT and have NDEBUG defined which leads to a compilation error. To see this compile the iterator_range.cpp example included with the library with the following defines: gcc -DBOOST_ENABLE_ASSERT_HANDLER -I/usr/local/src/boost-trunk/ -DNDEBUG -c iterator_range.cpp John.

4 3

Using Bind & Lambda
by Robert Jones 11 Jun '08

11 Jun '08

This is my first time posting to this list, so I hope you'll forgive me it my etiquette is awry. I'm new to the Boost libraries, and just learning my way around, guided mainly by Bjorn Karlson's book - any other book recommendations would be most welcome btw, and there a few things I just don't quite get! First - is "boost/lambda/bind.hpp" equivalent to "boost/bind.hpp"? Should it ever be necessary to include both? Second - Is it true that a lambda function is syntactically and semantically the same as a bind finction, eg Given:- int foo( int i ) { return i + 3; } Are these two expressions the same? bind( &foo, _1 ); _1 + 3; Finally (for now!) - why is this wrong? struct A { string identity( string ) const; } a; vector<string> v; for_each( v.begin(), v.end(), cout << bind( &A::identity, a, _1 ) << '\n' ); which is intended to output the return value of A::identity() for each element of v, separated by newlines. Thanks in advance for any guidance! Rob. -- ACCU - Professionalism in programming - http://www.accu.org

5 4

[python] working with raw const pointers
by Gennadiy Rozental 11 Jun '08

11 Jun '08

Hi, In our system there are a lot of pointers to const objects flying around. I've got several problems in this regard. They can be exhibited by the following code snippet: a.cpp : class A{}; void boo1( A const* ) {} void boo2( A* ) {} A const* make_const_a() { return (A const*)0; } A* make_non_const_a() { return (A*)0; } std::vector<A const*> make_vec() { return std::vector<A const*>( 1 ); } .... class_<A,boost::noncopyable>( "A", bp::no_init ); class_<std::vector<A const*> >( "AVec", bp::no_init ) .def( vector_indexing_suite<std::vector<A const*> >() ) ; def( "boo", &boo1 ); def( "boo", &boo2 ); def( "make_const_a", &make_const_a, return_value_policy<reference_existing_object>() ); def( "make_non_const_a", &make_non_const_a, return_value_policy<reference_existing_object>() ); def( "make_vec", &make_v ec ); Problem I. Access to mutable func Following python statement mymodule.boo( mymodule.make_non_const_a() ) seems to call incorrect function boo Problem II. Vector fails completely Following python statements vec = pytoolkit.make_vec(); for e in vec: pass produce TypeError: No to_python (by-value) converter found for C++ type: class A const * Any advise? Gennadiy

1 0

Threaded version
by albert meyburgh 11 Jun '08

11 Jun '08

Is there a threaded version of some of the boost components? For example, calling find on a container should be easily threaded. I realize the user could also easily thread this, but it seems like a trivial thing to just add support for in boost itself.

3 2

[python] frame corruption in python -> c++ -> python scenario
by Gennadiy Rozental 10 Jun '08

10 Jun '08

Hi, I am experiencing strange issue I can't find solution for. Sorry I'll need a bit lengthy explanation to describe it. I am using boost python for both automation and embedding the same time. I wrote trivial unit test module (can't use standard one). It has function check like this: def check( val ): check_impl( val ) def check_impl( val ): ... print "Error at line " + str(sys._getframe(2).f_lineno+1) ... ... This supposedly should produce error line location. And it does. Sometimes. test1.py: ... 8> check( False ) produces: Error in line 8: ... I've noticed though that in some cases line numbers are incorrect. I was able to narrow it down to the single function mymodule.foo() Before this function call line numbers are correct. After it incorrect. foo is a function implemented in c++ and exported using boost.python. I was able further to identify that the line at fault inside function foo implementation is the one invoking python function: void foo() { ... boost::python::object obj = ... obj(); ... } So. Finally the question: Is there way to fix/circumvent this somehow? Thanks, Gennadiy

1 0

boost::decimal ?
by Mårten Rånge 10 Jun '08

10 Jun '08

Is there some work been done on implementing a boost::decimal class? If there isn't my guess is that this would be very useful to C++ programmers out there if boost could provide this (we sure would dance with joy!). I would like to help but frankly I lack the competence on how to implement a decimal class efficiently and with the robustness it needs. A decimal class would be very useful for ERP and Financial systems as in many cases of the typical customers work with the decimal system instead of the binary system. It's hard to motivate for the customers why: 1 -0.2 -0.2 -0.2 -0.2 -0.2 doesn't add up to _exactly_ 0 when using double (windows calc.exe had this problem many years ago and it attracted lot of gleeful comments, so gleeful calc.exe uses infinite precision these days to represent rational numbers). In addition many sums are intended to become 0 at the end (the debit and credit should be balanced) so cancellation errors like above actually are quite common. The key issue is that common numbers such as 0.1, 0.2, and so on can't be exactly represented in a double. http://www2.hursley.ibm.com/decimal/ -- Mårten

4 3

[serialization] Base64 padding in basic_text_oprimitive
by Vyacheslav E. Andrejev 10 Jun '08

10 Jun '08

Hello, Code that pads base64 output with '=' looks in basic_text_oprimitive.ipp like following: std::size_t padding = 2 - count % 3; if(padding > 1) *oi = '='; if(padding > 2) *oi = '='; It will give the following transformations of a few simple strings: "012" -> "MDEy=" "0123" -> "MDEyMw" "01234" -> "MDEyMzQ" "012345" -> "MDEyMzQ1=" Which violates the principle of the base 64 that the encoded output length is multiple of 4 bytes. Right output for the samples above should be: "012" -> "MDEy" "0123" -> "MDEyMw==" "01234" -> "MDEyMzQ=" "012345" -> "MDEyMzQ1" The problem with the code above is that the condition (padding > 2) is never satisfied. I think that the proper code should look like std::size_t padding = 2 - count % 3; if (padding < 2) { *oi = '='; if (padding > 0) *oi = '='; } -- -- Vyacheslav Andrejev -- System Architect, Optech International, Inc.

2 3

[serialization] implementing class_name_optional_type
by Esteve Fernandez 10 Jun '08

10 Jun '08

Hi all, I'm working on the "YAML archives for Boost.Serialization" GSoC project and I've already implemented a working output archive. However, since YAML requires the class name to be always serialized, I had to implement the serialization of a new type called class_name_optional_type in basic_oarchive.cpp, which lets the archive decide if class names should always be written. It was discussed in the past [1], but never implemented, AFAIK. Cheers. PS: should I send the patch to this list or is it better to create a new ticket? 1 - http://lists.boost.org/boost-users/2006/06/20319.php

2 9

Request: RDTSC
by dherring＠ll.mit.edu 10 Jun '08

10 Jun '08

The CPU timestamp counter is a high-resolution clock source useful for profiling code or controlling tight timing loops. AMD/Intel CPUs support this via the RDTSC opcode, and other CPUs often provide similar functionality. Apparently MS compilers support an __rdtsc() intrinsic, but I couldn't find one for gcc. The code at http://www-unix.mcs.anl.gov/~kazutomo/rdtsc.html seems to do a good job for gcc; it avoids a couple bugs in code found at other sites. Any possibility of a raw boost::rdtsc() or boost::readTSC()? Something that doesn't compile on unsupported architectures would be good. Other code can worry about whether this is an accurate time source[1]; I just want a portable way to access the raw counter itself. Thanks, Daniel [1] Search "rdtsc site:boost.org"

4 3