February 2010 - Boost - lists.preview.boost.org

[date_time] int64 overflow bug with subsecond duration
by Zachary Turner 25 Feb '10

25 Feb '10

Consider the following code, very simple in nature: static const boost::gregorian::date windows_epoch_base(1601, 1, 1); date_time d(windows_epoch_base, microseconds(12908391903991608)); This should resolve to sometime in early 2010 Now, the problem is that this number is already really big, not that many factors of 10 away from the int64 overflow boundary. microseconds() is really just the subsecond_duration<> template, and on my machine it's being parameterized as subsecond_duration<boost::posix_time::time_duration, 1000000>. The code for the constructor is as follows: explicit subsecond_duration(boost::int64_t ss) : base_duration(0,0,0, ss*traits_type::res_adjust() / frac_of_second) {} res_adjust is also resolving to 1000000. So compiler is doing 12908391903991608*res_adjust first, which causes an overflow. I think it should be doing ss * (traits_type::res_adjust() / frac_of_second) instead. I don't know a whole lot about date_time internals, so I thought I would ask here before just submitting a bug. Is the proposed fix correct? I made the change locally and it does indeed appear to resolve the problem. Zach

2 3

[move] will boost::rv work with types with virtual destructor?
by DE 25 Feb '10

25 Feb '10

sorry for a dumb question i noticed boost move lib mentioned in a recent message and decided to look at it i noticed that destructor for boost::rv class is declared but not defined so it can no longer be generated automatically then if an arbitrary type has a virtual destructor and we want to parameterize the boost::rv template with it a vtable for the boost::rv<type> must be generated among other things it must hold a pointer to rv<type>::~rv which is not defined is that really a problem? -- Pavel

2 2

Spirit V1.8 to V2.1 Porting Errors
by Alireza Bemanian 25 Feb '10

25 Feb '10

Hi, Trying to port the code to V2.1, a number of redefinition errors get generated by the compiler. The first instance is the extern InterLockedIncrement variable at line 73 of the interlocked.hpp file. The code is segmented for different platforms using #if #elif and so on directives. The project is at the same time ported from Visual Studio 2008 to 2010 RC1, and the redefinition error is caused by compiler not excluding parts of that file. The define by the Visual Studio is the _WIN32_WINNT base on the use of the 32 bit development PC. This definition is not listed by the #elif chain within the interlocked.hpp code and it seems that BOOST_USE_WINDOWS_H is used instead. Knowing the segments are coded to be mutually exclusive, still the code generates errors? Regards, Reza Bemanian

2 4

Re: [boost] is review system in place is extremely slow? (wasRe:[rfc] rcpp)
by Pete Bartlett 25 Feb '10

25 Feb '10

>From Robert Ramey > > to move those 3 libraries to the 'detail' namespace > > of Boost.Task and have review as it is, as opposed to waiting. What > > do you think? >I think I caught hell for doing something similar in the serialization >library. I had to make a number of components such as >BOOST_STRONGTYPEDEF, state_saver, smart_cast, etc. >which I put into boost - (not detail) and year afterwards this >was raised as a huge problem. And this was even though the >components had been their through two reviews. So I would >be careful about doing this. This seems a little more negative than I remember. If we are thinking of the same thread (see e.g. http://lists.boost.org/Archives/boost/2007/11/130567.php ) then a lot of the discussion was about headers directly in the boost root directory rather than a subdirectory, rather the issue at hand. My opinion is that there is no overarching programming design deity that is going to come down and smite us if we be a bit pragmatic. So what if Move [*], Fiber and Atomic get "smuggled" in if Task gets reviewed and accepted first? They don't have to be documented as accepted - on the contrary the documentation could have appropriate warnings as *not* formally accepted apart for indirect use via Task. >Another issue is: if Boost.Task depends upon Boost.Fiber >and Boost.Atomic, what happens if the Boost.Fiber or >Boost .Atomic are not approved? Sometimes the laudable goal of perfection turns open source projects into games of Nomic! Again, pragmatism should play a role here else Boost will never be the rich and full set of libraries that it fantastically could be. Pete [*] Move seems so core these days - another bit of pragmatism would see that fast-tracked through.

5 5

typeof or similar
by Steffen Roeber 25 Feb '10

25 Feb '10

Hi, I have 3 classes: 1. class Base { public: virtual Base *createDerived() { return 0; } }; 2. template <class T> class BaseT : public Base { public: virtual Base *createDerived() { return new DerivedT<T>(); //nicht smart, because DerivedT must be known } }; 3. template <class T> class DerivedT : public BaseT<T> { }; So far so good. Or not. But now the problems are starting. Base *base = new BaseT<int>; This object it created at a place that is not known where I want to create: Base *derived = base->createDerived(); //base is a Base * The result is an object that has the same template type as 'base'. Finally I got my desired result, but the problem is the relation downward relation between BaseT and DerivedT. My preferred way would be something like a factory: class Factory { public: template <class T> static Base *createDerived<T>() { return new DerivedT<T>(); } }; But how can I tell the factory which type to be used. Base *derived = Factory::createDerived<sameAsBase> (base); //base is a Base * I think there is no "normal" C++ language support. But Boost has some methods that by-pass some language constraints. I took a loot at something like 'typeof'. But I don't get the things together. Has anybody any idea? Or is it impossible to do this? Kind regards Steffen

4 3

phoenix docs hidden
by Neal Becker 25 Feb '10

25 Feb '10

In 1.42, there is no link at top level docs to phoenix. It's fairly hidden under spirit. Isn't phoenix intended to be used independently?

2 1

[phoenix] typo in docs
by Neal Becker 25 Feb '10

25 Feb '10

In algorithms doc: for_each(r, c) << Shouldn't that say 'foreach(r, f)' to match this: for_each(begin(r), end(r), f)

1 0

[fusion] make_unfused_generic broken
by Rutger ter Borg 25 Feb '10

25 Feb '10

Hello, apparently the header #include <boost/fusion/functional/generation/make_unfused_generic.hpp> got broken somewhere in the upgrade process, including it yields lots of errors. Boost version 1.42. (replaced with make_unfused?) Cheers, Rutger

2 1

Re: [boost] [msm] scalability
by Christophe Henry 25 Feb '10

25 Feb '10

>Sorry for the delay. That is awesome Christoffe! I've turned off /Gm for >all my projects and my state machine now compiles in the 6-sec range, >and Incredibuild is able to keep the entire compile farm running at >capacity. You shaved >30% off my full build times. Great :) In case your state machines grow bigger, you can also give a try with the "faster compile" policy which I will further improve after inclusion. I think we should make /Gm a forbidden option. I will of course write about it in MSM's doc but it would be interesting to see which libraries also profit from disabling this option and we could add the tip to their doc too. Christophe

1 0

Re: [boost] [Review] ITL review starts today, February 18th
by Jeff Flinn 25 Feb '10

25 Feb '10

Hartmut Kaiser wrote: > Hi all, > > The formal review of the Interval Template Library (ITL) starts today, > February 18th, 2010 and will end February 27th, 2010. ITL is being developed by Joachim Faulhaber. > > ------------------------------------------------ > > About the library: > > The Interval Template Library (ITL) is a library of interval containers that > facilitates computations on collections of intervals and associated values. ... Hmm, not much discussion yet. While this is not my review I'll make a few comments. I was able to port at least my necessary portions of split interval map to CodeWarrior 9.4 and was successful in using it to generate a column coverage graph of what is effectively the columns of possibly large sparse matrices. I would have expected the containers to be templated on an interval type, negating the need for run-time determination of boundtype. With the current design, a rather heavy interval is required with it's boundtype data member. My domain type is a simple int, an index into a row of a sparse matrix. Our in-house Interval class is always right_open, so there is only the need for a begin and end value leading to minimal memory usage. The current library design requires me to convert between ITL intervals and my own application's intervals. I'd have like to be able to provide a model of an interval concept that the ITL container would use. I wasn't able to find any description of the class invariants for interval. For example in our own Interval class the upper/end value must be greater than/equal to the lower/begin value which is a precondition to the constructor taking a pair of domain values. We do provide a static makeNormalized method to produce a proper interval when the order of the domain arguments is unknown, which is not very often in our usage so you don't pay that cost. In my case I hadn't the need for 'reversed' intervals, but I could see that being required for some use cases. Does ITL handle that? Along the same lines I've found little need for the mutators. An interval really only needs default and copy constructors, and one taking the upper and lower values. All mutations merely construct a new interval using calculated values and begin/end values from source intervals. Jeff

6 15