January 2012 - Boost - lists.preview.boost.org

[scope_exit] improvements
by Lorenzo Caminiti 15 Jan '12

15 Jan '12

Hello all, I've been working with Alex to make the following improvements to Boost.ScopeExit (as they were discussed during Boost.Local's review). The update docs: https://svn.boost.org/svn/boost/sandbox/closure/libs/scope_exit/doc/html/in… 1) It is possible to use variadic macros on compilers that support them: int r = 0, v = 0; BOOST_SCOPE_EXIT(&r, v) { r = v; } BOOST_SCOPE_EXIT_END To support compilers without variadics and for backward compatibility with previous ScopeExit versions, all macros also accepts pp-sequences: BOOST_SCOPE_EXIT( (&r) (v) ) { r = v; } BOOST_SCOPE_EXIT_END 2) It is possible to capture no variable using `void`: BOOST_SCOPE_EXIT(void) { std::cout << "cout is a global variable << std::endl; } BOOST_SCOPE_EXIT_END This same syntax works on compilers with and without variadics. 3) From within member functions, it is possible to capture the object this using `this_`: BOOST_SCOPE_EXIT(&r, v, this_) { r = this_->f(v); } BOOST_SCOPE_EXIT_END On C++11, you can also (but you don't have to) use `this`: BOOST_SCOPE_EXIT(&r, v, this) { // C++11 only. r = this->f(v); } BOOST_SCOPE_EXIT_END 4) On C++11, you can (but you don't have to) use `;` instead of BOOST_SCOPE_END. 5) On C++11, there is an additional macro that allows to capture all variables using lambda-style `&` or `=`: BOOST_SCOPE_EXIT_ALL(&, v, this) { // C++11 only. r = this->f(v); }; Or equivalently: BOOST_SCOPE_EXIT_ALL(=, &r) { // C++11 only. r = this->f(v); }; I still have to move these changes from the sandbox into trunk. Comments are welcome. Thanks, --Lorenzo

2 1

[typeof] TYPEOF(this) on MSVC 7.1 and 8.0
by Lorenzo Caminiti 14 Jan '12

14 Jan '12

Hello all, BOOST_TYPEOF(this) generates error C2355 on MSVC 7.1 and 8.0 because of a MSVC bug. The code below was suggested to workaround the problem but it only works with native typeof. Is there a way to make BOOST_TYPEOF(this) work on MSVC 7.1 and 8.0 with Boost.Typeof in emulation mode? // Adam Butcher's workaround to deduce `this` type on MSVC revision < 10. // Boost.Typeof for VC71's typeid-based workaround does not work to determine // `this` type due to error C2355 being incorrectly reported. The typical // avoidance strategy implemented below is to make an indirect compile-time // constant by assigning an enum and use that as type-index-- this only works // with the sizeof() approach and not with the typeid() approach. This does not // work in typeof emulation mode. #if BOOST_WORKAROUND(BOOST_MSVC, >= 1300) && /* MSVC 7.1 */ \ BOOST_WORKAROUND(BOOST_MSVC, <= 1400) && /* MSVC 8.0 */ \ !defined(BOOST_TYPEOF_EMULATION) /* doesn't work in typeof-emu mode */ namespace boost { namespace scope_exit { namespace aux { namespace msvc_typeof_this { template<int Id> struct msvc_typeid_wrapper { typedef typename boost::type_of::msvc_extract_type<boost::mpl::int_<Id> >::id2type id2type; typedef typename id2type::type type; }; template<> struct msvc_typeid_wrapper<4> { typedef msvc_typeid_wrapper<4> type; }; template<typename T> struct encode_type { BOOST_STATIC_CONSTANT(unsigned, value = boost::type_of::encode_counter<T>::count); typedef typename boost::type_of::msvc_register_type<T, boost::mpl::int_<value> >::id2type type; BOOST_STATIC_CONSTANT(unsigned, next = value + 1); }; template<class T> struct sizer { typedef char(*type)[encode_type<T>::value]; }; template<typename T> typename boost::disable_if< typename boost::is_function<T>::type , typename sizer<T>::type >::type encode_start(T const&); template<typename T> typename boost::enable_if< typename boost::is_function<T>::type , typename sizer<T>::type >::type encode_start(T&); template<typename Organizer, typename T> boost::type_of::msvc_register_type<T, Organizer> typeof_register_type(const T&, Organizer* = 0); } } } } #define BOOST_SCOPE_EXIT_AUX_TYPEDEF_TYPEOF_THIS_INDEX_(id) \ BOOST_PP_CAT(boost_se_thistype_index_, id) #define BOOST_SCOPE_EXIT_DETAIL_TYPEDEF_TYPEOF_THIS(id, new_type) \ enum { \ BOOST_SCOPE_EXIT_AUX_TYPEDEF_TYPEOF_THIS_INDEX_(id) = sizeof( \ *::boost::scope_exit::aux::msvc_typeof_this::encode_start(this)) \ }; \ typedef ::boost::scope_exit::aux::msvc_typeof_this::msvc_typeid_wrapper< \ BOOST_SCOPE_EXIT_AUX_TYPEDEF_TYPEOF_THIS_INDEX_(id) \ >::type new_type; #else // typeof(this) on other compilers are OK. #define BOOST_SCOPE_EXIT_DETAIL_TYPEDEF_TYPEOF_THIS(id, new_type) \ typedef BOOST_TYPEOF(this) new_type; #endif // typeof(this) For example: // within a member function BOOST_SCOPE_EXIT_DETAIL_TYPEDEF_TYPEOF_THIS(__LINE__, this_type); this_type obj = this; Thanks. --Lorenzo

1 0

[context] mini-review comments
by Jeffrey Lee Hellrung, Jr. 14 Jan '12

14 Jan '12

[Late, I know] [I'm going off http://ok73.ok.funpic.de/boost/libs/context/doc/html/index.html] [Apologies if these are just repeating previous comments!] How do you store the function and function arguments passed to the context constructor? Are they bundled into memory allocated by the stack allocator? I couldn't find it in the documentation, and whatever the answer, seems like something that should be mentioned. "The maximum number of arguments of *context-function* is defined by BOOST_CONTEXT_ARITY (default is 10) and might be changed." Hopefully you mean, more specifically, that BOOST_CONTEXT_ARITY might be *increased*. Also, consider using variadic templates if available? Great examples! Would it be possible to also show the expected output, just so the reader can verify that he or she understands what should be happening? Given that void pointers are used to communicate between contexts, I almost would've expected the context constructor to simply take a void (*)( void* ) argument and context::start to take a void* argument which would be passed to the function pointer. Can you elaborate on the rationale for the low-level interface for communicating between contexts while still providing a relatively high level interface for specifying the context function? The documentation for the stack allocator concept doesn't completely define the semantics of the size parameter of stack_allocator::allocate/deallocate. I presume it is the stack size in bytes, but that should be explicit in the documentation. And I presume allocate is only required to return a pointer to *at least* size bytes of free memory. Also, you don't mention alignment requirements; what are they? I'm somewhat surprised default_stacksize() always returns 256*1024. I think it might be safer to make this implementation-defined, but some minimum size (as with everything else in C++...). Also, strictly speaking, the documentation is incorrect. It reads "Defines a default stack size of 256 kB", but defalt_stacksize() doesn't define anything, it just returns a number. And this description is inconsistent with the remaining function descriptions. I would do: "Returns the default stack size, which is implementation-defined but at least 256 kB." In "Low level API (boost_fcontext_t)", s/stated/started/. Also, what's the magic number 262144? Oh, I see, 262144 = 256 * 1024...maybe the latter expression would be more intuitive? Again, I can guess what the output of the example would be, but would be helpful to have the expected output handy. Just a nitpick on member variable names. boost_fcontext_stack::ss_base/ss_limit: what does "ss" mean? boost_fcontext::fc_stack/fc_link: I presume the prefix "fc" is short for "fcontext", in which case, isn't it redundant? Normally I wouldn't be picky about member variable names but these are precisely the documented API. In "Rationale", you give "[o]ther low level APIs and their caveats"; I'm guessing these are, specifically, alternatives to boost_fcontext_t but are *deficient* in some respect. It might clear things up if you make that explicit, and further group "setjump()/longjmp()", "ucontext_t", and "Windows fibers" into a common heading. Overall, I think the documentation and interface are good. I didn't look at the implementation, but I did see that someone else did, so...I'm not too worried about a correct implementation. As long as there's a full suite of tests. I also didn't try to use the library, but now I'll have to find an excuse :) These are all relatively minor comments, and although I hope Oliver will address them, it won't hold up my approval of Boost.Context. - Jeff

3 7

[fusion] Getting the adapted structure's member names
by Andy Venikov 14 Jan '12

14 Jan '12

struct alien { std::string color_; std::string planet_; bool benevolent_; }; BOOST_FUSION_ADAPT_STRUCT( alien, (std::string, color_) (std::string, planet_) (bool, benevolent_) ) I know how I can get to the actual member values, i.e. I could check at<2>(anAlien) to see if I should run away or not. But what if I wanted to get to the member names? (In this example it would be "color_", "planet_" and "benevolent_") It could be very useful for example for serialization that stores not just values but also the names of the values. Printing structures to stdout for inspection would be a breeze too. Since BOOST_FUSION_ADAPT_STRUCT is a macro, surely the member names could be strigized and stored somewhere. Thanks, Andy.

2 1

__COUNTER__ vs __LINE__
by lcaminiti 13 Jan '12

13 Jan '12

Hello all, Many of Boost macro APIs PP_CAT symbols with either __LINE__ or __COUNTER__ to make them unique (MPL_STATIC_ASSERT, SCOPE_EXIT, etc). AFAIK, __LINE__ is C++03 standard while __COUNTER__ is only a MSVC extension. I think using __COUNTER__ instead of __LINE__ might allow to expand a macro multiple times on the same line... but a part from that: What's the reason for using __COUNTER__ on MSVC instead of just using __LINE__ for all compilers? Thanks a lot. --Lorenzo -- View this message in context: http://boost.2283326.n4.nabble.com/boost-COUNTER-vs-LINE-tp4286398p4286398.… Sent from the Boost - Dev mailing list archive at Nabble.com.

6 6

[Boost.Move] A few notes
by Dan Ivy 13 Jan '12

13 Jan '12

Hi, I was recently trying out Boost.Move and a few issues worth sharing surfaced: 1. It would be helpful to have configuration macros to force emulation mode, even on C++11 compilers, as well as to disable move semantics altogether (that is, the conversion operators to boost::rv& should be disabled, and move/forward should return lvalue-references. BOOST_RV_REF and friends should remain intact, so that overloads remain unique.) In many cases, the higher level semantics of a program are expected to be identical under all three configurations, so having a quick way to switch between them is useful during testing/debugging. 2. Boost.Move is a little bit too opaque, as it stands. What's really missing are Boost.Move-aware type traits. Things like add_rvalue_reference are often necessary to calcuate return types of move-aware generic functions, and so on. Whether this belongs in Boost.TypeTraits or Boost.Move is a separate question. Likewise, there should be type traits to calculate the return types of boost::move and, in particular, boost::forward. On C++11, the return type of forward<T> conincides with add_rvalue_reference<T>, but not so in emulation mode, hence the necessity for this trait. 3. For some reason, the emulated boost::move is written so it doesn't accept temporaries. This doesn't play too nicely with forwarding: Unavoidably on C++03, there has to be made a choice whether forwarding functions use non-const or const-qualified references (bad wording, but you get the point), where the former accepts modifiable lvalues and rejects temporaries, and the latter does the opposite (well, sort of). Boost.Move chooses the latter. This means that while BOOST_FWD_REF parameters accept temporaries, these temporaries are bound as lvalues, which pretty much defeats the purpose of the library (note that these temporaries aren't eligible for copy-elision, since they're bound to a reference.) While there's not much you can do about it on the callee-side, callers could use boost::move on the temporaries to force them into rvalue-refs. Only that they can't. This creates a somewhat confusing situation where forwarding functions are able to take temporaries, but they won't move them, while the only things that can actually be moved are lvalues... Ultimately, this is what I suggest: First of all, have boost::move accept temporaries. Secondly, seriously consider changing the definition of BOOST_FWD_REF(T) from const T& to T&, as this would acheive two desirable, IMO, goals:: a. Forwarding functions would accept modifiable lvalues (and keep them modifiable). b. Forwarding functions would REJECT temporaries, UNLESS they're passed through boost::move, which assures that they're treated as rvalues. A few notes on the implementation: 4. There are many one-liner functions floating around that aren't declared inline. This is a big-deal for less-capable compilers, such as Sun and older versions of GCC. In fact, they simply won't inline boost::move calls without it. 5. boost::rv<T> unconditionally inherits from T, with the assumption that it would never get instanciated for non-class types, since it is only ever used as a reference. This is a false assumption in general. In the context of overload-resolution, the compiler is allowed, though not required, to instanciate the types of function parameters, even if the best overload can be determined without doing this. Consider this move-aware, but not very useful vector class: #include <boost/move/move.hpp> template <typename T> struct vector { void push_back(const T&) {} void push_back(BOOST_RV_REF(T)) {} }; int main() { vector<int> v; v.push_back(123); } The last line of main could potentially instanciate rv<int>, breaking the program. And it does. Sun falls right into this trap. It doesn't take much to make it break on GCC either, for the same reason. Fortunately, the fix is trivial: specialize boost::rv correctly for non-class types. All comments refer to the 1.48 release. I don't know what's going on on trunk. HTH

5 14

Release 1.49
by Barend Gehrels 13 Jan '12

13 Jan '12

Hi, I did not see (m)any recent mail traffic on the list w.r.t. release 1.49. According to the Calendar it was to be merged last Monday (major changes) and next Monday (all changes). Is this planning still up-to-date? Thanks, Barend

2 2

[interprocess] Crash of singleton_module in Windows
by JOAQUIN M. LOPEZ MUÑOZ 13 Jan '12

13 Jan '12

Steps to reproduce the problem: * Create singleton_dll.dll. * Create singleton_exe.exe, which uses singleton_dll.dll. * Run. On exit, an access violation happens. Seemingly, the problem happens when two different singletons are created, one in the main executable and the other in a DLL. If both singletons are created in the same module everything's OK, though. Best regards, Joaquín M López Muñoz Telefónica Digital ________________________________ Este mensaje se dirige exclusivamente a su destinatario. Puede consultar nuestra política de envío y recepción de correo electrónico en el enlace situado más abajo. This message is intended exclusively for its addressee. We only send and receive email on the basis of the terms set out at. http://www.tid.es/ES/PAGINAS/disclaimer.aspx

4 7

[bjam] if MSVC 7.1 or 8.0 compile-fail
by lcaminiti 13 Jan '12

13 Jan '12

Hello all, If there a way to program the following within a Jamfile? import testing ; if(MSVC 7.1 or MSVC 8.0) { # How do I program this? [ compile-fail native_tu_test.cpp native_tu1.cpp native_tu2.cpp : : : <define>BOOST_TYPEOF_NATIVE ] } else { [ run native_tu_test.cpp native_tu1.cpp native_tu2.cpp : : : <define>BOOST_TYPEOF_NATIVE ] } Thanks a lot. --Lorenzo -- View this message in context: http://boost.2283326.n4.nabble.com/boost-bjam-if-MSVC-7-1-or-8-0-compile-fa… Sent from the Boost - Dev mailing list archive at Nabble.com.

2 1

[Boost.Variant] Parameterized Visitors
by Andy Jost 13 Jan '12

13 Jan '12

New poster. When I'm using Boost.Variant, I often want to pass additional parameters to a visitor. A common case is when the variant is a class member and I want to have access to other members (or even "this") in the visitor. I muddled through for a long time writing visitors with explicit constructors and data members to take whichever parameters I needed, but I finally got around to making this generic with Boost.Fusion and I love it. I'm wondering whether I should submit this for consideration as an addition to the Boost.Variant library and how. First a few questions: 1. Is there already an equivalent solution that I should be aware of? 2. Should this be considered for an addition to Boost? 3. If so, how do I begin the review process? 4. This is a small change (400 or so lines of code). Is there anyone willing to help with the logistics, testing, and/or documentation (perhaps in exchange for a credit in the docs)? Here's an example, which should be fairly self-explanatory: struct A {}; struct B {}; struct C {}; struct my_visitor1 : param_visitor<my_visitor1, bool> { result_type operator()(int, float, char, A) const {...} result_type operator()(int, float, char, B) const {...} result_type operator()(int, float, char, C) const {...} }; struct my_visitor2 : fused_param_visitor<my_visitor2, bool> { template<typename Sequence> result_type operator()(Sequence const &, A) const {...} ... }; int main() { boost::variant<A,B,C> abc; boost::apply_visitor(my_visitor1().tie(1, 3.14, 'k'), abc); boost::apply_visitor(my_visitor2().tie(1, 3.14, 'k'), abc); return 0; } In terms of technical detail, there's not much to it. The base classes provide tie() and load() methods to take arguments by reference or value, respectively. A call to one of those returns a helper object that contains a copy of the derived visitor and a fusion::vector of arguments. Visiting that causes the user-defined visitor to be called with all of the arguments. For parameter forwarding, I use an implementation of perfect forwarding based on boost/flyweight/detail/dyn_perfect_fwd.hpp (I used a default limit of 4 parameters). Thanks, Andy

2 2