- Boost - lists.boost.org

[task] Minor issues
by Chard 19 Jun '09

19 Jun '09

I've been trying out the task library from the vault and have found a few issues. The code does not compile with MSVC 7.1; this appears to be a compiler non-conforming issue rather than code error: The contained impl_pool class inside detail::worker requires access to the static pool class, but even though the detail::worker class is a friend, VC7 doesn't allow impl_pool access. Adding: template< typename Pool > friend class detail::worker::impl_pool; to the class in static_pool.hpp gets around this. <> I believe there is another error in the task config.hpp file; it has the line: # if defined(BOOST_TASK_DYN_DLL) || defined(BOOST_ALL_DYN_LINK) I believe BOOST_TASK_DYN_DLL should be BOOST_TASK_DYN_LINK, following the guide from the main boost config.

2 1

[polygon] GTL name change and pre-review issues
by Simonson, Lucanus J 19 Jun '09

19 Jun '09

I am changing the name of my library from GTL (Geometry Template Library) to boost::polygon and narrowing the scope from "computational geometry" to "polygon manipulation". This scope precisely describes the current scope of what is implemented. It also clarifies the position of the library relative to similar proposals. I have already moved the code into boost::polygon namespace, added dependency on boost::enable_if and gotten the unit tests compiling and passing in windows with the new namespace and dependency. I wanted to ask about the gtl_yes and gtl_no types I use for metaprogramming, defined thus: struct gtl_no { static const bool value = false; }; struct gtl_yes { typedef gtl_yes type; static const bool value = true; }; I need to define type as gtl_yes in gtl_yes and leave it undefined in gtl_no to enable early SFINAE to work around an EDG compiler bug: template <typename geometry_type_1, typename coordinate_type_1> typename enable_if< typename gtl_and< typename is_mutable_polygon_90_set_type<geometry_type_1> #ifdef __ICC ::type #endif ::type, typename gtl_same_type<typename geometry_concept<coordinate_type_1>::type, coordinate_concept>::type> #ifdef __ICC ::type #endif ::type, geometry_type_1>::type operator-(const geometry_type_1& lvalue, coordinate_type_1 rvalue) { geometry_type_1 retval(lvalue); retval -= rvalue; return retval; } Here I am short circuiting template instantiation in the case that ICC (using EDG) is detected by calling the gtl_yes/gtl_no metafunction. This would be a syntax error in MSVC, of course. I need to do this for arithemetic operators because EDG incorrectly tries to match them to operators between pointer and unnamed enum type values in the stl and throws a syntax error when I refer to the unnamed enum type in the declaration of the operator. By SFINAEing out early I can work around this bug, which is fixed in the latest EDG versions, but I need to support older versions as well. Is it acceptable to continue to use my own definitions for true_type/false_type, or does someone have a suggestion for a different workaround that will allow me to replace gtl_yes/gtl_no with boost::true_type, boost::false_type? By the way, complete documentation for my library is now available at https://svn.boost.org/svn/boost/sandbox/gtl/doc/index.htm Feedback on the documentation before the formal review will help me get ready for the review. I'll be checking review ready code into the sandbox later today along with unit tests. If someone can point me to setup my unit test source file to work with the boost unit testing system I'd greatly appreciate it. Right now it just uses returning an error code to indicate test failure, and it is all or nothing in that sense. Regards, Luke

1 0

Proposed addition to the iterator library for iterating over all tuples.
by Alex Mendes da Costa 19 Jun '09

19 Jun '09

Hi folks: I quite often have to write code to do the following: Given N sequences, generate all the N-tuples s.t. the i'th element in each tuple is drawn from the i'th sequence. For example, if sequence S1 is { 1, 2, 3 } and sequence S2 is { 10, 20 }, I need to generate these tuples: (1, 10), (2, 10), (1, 20), (2, 20). I think this problem could be solved elegantly with a iterator similar to zip_iterator. I have some questions: (1) Is there already a way to do this in boost? (2) If not, would this be a worthwhile addition to the iterator library? (3) If so, could I contribute code for this, or must the existing library owners write it? (4) If so, how would I go about getting my code added? Most of the docs seem to be about submitting entire libraries, rather than extensions to existing libraries. Thanks, Alex

4 18

Boost.Bloom_filter now in Sandbox (WAS Re: Interest in a Bloom Filter?)
by Dean Michael Berris 19 Jun '09

19 Jun '09

On Mon, Jun 8, 2009 at 2:55 PM, Dean Michael Berris<mikhailberis(a)gmail.com> wrote: > On Mon, Jun 8, 2009 at 2:47 PM, Joaquin M Lopez Munoz<joaquin(a)tid.es> wrote: [snip] > > Thanks for your insights! I'll make the changes and then maybe put it > in the sandbox along with the documentation that I'll be working on. > The implementation is now available in Boost's Sandbox. You can check out the implementation along with the documentation via: svn co https://svn.boost.org/svn/boost/sandbox/bloom_filter/trunk/ Can someone please add this to the Libraries Under Construction page? I'd also like to request to add this to the formal review queue (if this qualifies for a fast track review, that would also be an option I'm open to). Thanks in advance, and I hope this helps. -- Dean Michael Berris | Software Engineer, Friendster, Inc. blog.cplusplus-soup.com | twitter.com/mikhailberis | linkedin.com/in/mikhailberis | profiles.friendster.com/mikhailberis | deanberris.com

9 33

[boost::noncopyable] compiler warnings with the -Weffc++ flags
by Anand Lakshminath 19 Jun '09

19 Jun '09

Hi, Deriving from the boost::noncopyable class generates warnings when I compile with the -Weffc++ (effective C++) flags. There are two types of warnings: 1) Non-virtual destructor warning (base class 'class boost::noncopyable_::noncopyable' has a non-virtual destructor). Here's a testcase: #include <boost/noncopyable.hpp> #include <iostream> namespace { class DontTreadOnMe : private boost::noncopyable { public: DontTreadOnMe() { std::cout << "defanged!" << std::endl; } }; // DontTreadOnMe } // unnamed namespace main() { DontTreadOnMe object1; } g++ -I/vobs/3rdparty/BOOST/boost_1_38_0/boost_inc -Weffc++ main.cpp main.cpp:10: warning: base class 'class boost::noncopyable_::noncopyable' has a non-virtual destructor 2) I have disabled the copy constructor/assignment operator in a class that has pointer data members by deriving from boost::noncopyable. However, the compiler generates the following warning: warning: 'class<unnamed>::DontTreadOnMe' has pointer data members main.cpp:10: warning: but does not override '<unnamed>::DontTreadOnMe(const <unnamed>::DontTreadOnMe&)' This kind of forces me to explicitly override the copy constructor and assignment operator instead of deriving from noncopyable. Here's the full test case: #include <boost/noncopyable.hpp> #include <iostream> namespace { class DontTreadOnMe : private boost::noncopyable { public: DontTreadOnMe() { std::cout << "defanged!" << std::endl; } private: char* buf; }; // DontTreadOnMe } // unnamed namespace main() { DontTreadOnMe object1; } g++ -I/vobs/3rdparty/BOOST/boost_1_38_0/boost_inc -Weffc++ main.cpp main.cpp:10: warning: base class 'class boost::noncopyable_::noncopyable' has a non-virtual destructor main.cpp:10: warning: 'class<unnamed>::DontTreadOnMe' has pointer data members main.cpp:10: warning: but does not override '<unnamed>::DontTreadOnMe(const <unnamed>::DontTreadOnMe&)' main.cpp:10: warning: or 'operator=(const <unnamed>::DontTreadOnMe&)' Is there a workaround and/or fix? I am using gcc 4.0.1. Thanks, Anand

5 11

Re: [boost] Proposal: Monotonic Containers - Comparison with boost::pool and boost::fast_pool
by Christian Schladetsch 19 Jun '09

19 Jun '09

Hello, I ran the following test http://tinyurl.com/l89llq This compares std::allocator, boost::fast_pool_allocator, boost::pool_allocator and monotonic::allocator. The only test in which monotonic allocation fails to perform faster than the other schemes is thrash_pool_sort_list_int against fast_pool. The reason for this is that fast_pool effectively stores a pre-generated array of correctly-aligned chunks, so allocation against this is a pointer increment. Fast pool can do this, because it stores different pools for different-sized blocks. Rather, the monotonic allocator uses a single buffer for all allocations, so it has to align each time. There is an optimisation I can introduce between a monotonic::allocator<T> and the underlying storage that can improve performance slightly for this case. The downside for fast_pool is that the user has to keep track of each different-sized chunk that can possiblly be used, and it doesn't differentiate on type. From the documentation: "Since the size of T is used to determine the type of the underlying Pool, each allocator for different types of the same size will share the same underlying pool. The tag class prevents pools from being shared between pool_allocator and fast_pool_allocator. For example, on a system where sizeof(int) == sizeof(void *), pool_allocator<int> and pool_allocator<void *> will both allocate/deallocate from/to the same pool." This is problematic for cases where a given allocator is rebound to a different type, as happens for the node-based containers. I am unsure what the best practise is here, or if boost::pool and boost::fast_pool simply leak in this case. You will also not that fast_pool is very slow for thrash_pool_sort, which sorts different-sized vectors. Here, the same mechanism that made it %10-%20 faster than monotonic using lists makes it many times slower when using vectors. In the following tables, the first row is the maximum number of elements in the relevant container; all internal columns are in seconds. Results for MSVC /O2. thrash_pool count fast_p pool std mono local fp/mono pool/mono 10 0.015 0.019 0.02 0.005 0.005 300% 380% 210 0.035 0.037 0.038 0.01 0.009 350% 370% 410 0.051 0.055 0.04 0.013 0.013 392.3% 423.1% 610 0.069 0.072 0.046 0.017 0.017 405.9% 423.5% 810 0.088 0.093 0.048 0.021 0.022 419% 442.9% 1010 0.105 0.111 0.053 0.026 0.028 403.8% 426.9% 1210 0.125 0.138 0.06 0.031 0.03 403.2% 445.2% 1410 0.142 0.143 0.061 0.034 0.033 417.6% 420.6% 1610 0.164 0.162 0.069 0.038 0.038 431.6% 426.3% 1810 0.179 0.184 0.072 0.043 0.042 416.3% 427.9% thrash_pool_iter count fast_p pool std mono local fp/mono pool/mono 10 0.027 0.032 0.032 0.012 0.011 225% 266.7% 210 0.152 0.146 0.118 0.102 0.101 149% 143.1% 410 0.27 0.263 0.199 0.191 0.189 141.4% 137.7% 610 0.4 0.372 0.286 0.283 0.284 141.3% 131.4% 810 0.509 0.487 0.376 0.372 0.372 136.8% 130.9% 1010 0.64 0.598 0.474 0.474 0.464 135% 126.2% 1210 0.765 0.721 0.579 0.563 0.592 135.9% 128.1% 1410 0.895 0.841 0.632 0.662 0.652 135.2% 127% 1610 0.991 0.923 0.701 0.714 0.759 138.8% 129.3% 1810 1.185 1.096 0.793 0.91 0.842 130.2% 120.4% thrash_pool_sort count fast_p pool std mono local fp/mono pool/mono 10 0.01 0.002 0.002 0.001 0.001 1000% 200% 110 0.091 0.011 0.011 0.008 0.009 1138% 137.5% 210 0.233 0.023 0.02 0.017 0.016 1371% 135.3% 310 0.495 0.034 0.026 0.025 0.025 1980% 136% 410 1.347 0.044 0.036 0.035 0.034 3849% 125.7% 510 1.411 0.058 0.045 0.042 0.042 3360% 138.1% 610 2.928 0.067 0.055 0.052 0.053 5631% 128.8% 710 2.54 0.078 0.064 0.065 0.061 3908% 120% 810 6.175 0.089 0.073 0.07 0.07 8821% 127.1% 910 3.976 0.104 0.081 0.082 0.08 4849% 126.8% thrash_pool_sort_list_int count fast_p pool std mono local fp/mono pool/mono 10 0.006 0.011 0.011 0.003 0.004 200% 366.7% 210 0.057 0.087 0.088 0.06 0.059 95% 145% 410 0.102 0.214 0.169 0.135 0.125 75.56% 158.5% 610 0.156 0.367 0.254 0.199 0.196 78.39% 184.4% 810 0.211 0.588 0.346 0.26 0.261 81.15% 226.2% 1010 0.269 0.837 0.459 0.351 0.35 76.64% 238.5% 1210 0.347 1.16 0.547 0.441 0.431 78.68% 263% 1410 0.406 1.504 0.632 0.54 0.511 75.19% 278.5% 1610 0.481 1.882 0.728 0.583 0.587 82.5% 322.8% 1810 0.559 2.32 0.817 0.671 0.674 83.31% 345.8% thrash_pool_map_list_unaligned count fast_p pool std mono local fp/mono pool/mono 10 0.002 0.003 0.004 0.001 0.001 200% 300% 210 0.033 0.083 0.079 0.029 0.029 113.8% 286.2% 410 0.067 0.214 0.164 0.058 0.061 115.5% 369% 610 0.102 0.404 0.246 0.091 0.092 112.1% 444% 810 0.139 0.646 0.338 0.122 0.126 113.9% 529.5% 1010 0.174 0.936 0.415 0.155 0.16 112.3% 603.9% 1210 0.219 1.271 0.51 0.19 0.197 115.3% 668.9% 1410 0.253 1.675 0.594 0.23 0.237 110% 728.3% 1610 0.297 2.144 0.678 0.273 0.274 108.8% 785.3% 1810 0.339 2.692 0.874 0.402 0.313 84.33% 669.7% Results for GCC -O2 thrash_pool count fast_p pool std mono local fp/mono pool/mono 10 0 0.01 0 0.01 0 0% 100% 210 0.01 0.01 0.01 0 0 inf% inf% 410 0.01 0 0.01 0.01 0 100% 0% 610 0.01 0.01 0.01 0 0 inf% inf% 810 0.01 0 0.01 0.01 0 100% 0% 1010 0 0.01 0 0 0.01 nan% inf% 1210 0.01 0.01 0.01 0 0 inf% inf% 1410 0.01 0 0.01 0 0.01 inf% nan% 1610 0 0.01 0.01 0 0 nan% inf% 1810 0 0.01 0 0 0.01 nan% inf% thrash_pool_iter count fast_p pool std mono local fp/mono pool/mono 10 0.01 0.01 0 0.01 0 100% 100% 210 0.05 0.06 0.02 0.02 0.02 250% 300% 410 0.12 0.1 0.04 0.03 0.03 400% 333.3% 610 0.16 0.15 0.05 0.05 0.04 320% 300% 810 0.2 0.21 0.06 0.06 0.07 333.3% 350% 1010 0.25 0.25 0.07 0.08 0.08 312.5% 312.5% 1210 0.29 0.31 0.08 0.08 0.09 362.5% 387.5% 1410 0.33 0.35 0.1 0.08 0.1 412.5% 437.5% 1610 0.38 0.37 0.11 0.12 0.12 316.7% 308.3% 1810 0.43 0.45 0.12 0.13 0.13 330.8% 346.2% thrash_pool_sort count fast_p pool std mono local fp/mono pool/mono 10 0.01 0.01 0 0 0 inf% inf% 110 0.06 0.01 0.01 0.01 0.01 600% 100% 210 0.19 0.02 0.02 0.02 0.01 950% 100% 310 0.47 0.02 0.02 0.02 0.02 2350% 100% 410 0.69 0.04 0.04 0.03 0.02 2300% 133.3% 510 1.99 0.05 0.04 0.03 0.04 6633% 166.7% 610 1.65 0.06 0.04 0.04 0.04 4125% 150% 710 4.73 0.07 0.05 0.04 0.05 1.182e+04% 175% 810 2.28 0.07 0.07 0.06 0.06 3800% 116.7% 910 4.29 0.09 0.05 0.08 0.06 5362% 112.5% thrash_pool_sort_list_int count fast_p pool std mono local fp/mono pool/mono 10 0 0.01 0 0 0 nan% inf% 210 0.03 0.05 0.04 0.04 0.03 75% 125% 410 0.07 0.15 0.1 0.08 0.06 87.5% 187.5% 610 0.1 0.26 0.12 0.11 0.12 90.91% 236.4% 810 0.13 0.39 0.19 0.18 0.14 72.22% 216.7% 1010 0.17 0.57 0.24 0.19 0.2 89.47% 300% 1210 0.24 0.74 0.29 0.23 0.24 104.3% 321.7% 1410 0.26 0.98 0.36 0.26 0.28 100% 376.9% 1610 0.27 1.21 0.37 0.3 0.33 90% 403.3% 1810 0.35 1.51 0.43 0.35 0.37 100% 431.4% thrash_pool_map_list_unaligned count fast_p pool std mono local fp/mono pool/mono 10 0 0 0 0 0 nan% nan% 210 0.02 0.05 0.03 0.02 0.02 100% 250% 410 0.02 0.11 0.06 0.04 0.04 50% 275% 610 0.06 0.19 0.11 0.07 0.07 85.71% 271.4% 810 0.06 0.33 0.13 0.09 0.08 66.67% 366.7% 1010 0.09 0.51 0.17 0.11 0.11 81.82% 463.6% 1210 0.12 0.65 0.21 0.13 0.14 92.31% 500% 1410 0.15 0.93 0.25 0.14 0.14 107.1% 664.3% 1610 0.17 1.11 0.27 0.19 0.15 89.47% 584.2% 1810 0.18 1.35 0.33 0.21 0.21 85.71% 642.9% Regards, Christian

1 1

From Concept class to type traits
by joel 19 Jun '09

19 Jun '09

Is there any deep explanation on how the concept_check class from Boost:/Concept works ? I'm trying to see how to turn a concept into a traits but all my attempt are failing miserably and I think that's because I don't get how the concept check work in itself. Regards -- ___________________________________________ Joel Falcou - Assistant Professor PARALL Team - LRI - Universite Paris Sud XI Tel : (+33)1 69 15 66 35

5 6

Fwd: bug fix in boost/libs/python/pyste/src/Pyste/declarations.py
by YuQian Zhou (周宇前) 19 Jun '09

19 Jun '09

Hi, I was unable to reach Dave, and https://svn.boost.org/trac/boost/newticket doesn't work for me. ("Secure Connection Failed"). So I'm posting here, can anyone take a look this? thanks. ---------- Forwarded message ---------- From: YuQian Zhou (周宇前) <zhou(a)google.com> Date: 2009/6/15 Subject: Fwd: bug fix in boost/libs/python/pyste/src/Pyste/declarations.py To: dave(a)boost-consulting.com Hi Dave, Can you tell me where I should submit this change? Thanks. ---------- Forwarded message ---------- From: Bruno Oliveira <nicoddemus(a)gmail.com> Date: 2009/6/15 Subject: Re: bug fix in boost/libs/python/pyste/src/Pyste/declarations.py To: "YuQian Zhou (周宇前)" <zhou(a)google.com> Hi YuQian, Unfortunately I don't maintain Pyste anymore; I advise you to send this fix to the boost-python list, since I don't have commit privileges anymore (I think). Best Regards, On Mon, Jun 15, 2009 at 5:32 PM, YuQian Zhou (周宇前) <zhou(a)google.com> wrote: > > Hi Bruno, > > I found a bug in boost/libs/python/pyste/src/Pyste/declarations.py (I > tried the latest version (Version 1.39.0) from boost.org, it has the > same bug.) > > The bug is: when paring an anonymous struct, the name passed to a > Declaration object is None, then on > libs/python/pyste/src/Pyste/GCCXMLParser.py: line 57 > > def AddDecl(self, decl): > if decl.FullName() in self._names: > > The call to FullName() will concatenate name, line 53: > > return namespace + self.name > > TypeError: cannot concatenate 'str' and 'NoneType' objects > > I made a fix below, please let me know if it's correct, and you can > include in the release. > > Thanks. > > YuQian > > Below is a test file: > > $ cat test.h > > typedef int __jmp_buf; > > typedef struct > { > struct > { > __jmp_buf __cancel_jmp_buf; > int __mask_was_saved; > } __cancel_jmp_buf[1]; > void *__pad[4]; > } __pthread_unwind_buf_t __attribute__ ((__aligned__)); > > > here is the diff of libs/python/pyste/src/Pyste/declarations.py > > 16d15 > < _anonymous_name_counter = 0 > 33,36d31 > < global _anonymous_name_counter > < if name is None: > < name = 'anonymous_name_%s' % _anonymous_name_counter > < _anonymous_name_counter += 1

1 0

Request: boost deprecated tag
by Chard 18 Jun '09

18 Jun '09

Given that there is a current move to rationalize the bind placeholders, and potentially introduce deprecated definitions, would it worth introducing a boost deprecated marker that makes use of compiler support to notify the user that they are using such definitions? I recently recompiled a project with BOOST_FILESYSTEM_NO_DEPRECATED (for no particular reason, except perhaps for tr1 readiness) and I was surprised how many uses of the old native_file_string and boost::native tags were buried in the code-base. Personally, I think it would be useful to be notified (or nagged :-)) if I was using deprecated functionality; I don't examine every comment of every function, and after successful compilation/unit-test they are easily forgotten. I don't know the best, generic, way of achieving this, but a simple approach that should work with MSVC and gcc would be to have: #ifndef BOOST_NO_DEPRECATED_WARNINGS // Don't nag me, compiler # if [BUILDING_WITH_MSVC] # define BOOST_DEPRECATED __declspec(deprecated) # else if [BUILDING_WITH_GCC] # define BOOST_DEPRECATED __attribute__ ((deprecated)) # else if [BUILDING_WITH_acme] # define BOOST_DEPRECATED ... # end if #endif #ifndef BOOST_DEPRECATED // No compiler support, just a tag # define BOOST_DEPRECATED #end if Then usage is: BOOST_DEPRECATED void TheFunction(...); The intent here is clear, and uncluttered, but may not be flexible enough. For example, a compiler may use syntax that declares its deprecated tag after the definition (which gcc supports, and it is usually shown that way in the guidelines). In which case, I expect the whole definition would have to be wrapped up in the macro. Also, there may be a case for separating the deprecated tags for functions, variables, types, etc.. As an (untested) illustration: #ifndef BOOST_NO_DEPRECATED_WARNINGS # if [BUILDING_WITH_MSVC] # define BOOST_DEPRECATED_FUNCTION(_F_) __declspec(deprecated) _F_ # else if [BUILDING_WITH_GCC] # define BOOST_DEPRECATED_FUNCTION(_F_) _F_ __attribute__ ((deprecated)) # else if [BUILDING_WITH_acme] # define BOOST_DEPRECATED_FUNCTION(_F_) ... # define BOOST_DEPRECATED_VARIABLE(_V_) ... # define BOOST_DEPRECATED_TYPE(_T_) ... # end if #endif #ifndef BOOST_DEPRECATED_FUNCTION # define BOOST_DEPRECATED_FUNCTION(_F_) _F_ #end if #ifndef BOOST_DEPRECATED_VARIABLE // Some compilers may use different syntax for variables, if not, // use same as deprecated function tag # define BOOST_DEPRECATED_VARIABLE(_V_) BOOST_DEPRECATED_FUNCTION(_V_) #endif #ifndef BOOST_DEPRECATED_TYPE // As for variables.. # define BOOST_DEPRECATED_TYPE(_T_) BOOST_DEPRECATED_FUNCTION(_T_) #endif Usage example: class path { ... BOOST_DEPRECATED_FUNCTION(string_type native_file_string() const) { return file_string(); } ... }; This is noisier, and may be worse if parentheses/comma expansion work-arounds are required, but it's an example. Personally, I prefer the simple BOOST_DEPRECATED, but I don't have experience with other compilers to know if that would suffice. Even if no support is available, having a benign tag in the definition provides information at the least. Regards...

2 2

Boost Trac/Subversion upgrade this Friday
by Doug Gregor 18 Jun '09

18 Jun '09

The Boost Trac and Subversion repository will be offline for a few hours this Friday, June 19th, while we upgrade to a new version of Trac. Mailing lists will not be affected. Here is the approximate down-time: Date: Friday June 19, 2009 Time: - 11:00am-2:00pm Pacific US time - 12:00pm-3:00pm Mountain US time - 1:00pm-4:00pm Central US time - 2:00pm-5:00am Eastern US time - 6:00pm-9:00pm GMT Cheers, - Doug

2 1