June 2009 - Boost - lists.boost.org

[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

Getting involved in Boost
by Anteru 18 Jun '09

18 Jun '09

Hi, I'm a Boost user since quite some time now (1.33 or so), however, I did not contribute a library to Boost, and I was also not too active on the news groups (a few posts so far, at most.) When I look at the review queue, I can see lots and lots of interesting libraries waiting. So my question is: How can I, primary a Boost user, contribute to the community to help with getting new libraries into Boost? Primary, I'd be interested in PIMPL and the ThreadPool, possibly in AutoBuffer and Sorting. Is there some recommended way of reviewing a library "out-of-order" with no review manager, such that the review is easily identifiable and can be found when it comes to the real review? For me, it seems to be no problem to gather reviews of a library up-front. This could make it easier for a review manager to do his work, as he could simply start a formal review only after a certain number of reviews has been done. Side note: I couldn't find a recent review status on the website -- under Development, Submissions, the latest report is from 2008, while the one latest one I could find is from June 2009. However, it does not say what the current status of the UUID is -- did I miss some obvious page with the status of all libraries? Cheers, Anteru

2 1

some mpl programming task -- don't get this compiled
by Peter Foelsche 18 Jun '09

18 Jun '09

I'm still unclear about some fundamentals of the mpl library. Could somebody have a look at this source code? I'm playing around with this since quite a while. I'm trying to convert some information into an array of strings. I'm using boost::mpl::map instead of vector as I want to overwrite points in a two dimensional matrix. So in fact this is a map indexed by a long containing another map indexed by a long containing char. Two dimensional quadratic matrix of char. Thanks Peter

2 5

[date_time] [#1861] Change for the default duration format
by Andrey Semashev 18 Jun '09

18 Jun '09

Hi, There's a Boost.DateTime ticket: https://svn.boost.org/trac/boost/ticket/1861 The initial problem was that the default handling of the %H format specifier (passing it to strftime) played badly with time durations that could be longer than 24 hours. The resolution was to provide additional %O format specifier that does not have 24 hour restriction. However the original poster proposes to change the default format of time durations from %H:%M:%S%F to %O:%M:%S%F so that formatted durations are always allowed to be longer than 24 hours. I mostly agree with this change, however it is clearly a breaking change that can affect error checking in some cases. What do the community and the maintainer (Jeff?) in particular think about this?

4 15

Proposal: MapReduce library (single machine)
by Craig Henderson 17 Jun '09

17 Jun '09

MapReduce is a programming model from Google that is designed for scalable data processing. Google's implementation is for scalability over many thousands of commodity machines, but there is value in using the idiom on multi-core processors to partition processing to efficiently use the CPU available, and avoid the complexities of multi-threaded development. I have a implemented a MapReduce runtime library in C++ using Boost and would like to see if there is interest from you all to submit the library to Boost. I have preliminary documentation on my website at http://www.craighenderson.co.uk/mapreduce/ Regards -- Craig Craig Henderson <http://craighenderson.co.uk/> http://www.craighenderson.co.uk http://www.siteupdatenotification.com

6 11