November 2008 - Boost - lists.boost.org

Why doesn't boost::exception derive from std::exception?
by Michael Marcin 07 Nov '08

07 Nov '08

Why doesn't boost::exception derive from std::exception? -- Michael Marcin

2 1

[1.37.0] boost test crashes under cygwin gcc-4.x
by Mat Marcus 06 Nov '08

06 Nov '08

When I run a simple test application with a single (empty) BOOST_AUTO_TEST_CASE, boost test causes a crash under cygwin using gcc-4.x. It also crashes for non-empty auto test cases. For example, the boost test regressions crash: ### user-config.jam contains: using gcc : 4.3.2 : /bin/g++-4.exe ; ## cygwin gcc 4.x compiler $ cd /asl/staff/mmarcus/sandbox/boost_libraries/libs/test $ ../../tools/jam/src/bin.cygwinx86/bjam.exe gcc-4.3.2 link=static -q gives: $ ../../tools/jam/src/bin.cygwinx86/bjam.exe gcc-4.3.2 link=static -q ...patience... ...found 1528 targets... ...updating 341 targets... gcc.link ../../bin.v2/libs/test/example/unit_test_example_11.test/gcc-4.3.2/debug/link-static/unit_test_example_11 testing.capture-output ../../bin.v2/libs/test/example/unit_test_example_11.test/gcc-4.3.2/debug/link-static/unit_test_example_11.run ====== BEGIN OUTPUT ====== EXIT STATUS: 53 ====== END OUTPUT ====== (failed-as-expected) ../../bin.v2/libs/test/example/unit_test_example_11.test/gcc-4.3.2/debug/link-static/unit_test_example_11.run **passed** ../../bin.v2/libs/test/example/unit_test_example_11.test/gcc-4.3.2/debug/link-static/unit_test_example_11.test gcc.link ../../bin.v2/libs/test/example/unit_test_example_12.test/gcc-4.3.2/debug/link-static/unit_test_example_12 **passed** ../../bin.v2/libs/test/example/unit_test_example_12.test/gcc-4.3.2/debug/link-static/unit_test_example_12.test gcc.link ../../bin.v2/libs/test/example/est_example1.test/gcc-4.3.2/debug/link-static/est_example1 testing.capture-output ../../bin.v2/libs/test/example/est_example1.test/gcc-4.3.2/debug/link-static/est_example1.run ====== BEGIN OUTPUT ====== EXIT STATUS: 53 ====== END OUTPUT ====== PATH=//bin://lib://lib32://lib64:$PATH export PATH "../../bin.v2/libs/test/example/est_example1.test/gcc-4.3.2/debug/link-static/est_example1" > "../../bin.v2/libs/test/example/est_example1.test/gcc-4.3.2/debug/link-static/est_example1.output" 2>&1 status=$? echo >> "../../bin.v2/libs/test/example/est_example1.test/gcc-4.3.2/debug/link-static/est_example1.output" echo EXIT STATUS: $status >> "../../bin.v2/libs/test/example/est_example1.test/gcc-4.3.2/debug/link-static/est_example1.output" if test $status -eq 0 ; then cp "../../bin.v2/libs/test/example/est_example1.test/gcc-4.3.2/debug/link-static/est_example1.output" "../../bin.v2/libs/test/example/est_example1.test/gcc-4.3.2/debug/link-static/est_example1.run" fi verbose=0 if test $status -ne 0 ; then verbose=1 fi if test $verbose -eq 1 ; then echo ====== BEGIN OUTPUT ====== cat "../../bin.v2/libs/test/example/est_example1.test/gcc-4.3.2/debug/link-static/est_example1.output" echo ====== END OUTPUT ====== fi exit $status ...failed testing.capture-output ../../bin.v2/libs/test/example/est_example1.test/gcc-4.3.2/debug/link-static/est_example1.run... ...failed updating 1 target... ...updated 7 targets... ### Similar problems occur with when I use a manually-compiled gcc-4.3.0. In that case, my output for an empty auto test case looks like: ====== BEGIN OUTPUT ====== Running 1 test case... *** No errors detected 2 [main] any_regular_test 66836 _cygtls::handle_exceptions: Exception: STATUS_ACCESS_VIOLATION 413574 [main] any_regular_test 66836 open_stackdumpfile: Dumping stack trace to any_regular_test.exe.stackdump EXIT STATUS: 35584 ====== END OUTPUT ====== Stepping through in gcc shows that things run fine through the end of main, but then a problem occurs at static destruction time. My gdb skills are not sufficient to give much more info. These problems are new to 1.37.0--I've been using the unit tests successfully up through 1.36.0 (with the alt stack patch). The stack dump file for the manually-compiled gcc-4.3.0 follows. Can someone offer a suggestion? Thanks, Mat ### Exception: STATUS_ACCESS_VIOLATION at eip=610C38B4 eax=00000001 ebx=00000001 ecx=61108B2C edx=00BE0000 esi=61102F64 edi=FFFFFFF9 ebp=0022CAC8 esp=0022CA40 program=c:\asl\staff\mmarcus\sandbox\built_artifacts\test\unit_tests\any_regular\any_regular_test.test\gcc-4.3.0\debug\architecture-x86\asl-dev\instruction-set-i486\link-static\target-os-cygwin\threadapi-pthread\threading-multi\any_regular_test.exe, pid 66836, thread main cs=001B ds=0023 es=0023 fs=003B gs=0000 ss=0023 Stack trace: Frame Function Args 0022CAC8 610C38B4 (004742A0, 00BE2BE0, 00000001, 004311E0) 0022CB58 00428850 (004742A0, 61102EDC, 0022CB78, 610046C2) 0022CB78 0040682F (004737DC, 61004690, 00000000, 0022CBE0) 0022CBA8 610F7DD3 (00000000, 00000000, 0022CBD8, 00000000) 0022CBC8 610E1558 (00000000, 00000000, 0022CD98, 610060E0) 0022CBD8 61004DD1 (00000000, 00BE2580, 00BE0090, 0022CB60) 0022CD98 610060E0 (00000000, 0022CDD0, 61005450, 0022CDD0) 61005450 61004416 (0000009C, A02404C7, E8611021, FFFFFF48) 2046399 [main] any_regular_test 66836 _cygtls::handle_exceptions: Exception: STATUS_ACCESS_VIOLATION 2114906 [main] any_regular_test 66836 _cygtls::handle_exceptions: Error while dumping state (probably corrupted stack)

5 7

[serialization] load-only or save-only types?
by David Abrahams 06 Nov '08

06 Nov '08

Hi, I want to send out instances of one type and receive the same data format but build instances of a different type. Obviously I can just define serialize() for each of them, but can I define only save for the outgoing type and load for the incoming one, causing serialization in the other direction to be a compile-time error? Not a big deal, but it would be more expressive that way. TIA, -- Dave Abrahams BoostPro Computing http://www.boostpro.com

3 5

make_shared and allocate_shared for classes with private constructors
by Michael Marcin 06 Nov '08

06 Nov '08

I have some classes that derive from enable_shared_from_this which I've given private constructors and static creation methods which return a shared_ptr. This prevents someone from using the class inappropriately. It seems obvious that at a minimum using make_shared internally here would be an improvement. Unfortunately this requires that make_shared be made a friend for every arity and probably won't work with the variadic template implementation of make_shared. Is it possible to make declaring friendship for make_shared and allocate_shared simpler? I thinking something like serialization does with friend class boost::serialization::access. Thanks, -- Michael Marcin

2 3

location for PDF docs?
by Eric Niebler 06 Nov '08

06 Nov '08

I can't remember if we ever settled on a location for PDF documentation. Did we? And should we make it official by linking to it from boost.org? FWIW, I've uploaded a PDF version of Proto's docs to: http://www.boostpro.com/vault/index.php?directory=PDF%20Documentation Aside: Proto's reference section for 1.37 is /much/ improved. -- Eric Niebler BoostPro Computing http://www.boostpro.com

6 28

[FileSystem] portable Temp, Desktop & Applications directories?
by Jeff Flinn 06 Nov '08

06 Nov '08

Is there any existing facility to get in particular the path to the user's temp directory in FileSystem? I couldn't find one in boost_1_35_0, but I've might have missed it. Any plans for this functionality? On Win and Mac there's getenv TEMP/TmpDir, or is there a better way. Thanks, Jeff

3 3

Reflection?
by Jean-Louis Leroy 06 Nov '08

06 Nov '08

Is runtime reflection going to be available in boost soon? The last I have found on the subject is http://lists.boost.org/Archives/boost/2007/08/126006.php but I see nothing in the sandbox nor the vault. I have toyed with an implementation myself. It supports member variables and functions, multiple and virtual inheritance and object creation. Here is what it looks like. First a small hierarchy: class animal { /* ... */ }; class mammal : public virtual animal { /* ... */ }; class carnivore : public virtual animal{ /* ... */ }; class dog : public mammal, public carnivore{ /* ... */ }; Reflection info is filled like this: reflection_class<animal>::name("animal"); reflection_class<animal>::variable("age", &animal::age); // etc reflection_class<mammal>::name("mammal"); reflection_class<mammal>::base<animal>(virtual_); // etc reflection_class<carnivore>::name("carnivore"); reflection_class<carnivore>::base<animal>(virtual_); // etc reflection_class<dog>::name("dog"); reflection_class<dog>::base<mammal>(); reflection_class<dog>::base<carnivore>(); // a function reflection_class<dog>::function("bark", &dog::bark); // overloaded functions reflection_class<dog>::function<string (dog::*)(void) const>("name", &dog::name); reflection_class<dog>::function<void (dog::*)(const string&)>("name", &dog::name); // constructor reflection_class<dog>::constructor<void* (*)()>(); // etc And here how it's used: animal* pa(new dog); object o(downcast(pa)); // recover dynamic type // access member variable inherited from virtual base `animal' member_variable<int> dog_age(o, "age"); int age = o[dog_age]; // call member function member_function<string (object::*)() const> dog_name(o, "name"); string name = o[dog_name](); // create a dog using default ctor object o("dog"); Ok, there is a bit more to it (e.g. support macros, ambiguity detection, handling of shadowed members, overloaded ctors) but the important part is shown above. All in all it is a foundation upon which one can build multimethods, generic object processing (like Perl's data::Dumper), serialization, object-relational mapping and all sorts of nice tools. Does anybody see interest in this work? Jean-Louis

6 15

Is Boost.Astronomy feasible?
by Mika Heiskanen 06 Nov '08

06 Nov '08

Hello, I would like to think that it would be possible to create a library for common astronomical calculations, and I think boost might be the proper place to create one for the C++ community. First a definition: by common calculations I mean those most relevant to every dat life. For normal people this would pretty much exclude most astronomical calculations except those concerning the sun and the moon. Some examples on what calculations I think would be common are: o sunrise, sunset, noon o sun elevation, azimuth, declination o length of the day o phase of the moon, and whether the visible fraction is increasing or not However, there are some complications: o There are several ways to define twilight times, for example civil, nautical and astronomical. o The sun may set/rise more than once during a day in polar regions, or it may do neither. o What is the definition for the length of the day if sunrise/sunset occurs more than once during a day? o Are separate methods needed for testing whether the given date is of type "polar day" or "polar night" or "midnight sun"? The definitions for the three seem to be a bit ambiguous. For example, it is perfectly possible for the sun to be up at midnight but still to set during the day, in which case you would have "midnight sun" but not necessarily a "polar day", when the sun shines 24h. o Most calculations are intimately tied to the local time, so the API should rely on Boost.Date_time. Many places refer to Meeus's work for the most accurate calculations. The formulas have been ported to C++ in the form of the AA++ (astronomical algorithms) library. However, the library seems to be aimed for experts and there are no easy recipies for solving the common problems mentioned above. The questions I would like to pose are then: o Would a boost::astronomy namespace as implied above be a good addition to boost? o Are there any experts out there who understand the subject matter well enough to provide a robust implementation or atleast guidance on the matter? For example, NOAA provides web based solar position and sunrise calculators. According to the documentation the calculators are based on the work of Meeus. However, there are some clear design issues on how things should be handled in C++ versus the JavaScript embedded into the NOAA pages. Any thoughts on the matter? Regards, --> Mika Heiskanen

5 8

[c++0x] Boost implementations of C++0x standard library components?
by Beman Dawes 05 Nov '08

05 Nov '08

I want to submit an improved timer to Boost. I'd like it to use the C++0x standard library's <chrono> and <ratio> headers. Because it will be many years before such C++0x are universally available, I think Boost should include an implementation of these libraries. Howard Hinnant has kindly contributed his demo implementation of <chrono> and <ratio> as a starting point, so the hard stuff is already done. I'm starting to boostify his code. That brings up questions of header and namespace names. Presumably what we do with these two small headers will set the pattern for any other C++0x standard library headers we include in Boost. Here is a strawman proposal for how we organize C++0x standard library components we choose to add to Boost: * Such libraries go through the regular formal review process, although it will need to be adjusted a bit since the interface is already known and frozen. * C++0x standard library header naming convention is followed. Thus these will be named <chrono> and <ratio> rather than <chrono.hpp> and <ratio.hpp>. * C++0x standard library namespaces are used. Thus namespace std and std::chrono rather than boost and boost::chrono. * Directory structure: boost-root boost c++0x // <ratio>, <chrono>, and other c++0x std lib headers go here ... libs c++0x chrono build doc src // if needed test ratio build doc src // if needed test ... This allows a user to refer to say the ratio header as <boost/c++/ratio>, <c++/ratio>, or just plain <ratio>. The user would have to provide an additional include path of boost-root/boost or boost-root/boost/c++0x for the second and third usages to work. Comments? --Beman

4 5

[MultiIndex] interface comments
by David Abrahams 05 Nov '08

05 Nov '08

Just reading through the docs for this very useful library and I had a few notes: 0. The term /typelist/ (italicized) is used in the tutorial without definition. Do you mean an MPL type sequence, or something else? 1. Seems like the tags functionality could take advantage of Boost.Parameter to provide a nicer interface and save some code. 2. The "shortcuts" save exactly 1 character. Each index of a multi_index_container uses its own iterator types, which are different from those of another indices. As is the rule with STL containers, these iterators are defined as nested types of the index: employee_set::nth_index<1>::type::iterator it= es.get<1>().find("Judy Smith"); <snip> Additionally, multi_index_containers provide shortcut definitions to the iterator types of their constituent indices: employee_set::nth_index_iterator<1>::type it= es.get<1>().find("Judy Smith"); There is a variation of the expression above for use with tags: employee_set::index_iterator<name>::type it= es.get<name>().find("Judy Smith"); // get<1> would also work Notice the column in which "it" appears in both versions. IMO these interfaces should be deprecated (i.e. not documented, and eventually removed) as they provide no readability advantage, hurt learnability, and complicate usage. Regards, -- Dave Abrahams BoostPro Computing http://www.boostpro.com

3 4