September 2004 - Boost - lists.preview.boost.org

[bind] Intel/Linux failures
by Aleksey Gurtovoy 20 Sep '04

20 Sep '04

Peter, Could you please comment on "bind_cv_test" and "bind_stateful_test" tests' failures on all Intel compilers on Linux -- http://tinyurl.com/5p3or? Should we mark these up as expected? -- Aleksey Gurtovoy MetaCommunications Engineering

2 4

RE: [boost] platform specific structures
by Puverle, Tomas (IT) 20 Sep '04

20 Sep '04

> The Win32 API provides InterlockedXXX for atomic operations > on long data types, e.g. InterlockedIncrement. I am not sure > about other OS's, but it should be possible to define an > atomic_long class: > > class atomic_long > { > private: long val; > public: > inline atomic_long & operator++() > { > InterlockedIncrement( &val ); > return( *this ); > } > // ... > }; I think this design could cause problems - presumably, there would need to be a way to test the value of the atomic_long, such as operator==(long), but that could lead to problematic code: if (++myAtomicLong == someValue) //... else //... An easy fix would be to make the return type void. Tom -------------------------------------------------------- NOTICE: If received in error, please destroy and notify sender. Sender does not waive confidentiality or privilege, and use is prohibited.

2 1

Thread/intel-8.0-linux regressions
by Aleksey Gurtovoy 20 Sep '04

20 Sep '04

Does anybody have any ideas what's up with this (http://tinyurl.com/6vecl) Linker output []: Chmod1 ../bin/boost/libs/thread/test/test_tss.test/intel-8.0-linux/debug/threading-multi/test_tss Run output []: EXIT STATUS: 139 ? All thread tests are failing in the similar manner. This is really unpleasant given that the library worked on Intel 8.0/Linux in 1.31. -- Aleksey Gurtovoy MetaCommunications Engineering

3 5

[thread] Build failures for CW, found the xtime problem...
by Rene Rivera 20 Sep '04

20 Sep '04

After some investigation I found the problem with the build failures for xtime.cpp on CW platforms: http://tinyurl.com/5nwxo [1] The issue is the placement of the definition of __STDC_CONSTANT_MACROS to bring in the *INT*_C() constant macros from cstdint. Currently the top of xtime.cpp looks like so: =================================================================== #include <boost/thread/detail/config.hpp> #include <boost/thread/xtime.hpp> #if defined(BOOST_HAS_FTIME) # include <windows.h> # define __STDC_CONSTANT_MACROS # include <boost/cstdint.hpp> #elif defined(BOOST_HAS_GETTIMEOFDAY) # include <sys/time.h> #elif defined(BOOST_HAS_MPTASKS) # include <DriverServices.h> # include <boost/thread/detail/force_cast.hpp> #endif =================================================================== But that placement is not early enough as the <boost/cstdint.hpp> is included from <boost/thread/xtime.hpp>. Moving the definition above the <boost/thread/xtime.hpp> included fixes that problem: =================================================================== RCS file: /cvsroot/boost/boost/libs/thread/src/xtime.cpp,v retrieving revision 1.12 diff -u -r1.12 xtime.cpp --- xtime.cpp 16 Sep 2004 20:39:49 -0000 1.12 +++ xtime.cpp 17 Sep 2004 16:51:04 -0000 @@ -11,11 +11,14 @@ #include <boost/thread/detail/config.hpp> +#if defined(BOOST_HAS_FTIME) +# define __STDC_CONSTANT_MACROS +#endif + #include <boost/thread/xtime.hpp> #if defined(BOOST_HAS_FTIME) # include <windows.h> -# define __STDC_CONSTANT_MACROS # include <boost/cstdint.hpp> #elif defined(BOOST_HAS_GETTIMEOFDAY) # include <sys/time.h> =================================================================== OK to commit?? [1] http://www.meta-comm.com/engineering/boost-regression/developer/output/RSI-… -- -- Grafik - Don't Assume Anything -- Redshift Software, Inc. - http://redshift-software.com -- rrivera/acm.org - grafik/redshift-software.com - 102708583/icq

3 2

Re: [boost] platform specific structures
by Reece Dunn 20 Sep '04

20 Sep '04

Peter Dimov wrote: >John Maddock wrote: >>>I'm interested in getting something like "needs_lock" below into the >>>normal traits, perhaps as part of boost thread. It informs you >>>whether of not an operation on such a contiguous block needs a lock >>>to be atomic. >>> >>>For example, on ia32 32 bit aligned ops are atomic, on ia64 it is 64 >>>bit. doubles on ia32 need locking to be atomic, on ia64 they don't. >>>sizeof(void*) works generically for these two platforms as is >>>included in the code below. A default, safety first implementation >>>might return true always for needs lock, or perhaps, true for all >>>sizeof's greater than a byte. >> >>Is it the case that we could actually rely on this? I thought for >>example that on IA32 operations were only atomic and thread safe when >>the assembly is prefixed by LOCK? > >AFAIK reads and ordinary writes are atomic without LOCK. Read-modify-write >operations need LOCK. The Win32 API provides InterlockedXXX for atomic operations on long data types, e.g. InterlockedIncrement. I am not sure about other OS's, but it should be possible to define an atomic_long class: class atomic_long { private: long val; public: inline atomic_long & operator++() { InterlockedIncrement( &val ); return( *this ); } // ... }; Regards, Reece _________________________________________________________________ It's fast, it's easy and it's free. Get MSN Messenger today! http://www.msn.co.uk/messenger

2 1

RE: [boost] platform specific structures
by Puverle, Tomas (IT) 20 Sep '04

20 Sep '04

(first post to this forum, so be nice guys! :) > For example, on ia32 32 bit aligned ops are atomic But you will also need to express the constraint that the data be aligned? > , on ia64 > it is 64 bit. However, remember that the fact that a given data type can be written in a single operation != visibility across all processors in an MP system. > doubles on ia32 need locking to be atomic, on > ia64 they don't. I think that the later models of ia32 can write up to 8 bytes at a time. > Any suggestions on best practice for this? I think you might want to send this post to comp.programming.threads too - you may get some interesting feedback. I think Alexander Therekhov has some class wrapper for atomic types that I am sure he'd be keen to discuss, but unfortunately I've never had the time to review it. Tom -------------------------------------------------------- NOTICE: If received in error, please destroy and notify sender. Sender does not waive confidentiality or privilege, and use is prohibited.

2 1

[Output Formatter]: Cascading Style Sheets (CSS) proposal
by Reece Dunn 20 Sep '04

20 Sep '04

Throughout this e-mail, I assume the following aliases: namespace io = boost::io; namespace fmt = boost::io::format; namespace css = boost::io::css; NOTE: This is a design outline for features in a future version of the library. David Abrahams has suggested the ability to specify default formatting on a stream. I have implemented a simple version of this so that you can do: io::sequence_decorators< char > seq( "; " ); std::cout << '"' << seq.open( std::cout ) << "\"\n" // "[ " << '"' << seq.close( std::cout ) << "\"\n" // " ]" << '"' << seq.separator( std::cout ) << "\"\n"; // "; " where separator( std::cout ) uses the separator value specified by seq and the others use the values set on the std::cout stream. To change these defaults, you can use: std::cout << io::cdecorate( "[{ ", " }]" ); // [{ a, b, c }] The problem with this is that it changes the defaults for all the formatting used, thus you cannot have different defaults for sequences and n-ary types. Cascading Style Sheets (CSS) are a web standard [http://www.w3.org/Style/CSS/] used for controlling how HTML tags are displayed to different devices (screen, printer and speech). == Inlined Styles In HTML you can specify "inlined" styles using a style attribute, e.g.: <div style = "color: red;">Red Text</div> My library equivalent is to use the decorate() function, e.g.: fmt::range().decorate( "<-- ", " -->" ); == Dynamic Styles You can also set CSS properties within script code, e.g. tag.style.fontFamily = "Times New Roman"; This style behaviour is available in the new decorators, for example: io::wrapper_decorators< char > wrap; wrap.open = "<< "; wrap.close = " >>"; A more complex example that uses nested format objects: fmt::container_t< char, fmt::pair_t< char > > my_fo = fmt::container( fmt::pair()); my_fo.elem.open = "<< "; my_ob.elem.close = " >>"; my_ob.separator = " | "; std::cout << io::object( ob, my_ob ); // output: [ << a, 5 >> | << b, -1 >> ] == Cascading Style Sheets HTML provides a <style> tag to specify styling across the document without linking to an external file, e.g.: <style>h1{ color: orange; font-size: 2pc; }</style> The equivalent within my library would be to define the styling on a stream, e.g.: std::cout << css::element( "pair" ) [ ( css::attribute( "open" ) = "<< " ) + ( css::attribute( "close" ) = " >>" ) ] << css::element( "container" ) [ css::attribute( "separator" ) = " | " ] << io::object( ob ) // output: [ << a, 5 >> | << b, -1 >> ] << css::element( "pair" ) [ css::attribute( "separator" ) = " = " ] << io::object( ob ); // output: [ << a = 5 >> | << b = -1 >> ] HTML can import styling from an external CSS file. The equivalent would be to use an object to hold the styling, e.g.: css::stylesheet< char > style = css::element( "pair" ) [ css::attribute( "separator" ) = " = " ] + css::element( "container" ) [ ( css::attribute( "open" ) = "{ " ) + ( css::attribute( "close" ) = " }" ) ] ; std::cout << style << io::object( ob ); // output: { [ a = 5 ], [ b = -1 ] } It should be possible to provide the ability to save/load styling to/from a file and allow copying of style objects, e.g.: css::stylesheet< char > style2( "mystyles.style" ); // load styling style.load( "newstyles.style" ); style = style2; style.save( "stylings.sty" ); NOTE: The css::stylesheet class will store the styles using something like: std::map< std::string, std::map< std::string, std::string > > therefore it will be possible to use my library to implement the load and save facilities! Regards, Reece _________________________________________________________________ It's fast, it's easy and it's free. Get MSN Messenger today! http://www.msn.co.uk/messenger

5 4

Thread/intel-win32-8.1 failures
by Aleksey Gurtovoy 20 Sep '04

20 Sep '04

As shown in http://tinyurl.com/6ukqm, _all_ thread library tests fail on Intel 8.1 with memory access violation (!). That's especially weird because Intel 8.0 shows quite the opposite results -- it passes all the tests. Could somebody with local access to this compiler please take a look at this? -- Aleksey Gurtovoy MetaCommunications Engineering

3 6

Re: Lexical cast suggestion
by Martin Adrian 20 Sep '04

20 Sep '04

>From: Terje Slettebø <tslettebo <at> broadpark.no> >Subject: Re: Lexical cast suggestion >Newsgroups: gmane.comp.lib.boost.devel >Date: Sat, 18 Sep 2004 14:23:50 +0200 >>From: "Martin" <adrianm <at> touchdown.se> >> I have tried to use lexical_cast many times but I always run into problems >> sooner or later due to the lack of locale support. >This issue has come up with quite regular intervals. The current docs says a >little bit about this: >"Where a higher degree of control is required over conversions, >std::stringstream and std::wstringstream offer a more appropriate path." My concern is that the current implementation gives you a false sense of portability. The documentation says nothing nothing about locales. Even the test cases for lexical_cast assumes the default "C" locale. E.g. BOOST_CHECK_EQUAL("1.23", lexical_cast<std::string>(1.23)); I think there should be a big warning about using lexcial_cast for convering numbers to/from text. The only safe place I see is when the resulting string is used to create user messages. A quick scan shows that at least "program options" uses lexical_cast for other things than user messages. lexical_cast is used to validate entries and to create default values. This means that the options must be specified in the locale the application uses (which might not be the same as the computer is running). This might be intentional but it is not documented so I have no idea how program options works in my locale.

2 1

Linux regressions
by Aleksey Gurtovoy 20 Sep '04

20 Sep '04

A clarification for developers monitoring regression reports: a significant number of Linux regressions showing up in the current reports (http://www.meta-comm.com/engineering/boost-regression/developer/summary.html) is NOT a result of some recent checkin. Most of these failures were there for some time, but weren't highlighted as regressions because the report generation tools didn't have the data for this platform to compare against. This was fixed yesterday. Similarly, new regressions for "cw-8.3" toolset are only a result of updating the old toolset name ("cwpro83") in the 1.30.2 results. I hope we can take care of these new issues quickly, although "gcc-2.95.3-stlport-4.5.3-linux" in particular doesn't look that good. Is it still a widely-used configuration? -- Aleksey Gurtovoy MetaCommunications Engineering

4 6