- Boost - lists.boost.org

Interest in HeapAware mixin?
by Corey Goldfeder 03 Jul '08

03 Jul '08

Hi everyone, Would there be any interest in a mixin class that signs an object as being on the stack or the heap? Any class that needs to know if it can delete 'this' can (possibly multiply) inherit from class HeapAware, and will gain a member function on_heap() that returns whether the object was dynamically allocated. A similar idea is discussed in More Effective C++, item 27, but unlike that approach, my implementation works correctly for arrays. It also does all the work during construction, so that checking on_heap() just returns a preset bool. The extra cost at construction is a small constant. It is not yet thread safe, but could be made so reasonably easily. Unlike some similar classes, it uses nothing compiler specific, even for arrays, which is where people usually need to "cheat". In particular, for arrays it makes no assumptions about where in the allocated memory the objects are constructed, and so alignment issues or extra compiler bookkeeping don't cause any problems. I currently use this class within a 3d mesh library I maintain. I have triangles that belong to multiple meshes, and I want updates to a triangle to be noticed by all meshes that contain it, so that they can mark their cached mesh statistics as being out of date. This means I need an event listener mechanism. I also want triangles ownership to be shared among the meshes, so that deleting a mesh deletes any triangles that no other mesh currently contains. I built a two way smart pointer / event listener class that maintains a single list that serves as both listeners and owners. The HeapAware class that I am proposing for Boost was invaluable to me, because it allowed me to disable self deletion for triangles on the stack. The snippet below demonstrates how HeapAware can be used. (Obviously this a trivial example). My implementation produces the correct output for all these cases, as well as for globals, statics, and objects created with placement new and placement new[]. It also works correctly with copy construction, which some other implementations fail at. class Widget : public HeapAware {}; int main() { Widget stack_w; Widget* heap_w = new Widget; Widget stack_copy(*heap_w); Widget* heap_copy = new Widget(stack_w); Widget stack_array_w[2]; Widget* heap_array_w = new Widget[2]; printf("Stack Widget: %d\n", stack_w.on_heap()); printf("Heap Widget: %d\n", heap_w->on_heap()); printf("Stack copy of Heap widget: %d\n", stack_copy.on_heap()); printf("Heap copy of Stack widget: %d\n", heap_copy->on_heap()); printf("Stack array widget 0: %d\n", stack_array_w[0].on_heap()); printf("Stack array widget 1: %d\n", stack_array_w[1].on_heap()); printf("Heap array widget 0: %d\n", heap_array_w[0].on_heap()); printf("Heap array widget 1: %d\n", heap_array_w[1].on_heap()); return 0; } Thanks! -Corey Goldfeder

1 0

Exception lib documentation
by Emil Dotchevski 02 Jul '08

02 Jul '08

I have a question about the release procedure, do I need to do anything special to hook the Exception library documentation into the main Boost documentation? (The main Boost Exception documentation file is libs/exception/doc/boost-exception.html) Emil Dotchevski Reverge Studios, Inc. http://www.revergestudios.com/reblog/index.php?n=ReCode

3 5

[release][website] Adding 1.36 to the beta site and checklist additions
by Daniel James 02 Jul '08

02 Jul '08

Hi, I've just created a placeholder for the new release information in the beta site. I hope I'm not jumping the gun there. Also, I've added a few items to the release manager checklist which need to be done for the website. These tasks probably should be delegated but I thought they were worth including in the main checklist. Also, would it be possible to upload (or download for that matter) the release snapshots to the beta site? So we can see how well the new documentation works in the site before it is released. thanks, Daniel

1 0

Is "+" safe for file names?
by Rene Rivera 02 Jul '08

02 Jul '08

Just noticed this directory name: libs/filesystem/test/vc++.boost.filesystem Is "+" safe to use across file systems? Hm, actually I'm fairly sure more than one "." is not safe. For example in the MVS systems? -- -- Grafik - Don't Assume Anything -- Redshift Software, Inc. - http://redshift-software.com -- rrivera/acm.org (msn) - grafik/redshift-software.com -- 102708583/icq - grafikrobot/aim,yahoo,skype,efnet,gmail

7 11

Where can I find Phoenix 2.0?
by Mathias Gaunard 02 Jul '08

02 Jul '08

I discovered Phoenix 2.0 which looks fairly great according to the documention, however I can't find it anywhere. Is it what's in trunk/final/boost/spirit/phoenix/ on the spirit svn?

2 1

[circular_buffer] will release branch be updated with fixes from trunk?
by Thorsten Ottosen 02 Jul '08

02 Jul '08

Hi Jan, The old behavior of the default constructor is causing my release regressions to break. I'm sure the new behavior (not doing any allocation) will be most welcome in the release branch for other reasons as well. Do you plan on changing this in the release branch? best regards -Thorsten

1 0

Subversion problem(s)?
by Dean Michael Berris 02 Jul '08

02 Jul '08

I tried doing the following just now... Anything going on? $ svn co http://svn.boost.org/svn/boost/trunk boost-trunk-flux svn: REPORT request failed on '/svn/boost/!svn/vcc/default' svn: REPORT of '/svn/boost/!svn/vcc/default': 400 Bad Request (http://svn.boost.org) -- Dean Michael C. Berris Software Engineer, Friendster, Inc.

2 2

[asio] failure in ~wrapped_handler
by Dean Michael C. Berris 02 Jul '08

02 Jul '08

Hi everyone, I recently ran into a problem with Boost.Asio while using a slightly modified version of the HTTP 2 example server. The debugger trace (attached) shows where a segfault had occurred in the code. The short of it is that when ~wrapped_handler is called, the call to boost::detail::atomic_exchange_and_add fails when boost::detail::sp_counted_base::release is invoked. It occurs when a handler_wrapper is deleted/destructed (~handler_wrapper is called) in the strand_service. I'm just wondering what could be causing this -- it's worth noting that under not-so-heavy load, the problem does not manifest itself; but soon enough when the load on the server goes up (more incoming connections) the server crashes consistently. Any pointers where I should be looking? I'm using Boost.Trunk revision number 43296. -- Dean Michael Berris Software Engineer, Friendster, Inc.

3 2

Re: [boost] [system] boost compilation
by simonvifr＠voila.fr 01 Jul '08

01 Jul '08

Hello, Thank you for your help. The error is now fixed. My configuration was not correctly set. However, it still remains an error when building the C++ program. After including the line "#include <boost/asio.hpp>" the following messages are displayed : Linking console executable: .objs\Menu.o:Menu.cpp:(.text+0x19a3): undefined reference to `boost::system::get_system_category()' .objs\Menu.o:Menu.cpp:(.text+0x19bc): undefined reference to `boost::system::get_posix_category()' .objs\Menu.o:Menu.cpp:(.text+0x19d5): undefined reference to `boost::system::get_posix_category()' .objs\Menu.o:Menu.cpp:(.text+0x19ee): undefined reference to `boost::system::get_system_category()' .objs\Menu.o:Menu.cpp:(.text$_ZN5boost4asio5error19get_system_categoryEv[boost::asio::error::get_system_cathttp://mail3.voila.fr/webmail/fr_FR/write.html# Envoyeregory()]+0x7): undefined reference to `boost::system::get_system_category()' collect2: ld returned 1 exit status Process terminated with status 1 (0 minutes, 0 seconds) I think I need to link the "boost_system-iw-mt-1_35.lib" and/or the "boost_system-iw-mt.lib" libraries but I do not know how. I have tried without success. Thank you for your answer to fix this error. Regards. > Message du 01/07/08 à 16h36 > De : "Beman Dawes" <bdawes(a)acm.org> > A : boost(a)lists.boost.org > Copie à : > Objet : Re: [boost] [system] boost compilation > > > simonvifr(a)voila.fr wrote: > > > I contact you because an error occurs when compiling the "system" > > library with the following command prompt : bjam --with-system > > --toolset=intel --debug-configuration stage > > > > The intel compiler seems to be invoked correctly (see mail below). > > > > The message error is : > > > > error_code.cpp libs\system\src\error_code.cpp(90): error: namespace > > "std" has no member "strerror" const char * c_str = std::strerror( ev > > ); ^ compilation aborted for libs\system\src\error_code.cpp (code 2) > > > > My OS version is Windows XP. > > My boost version is 1.35.0 > > My boost-jam version is 3.1.16-1 (ntx86) > > My intel compiler version is 10.1.021 > > Each of those is exactly identical to the setup on my main development > machine. The Intel compiler is one of the ones I test with regularly. > > Yet the command line you specify about works fine for me. > > One possibility, and it is only a guess, is that your configuration is > finding the wrong set of C++ standard library include files. > > > Thank you for your answer to fix this error. > > HTH, > > --Beman > > _______________________________________________ > Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost > > ____________________________________________________ Sur le mail Voila, vous pouvez personnaliser l’anti-spam ! http://mail.voila.fr

1 0

DocBook -> Boost documentation?
by Boris 01 Jul '08

01 Jul '08

Is there any XSLT file to convert a DocBook to a Boost library documentation (the HTML files everybody uses)? I've been clicking around on the Boost website to find out how a library documentation is created but could only find http://www.boost.org/development/requirements.html#Documentation which refers to http://www.boost.org/doc/libs/1_34_1/more/writingdoc/index.html (which basically says everyone may do what he wants). Boris

3 4