- Boost-users - lists.boost.org

"Different" use of shared_ptr/weak_ptr
by Goran Pusic 26 Oct '05

26 Oct '05

Hi all! I am trying to use boost smart pointers for some multithreaded work in the following way: 1. I use shared_ptr to wrap stuff in thread 1 2. I create weak_ptr-s in thread 1 and pass them to thread 2 3. in thread 2, I lock my weak_ptr-s and use stuff I get from there 4. in thread 1 I may destroy my shared_ptr 5. next time I want to lock my weak_ptr, I get NULL and can't use stuff Basically, my idea is to use weak_ptr as sort of transport mechanism between threads, but with "stuff destroyed detection". I did some tryouts and this seems to be OK. However, I'd like to get some more authorative advice :-)) Can I do that? If no, why not, or, what would you suggest? Thanks, Goran. --------------------------------- Yahoo! FareChase - Search multiple travel sites in one click.

2 1

test: using floating_point_comparison.hpp in precompiled header
by Matt Amato 26 Oct '05

26 Oct '05

I don't know if this is a bug or not, but using boost/test/floating_point_comparison.hpp in a precompiled header under VC71 on Windows yields multiply defined symbol errors for each file using it. The workaround is to not use it in a precompiled header but only in the source files where it is needed. Any ideas? Linking... stdafx.obj : error LNK2005: "struct check_is_close_t::boost::check_is_close_t boost::test_tools::`anonymous namespace'::check_is_close" (?check_is_close@?A0x5e8fa061@test_tools@boost@@3Ucheck_is_close_t@23@A) already defined in TestExample.obj ..\..\..\..\..\bin\Debug.NET\Test.exe : fatal error LNK1169: one or more multiply defined symbols found

3 2

Extended Thread object
by Phil Endecott 26 Oct '05

26 Oct '05

Dear Experts, The boost::thread documentation says: Just as the lifetime of a file may be different from the lifetime of an iostream object which represents the file, the lifetime of a thread of execution may be different from the thread object which represents the thread of execution. In particular, after a call to join(), the thread of execution will no longer exist even though the thread object continues to exist until the end of its normal lifetime. The converse is also possible; if a thread object is destroyed without join() first having been called, the thread of execution continues until its initial function completes. That's fine, but it's not what I'm looking for right now. Can anyone point me to some code - perhaps adding a layer on top of boost::thread - that ties object lifetime to thread lifetime in some way? (I'm imagining a dynamically-allocated object the deletes itself when the thread's initial function terminates, and that kills the thread if it (the object) is deleted first.) The actual application is a sort of daemon; a main loop accepts tcp connections, and creates a new thread to handle each one. It normally wants to "fire and forget" these threads, and I imagine it creating them with new() and then letting them dangle, until they eventually terminate and delete themselves. My guess is that this could be done in just a few lines of code, but there are probably lots of race conditions to worry about. Any suggestions anyone? Cheers, --Phil.

3 2

thread: condition::notify() does not require a lock - why?
by Pete Chapman 26 Oct '05

26 Oct '05

condition::wait requires that a lock is acquired prior to calling it, but condition::notify does not. The Java threading API (which in my limited experience seems similar to boost::thread) requires that a lock is acquired prior to calling Object.notify(). Is this potentially a problem? For example: scoped_lock myLock(myMutex); while (!some_variable) { // Another thread now sets some_variable = true, and // calls myCondition.notify() // then... // will this wait indefinitely? myCondition.wait(myLock); } Clearly if the other thread in this example were forced to obtain a lock on myMutex before calling myCondition.notify(), there would be no danger. In practice this is what I do anyway, but the boost:thread API doesn't force me to. Why not? Is there a potential problem if I don't? Many thanks...

2 2

bind with boost::function
by Eric Pruneau 26 Oct '05

26 Oct '05

Hi Here is my problem. I need to specify a callback function for use with OpenCV cvcam dll. I modified OpenCV to accept a boost::function1<void,IplImage*> instead of a simple void pointer to a function. The function is called by openCv on every frame of a video stream. Im trying to have the callback function member of a class e.g Class A { ... void init() { ... boost::function1<void,IplImage*> Myfunct; Myfunct = boost::bind(&A::Callback, this); cvcamSetcallback(Myfunct); // OpenCV function to set the callback } void Start() { /* tell opencv to start grabbing frames and to call Callback on every frame */ } void Callback(IplImage* img) { /* process the image */ } ... }; Its not like I want to directly call the callback function and use it like a functor. I want OpenCv to call my member function. When I try to compile this it give me: error C2091: function returns function on a C++ .net compiler. Im really confused and I dont have a clue about how to solve this. Thanks in advance Eric _________________________________________________________________ Balayez vos courriels entrants et sortants et les pièces jointes et contribuez à éliminer les virus destructeurs susceptibles dy être intégrés. http://join.msn.com/?pgmarket=fr-ca&page=features/virus Commencez dès maintenant à profiter de tous les avantages de MSN Premium et obtenez les deux premiers mois GRATUITS*.

2 1

[iostreams] Patch for exception message in mapped_file
by kml＠elreki.net 26 Oct '05

26 Oct '05

Hi! I noticed there's a bug in the detail::cleanup_and_throw() function. It doesn't use the error message passed in the arguments, but always reports "failed getting file size". Please find attached a patch against the CVS version. Cheers, Kim

2 1

[variant] type ambiguity help
by Aaron Griffin 25 Oct '05

25 Oct '05

This isn't a full blown issue really, but it's probably a lack of understanding on my part. variant< typeA, typeB, int, double> v; gives me conversion ambiguities. I can remove the int and it works fine. In an attempt to kludge it, I changed the int to a rational<int>. However, now, "v = 12" is converted to a double, not to the rational. First question, would be, how exactly is the variant choosing to use the double instead of the rational? Does it have anything to do with order? Would <double,rational<int> > be different compared to <rational<int>,double> ? Is there any way to "slay" this ambiguity. It's a rather moot point, but I'd like to know if it's an int or a double, seperately, not one single double which would require me to determine if it is a whole number or not. Thanks in advance, Aaron

1 1

date_time:
by __PPS__ 25 Oct '05

25 Oct '05

Are there better conversion functions from ISO date format (calendar, ordinal, week dates) in basic and/or extended format? Should conversion from invalid iso date like "20050231" fail or produce same results as date(2005,02,31). (I know that it produces the same results, but is it right to convert invalid iso date representation to valid date?) Another thing I noted, some conversion functions accept const std::string &, but most of them accept plain std::string, isn't better to convert all of them to constant references??? If anybody interested, I wrote such function that accepts iso date string in any of the above mentioned 6 combinations and returns date. It runs 90-100 times faster than equivalent date_from_iso_string from boost. And yes, it does make sure that "20050229" fails

4 6

Problems recognizing directionality of an edge
by Jeffrey Holle 24 Oct '05

24 Oct '05

I am using boost 1.33.0. I'm the guy who is working with spanning trees. I've decided that two graphs are in order. 1. The data graph, of type Graph, is bidirectional. 2. The tree graph, of type Tree, is undirectional. Tree is undirectional because this is the only way that I get expected results from a dfs. However, the direction of the associated edge needs to be recorded as a property in the Tree graph. I'm having difficulty in initializing this property. The problem is that when I create the spanning tree, all I have is an Graph::edge_descriptor. I find that an edge will always be in the target vertex's in_edges container, and in the source vertex's out_edges container. This don't allow identification of the edge direction. Am I missing something?

2 1

[boost-users] [MPL.Math] More major changes
by Cromwell Enage 24 Oct '05

24 Oct '05

A new directory, boost/mpl/math/unsigned_big_integral_, holds everything pertaining to unsigned_big_integral. The corresponding tag and forward declaration also have their own header files now. Also, enlarge_integral and its forward declaration have their own files as well. This should maximize decoupling of the three entities. The big_integral.cpp example and test programs have been updated to reflect these changes. As for double, whole_part and rational_part have been improved, but you'll need to #include unsigned_big_integral and rational to use them. Major limitation: attempting to calculate the rational_part of any double smaller than 1e+30 will most likely break your compiler's back; unsigned_big_integral and rational still won't mix. Play around with the double.cpp example program at your own risk. Cromwell D. Enage __________________________________ Yahoo! Mail - PC Magazine Editors' Choice 2005 http://mail.yahoo.com

1 0