- Boost-users - lists.preview.boost.org

function docs
by Ihsan Ali Al Darhi 10 Sep '02

10 Sep '02

Hi... I downloaded Boost 1.28.0. I looked at the function library folder but I didn't see any documentation on how to use this library. Where's it? If there isn't, who would volunteer and explain how to use this library? I am using gcc 3.2 on Win2000 and sometimes on Win98. Mo

2 2

RE: [Boost-Users] Slightly Off-Topic, Books
by Jim.Hyslop 10 Sep '02

10 Sep '02

I wrote: > Your post is not "slightly" off topic, it is entirely off topic. Sorry, I didn't mean to come across so brusque. I should at least have included some references in that message: More information about what is on-topic can be found at http://groups.yahoo.com/group/Boost-Users/ Some forums you might consider are comp.lang.c++ and alt.learn.c++ (I _think_ that's the name of the group) -- Jim

2 1

Re: msvc-stlport build problems
by David Resnick 09 Sep '02

09 Sep '02

OK, here is the real fix to the problem. Rehash: Compiling for tool "msvc-stlport" fails with these errors repeated for many files: C:\PROGRA~1\MICROS~3\VC98\include\winbase.h(1011) : error C2733: second C linkage of overloaded function 'InterlockedIncremen t' not allowed C:\PROGRA~1\MICROS~3\VC98\include\winbase.h(1009) : see declaration of 'InterlockedIncrement' C:\PROGRA~1\MICROS~3\VC98\include\winbase.h(1018) : error C2733: second C linkage of overloaded function 'InterlockedDecremen t' not allowed C:\PROGRA~1\MICROS~3\VC98\include\winbase.h(1016) : see declaration of 'InterlockedDecrement' C:\PROGRA~1\MICROS~3\VC98\include\winbase.h(1026) : error C2733: second C linkage of overloaded function 'InterlockedExchange ' not allowed C:\PROGRA~1\MICROS~3\VC98\include\winbase.h(1023) : see declaration of 'InterlockedExchange' This problem occurs when the Platform SDK is installed. The PSDK include path must be searched before the standard MSVC includes. The bjam tool adds -I"C:\PROGRA~1\MICROS~3\VC98\include" to the cl command line, so that is searched before the PSDK include path which is specified in the environment. An immediate solution is to have the PSDK include path searched first, using this bjam command line: bjam -sTOOLS=msvc-stlport -sBUILD="debug release <stlport-iostream>on <sysinclude>C:\Progra~1\MIFD68~1\Include\." -sSTLPORT_PATH=c:\etc NOTE: "C:\Progra~1\MIFD68~1\Include\." is the short form of "C:\Program Files\Microsoft SDK\Include" on my machine. YMMV. This should probably be fixed in msvc-tools.jam which would somehow check if the PSDK is installed (by checking the environment?) and added to the header search path as needed. BTW: I've added this to http://www.crystalclearsoftware.com/cgi-bin/boost_wiki/wiki.pl?CommonBui ldProblems Regards, David Resnick MobileSpear Inc. -----Original Message----- From: David Abrahams [mailto:yg-boost-users@m.gmane.org] Sent: Friday, September 06, 2002 23:56 To: boost-users(a)yahoogroups.com Subject: [Boost-Users] Re: Re: msvc-stlport build problems "David Resnick" <dresnick(a)mobilespear.com> wrote in message news:DDE1CEBCDCEE614EA25D40FCE8EDD6BD0C48F6@msmail.mobilespear.com... > I've fixed the problem, here's my solution in case anyone else needs it: > > > The problem didn't have to do with the Platform SDK in the end. > > The configuration that I'm using is MSVC6 SP3 on Windows 2000. > > The fix was to add : > > flags msvc-stlport DEFINES : _NTOS_=1 ; > > to the "special flags for STLport 4.0" block in > tools\build\msvc-stlport-tools.jam (line 117, for example). Is that #define needed by STLPort, or is specific to the code you're trying to compile? If it's the former, we should patch the toolset. Please respond to jamboost(a)yahoogroups.com > I have a question for jam experts: How can I set this define (_NTOS_=1) > from the command line? bjam -sBUILD="<stlport-iostream>on <define>_NTOS_=1" ... ----------------------------------------------------------- David Abrahams * Boost Consulting dave(a)boost-consulting.com * http://www.boost-consulting.com Yahoo! Groups Sponsor ADVERTISEMENT Info: <http://www.boost.org> Wiki: <http://www.crystalclearsoftware.com/cgi-bin/boost_wiki/wiki.pl> Unsubscribe: <mailto:boost-users-unsubscribe@yahoogroups.com> Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.

1 0

boost:shared_ptr<> exception in checked_delete(..)
by Julia Donawald 09 Sep '02

09 Sep '02

Hi, I have written the following code to simulate my problem: #include <vector> #include <shared_ptr.hpp> class Foo { public: Foo() { }; virtual ~Foo() { }; }; class Vector_Foo { public: Vector_Foo() { }; virtual ~Vector_Foo() { }; void AddFooObject(Foo* pFoo) { FooDataPtr ptrFoo(pFoo); m_vecChildren.push_back(ptrFoo); }; protected: typedef boost::shared_ptr<Foo> FooDataPtr; typedef std::vector<FooDataPtr> FooVector; FooVector m_vecChildren; }; int main( ) { Vector_Foo vec; Foo* pFoo = new Foo(); vec.AddFooObject(pFoo); vec.AddFooObject(pFoo); return 0; } When I now leave the function main, and the vec-Object is deleted I got an exception in the function: template< typename T > inline void checked_delete(T * x) { typedef char type_must_be_complete[sizeof(T)]; delete x; } in checked_delete.hpp What have I done wrong? Why doesnt my code work? Should I use another class then shared_ptr to do that what I want, namely to insert 2 pointers to the same Object in a vector? Thanks in advance Julia

2 1

BUG in Regex++ 3.31
by Yaroslav Govorunov 09 Sep '02

09 Sep '02

Hello, All! Call to the regex_match(chkString,Match,Expression,match_not_dot_newline) causes program go into an infinite loop when chkString="http://localhost/ideas/exchange/June98/test/test.cfm" and Expression="^/(.*)*.cfm" Both Regex++ and test program are compiled with VC7.0 on WindowsXP. I have tried to play with optimisation settings, but this bug appears even in debug configuration. I understand that syntax of regular expression is incorrect, but library should rise error in this case. I hope that I'm doing something wrong and there is a workaround to this problem. With best regards, Yaroslav Govorunov, E-mail: slov(a)rsdn.ru http://www.rsdn.ru http://www.isapirewrite.com

3 5

Re: boost:shared_ptr<> exception in checked_delete(..)
by Julia Donawald 08 Sep '02

08 Sep '02

Hi, the folllowing works not too: #include <vector> #include <shared_ptr.hpp> class Foo { public: Foo() { }; virtual ~Foo() { }; }; class Double_Foo { public: Double_Foo() { }; Double_Foo(Foo* pFoo1, Foo* pFoo2) : m_Foo1(pFoo1), m_Foo2(pFoo2) { }; virtual ~Double_Foo() { }; protected: typedef boost::shared_ptr<Foo> FooDataPtr; FooDataPtr m_Foo1; FooDataPtr m_Foo2; }; int main() { Foo* pFoo = new Foo(); Double_Foo DoubFoo(pFoo, pFoo); return 0; } I got the same error as in my former post. I read a bit about shared_ptr. Maybe it has something to do with a cyclic, but I thought in anohter case of cyclic data, namely if I have an object A which points to B and B holds a member which points back to A.... that's in my case not? If it has something to do with cyclic data then how can I solve it with weak_ptr? Thanks in advance Julia

2 1

uBLAS: Bug in zero_vector.
by David Jung 08 Sep '02

08 Sep '02

Hi. The online documentation for zero_vector<> says: zero_vector<T> Constructs a zero vector. v (size) However, line 786 of boost/numeric/ublas/vector.hpp from latest CVS reads: zero_vector (size_type size, size_type index): size_ (size) {} and there is no single size_type size argument constructor. Isn't this a bug? (looks like a cut-paste from unit_vector<>) Cheers, -David Jung.

2 1

Re: Re: [Boost-Users] Compile problem with boost thread and function libraries.
by Matthew Dempsky 08 Sep '02

08 Sep '02

Nevermind, I'm an idiot... I forgot to add -I/usr/include/stlport to gcc's commandline. =/ -jivera

1 0

Re: Re: Compile problem with boost thread and function libraries.
by Matthew Dempsky 08 Sep '02

08 Sep '02

>Probably a silly question: but you did build libboost_thread with STLport >support turned on didn't you? Well, I didn't actually compile libboost_thread at all - it's prepackaged on Debian. -jivera

1 0

Question about functional operators on bit-ops (~, &, |, ^, +)
by Jeffrey C. Jacobs 08 Sep '02

08 Sep '02

This is more of a newbie question I admit but then that's appropriate here as I understand. :) Really, my question comes down to perhaps ignorance of the ISO spec (need to get around to buying a copy! -- and YES, I know EXACTLY what I'd be getting into! ;) and determine whether this is really just lack of conformance on my compiler's part (you know the one, using Dinkumware 1.0 for 5 years) or an issue addressed by boost. Anyway, cut to the chase: I am looking for unary (binary) functions of this form: // TEMPLATE STRUCT compliment template<class _Ty> struct compliment : std::unary_function<_Ty, _Ty> { inline _Ty operator()(const _Ty& _X) const {return (~_X); } }; // TEMPLATE STRUCT unary_no_op template<class _Ty> struct unary_no_op : std::unary_function<_Ty, _Ty> { inline _Ty operator()(const _Ty& _X) const {return (_X); } }; That is, wrappers for compliment (~), unary positive (+), binary bitwise-and (&), bitwise-or (|) and bitwise-XOR (^) as well as the second concept I listed: the unary_no_op, which though functionally similar to unary "+" is none the less unique in that is really does nothing, where as +_X calls _X.operarot+(). [Yes, I know the "no-op" unary is somewhat silly in its usefulness but I only suggest it for abstraction purposes. I can give an example how it would be useful, but the code would likely be flawed in other senses to I'll reserve that until I need to justify such a construct.] Anyway, so first: Am I just missing something? I don't think the standard defines a "compliment" functional object but if it does, can someone tell me? And if as I believe it's not, given the usefulness of such a template functional object (regardless of boost's Lambda lib), where would I find the definition in boost of such functors listed above? If it's not defined, is there any reason we'd not like to see such an operator standardized? I mean, I'd be happy to do the trivial work of create such a list for boost but then I can only believe if it was wanted it would have been done. Then again, it's wanted by me so I suppose it's not completely unreasonable to expect it to be here but that's a whole other issue. So is there a standard (std:: or boost::) complement class, bitwise_and, etc. that is NOT confused with the ISO646 definitions? Thanks in advance and hope I was clear and did not offend with this possibly silly question. Jeffrey.

1 0

Compile problem with boost thread and function libraries.
by Matthew Dempsky 08 Sep '02

08 Sep '02

Trying to compile the example at the bottom of the thread_group page (http://www.boost.org/libs/thread/doc/thread.html#example-thread_group) results in the following error for me: undefined reference to `boost::thread_group::create_thread(boost::function0<void, boost::empty_function_policy, boost::empty_function_mixin, allocator<boost::function_base> > const &)' (I also had to rename the variable count to something else because I was getting a "`int count' redeclared as different kind of symbol" error and change boost::mutex::lock to boost::mutex::scoped_lock.) I linked with -lboost_thread -lpthread -lstlport and am using Debian 3.0 with g++ 2.95.4, stlport 4.5.3, and boost 1.27. -jivera

2 1

RE: [Boost-Users] Re: msvc-stlport build problems
by David Resnick 06 Sep '02

06 Sep '02

I've fixed the problem, here's my solution in case anyone else needs it: The problem didn't have to do with the Platform SDK in the end. The configuration that I'm using is MSVC6 SP3 on Windows 2000. The fix was to add : flags msvc-stlport DEFINES : _NTOS_=1 ; to the "special flags for STLport 4.0" block in tools\build\msvc-stlport-tools.jam (line 117, for example). I have a question for jam experts: How can I set this define (_NTOS_=1) from the command line? Regards, David Resnick MobileSpear Inc. -----Original Message----- From: David Abrahams [mailto:yg-boost-users@m.gmane.org] Sent: Friday, August 30, 2002 19:38 To: boost-users(a)yahoogroups.com Subject: [Boost-Users] Re: msvc-stlport build problems "David Resnick" <dresnick(a)mobilespear.com> wrote in message news:DDE1CEBCDCEE614EA25D40FCE8EDD6BD0C48EE@msmail.mobilespear.com... > Hi, > > I'm trying to build boost version 1.28.0, msvc-stlport build, using the > following line: > > bjam -sTOOLS=msvc-stlport -sstlport-iostream=on -sSTLPORT_PATH=c:\etc -sstlport-iostream=on sets a Jam variable called "stlport-iostream" to "on". As a build feature, you need to set this up as a part of the BUILD variable: -sBUILD="<stlport-iostream>on" However, that doesn't appear to be related to the warnings you're reporting. > I have no problems building the msvc build. > > It seems to me that the problem has to do with the MS Platform SDK that > I have installed. Quite possibly. > When compiling STLPort (version 4.5.3) I had similar warnings. 2 work > arounds that I found there were either to remove the platform SDK from > the include enviroment variable, or (when I found it) to define > _STLP_NEW_PLATFORM_SDK in stl_user_config.h. > > I've tried variations of that with compiling Boost, but nothing makes a > difference. I've included the error messages that I get for most files. > > ---------------------------- CUT HERE ---------------------------- > vc-C++ > libs\regex\build\bin\libboost_regex.lib\msvc-stlport\debug\runtime-link- > dynamic\stlport-cstd-namespace-std\stlport-ios > tream-off\stlport-version-4.5.3\c_regex_traits.obj > c_regex_traits.cpp > C:\PROGRA~1\MICROS~3\VC98\include\winbase.h(1011) : error C2733: second > C linkage of overloaded function 'InterlockedIncremen > t' not allowed > C:\PROGRA~1\MICROS~3\VC98\include\winbase.h(1009) : see > declaration of 'InterlockedIncrement' > C:\PROGRA~1\MICROS~3\VC98\include\winbase.h(1018) : error C2733: second > C linkage of overloaded function 'InterlockedDecremen > t' not allowed > C:\PROGRA~1\MICROS~3\VC98\include\winbase.h(1016) : see > declaration of 'InterlockedDecrement' > C:\PROGRA~1\MICROS~3\VC98\include\winbase.h(1026) : error C2733: second > C linkage of overloaded function 'InterlockedExchange > ' not allowed > C:\PROGRA~1\MICROS~3\VC98\include\winbase.h(1023) : see > declaration of 'InterlockedExchange' > > > "cl" /Zm400 -nologo -GX -c -DBOOST_REGEX_NO_LIB=1 > -DBOOST_REGEX_STATIC_LINK=1 -D_STLP_NO_SGI_IOSTREAMS=1 -D_STLP_DO_IMP > ORT_CSTD_FUNCTIONS=1 -D_STLP_DEBUG=1 -D_STLP_DEBUG_UNINITIALIZED=1 > -D_STLP_NEW_PLATFORM_SDK=1 /Z7 /Od /Ob0 /GX /GR /MDd -I > "libs\regex\build" -I"C:\etc\boost_1_28_0" > -I"c:\etc\STLport-4.5.3\stlport" -I"C:\PROGRA~1\MICROS~3\VC98\include" > -Fo"libs > \regex\build\bin\libboost_regex.lib\msvc-stlport\debug\runtime-link-dyna > mic\stlport-cstd-namespace-std\stlport-iostream-off\s > tlport-version-4.5.3\c_regex_traits.obj" > -Tp"libs\regex\build\../src\c_regex_traits.cpp" > > ...failed vc-C++ > libs\regex\build\bin\libboost_regex.lib\msvc-stlport\debug\runtime-link- > dynamic\stlport-cstd-namespace-std\s > tlport-iostream-off\stlport-version-4.5.3\c_regex_traits.obj... > ---------------------------- CUT HERE ---------------------------- > > Thanks in advance, > > David Resnick > MobileSpear Inc. -- ----------------------------------------------------------- David Abrahams * Boost Consulting dave(a)boost-consulting.com * http://www.boost-consulting.com Yahoo! Groups Sponsor ADVERTISEMENT <http://rd.yahoo.com/M=233351.2287382.3722243.2225243/D=egroupweb/S=1705 006788:HM/A=1212975/R=0/*http://www.gotomypc.com/u/tr/yh/grp/300_mapG/g2 2lp?Target=mm/g22lp.tmpl> Info: <http://www.boost.org> Wiki: <http://www.crystalclearsoftware.com/cgi-bin/boost_wiki/wiki.pl> Unsubscribe: <mailto:boost-users-unsubscribe@yahoogroups.com> Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service <http://docs.yahoo.com/info/terms/> . [Non-text portions of this message have been removed]

4 4

Exposing static members using boost.python
by sashan 06 Sep '02

06 Sep '02

Hi Is the correct way to expose static members by writing a wrapper function for them? For example void wrap_fn() { Class::fn(); } . . . this_module.def(wrap_fn, "fn"); Thanks in advance Sashan

1 0

array.hpp problem with msvc7 and stlport
by li_lirong 06 Sep '02

06 Sep '02

I got many errors with boost\array.hpp after I chaged to use stlport instead of the one from msvc7. one of these errors is: d:\cpplibs\boost\boost\array.hpp(61) : error C2039: '_Ptrit' : is not a member of '_STL' I have tried to config stlport with STLP_NO_OWN_NAMESPACE and _STLP_DONT_RENAME_STD but the errors still exist. Is it a bug in array.hpp or stlport? Regards, Lirong

2 1

ref. to object functions from threads
by Luis De la Parra 06 Sep '02

06 Sep '02

hi all, is there a way to get a reference to the function object used by a boost::thread object?? when you create a new thread and feed it a function object it copies it and then starts a new thread of execution on this object.. well, I would like to send messages to the object in which the thread is running but I don't know how to do it with boos.threads... is there a way to tell boost:thread not to copy the function object but to use it directly?, or to ask the thread object to give a reference to the object? I am now doing a dirty hack which consists of the object putting itself (a pointer) in a static chained list during construction and remove itself in the destructor, this way I can call a static member function which returns this list to get a hold of the object. There is also a static mutex protecting this list to make sure no other objects are created / destructed while one of them is inserting or removing itself from the list. it is working, but I don't think this is a very "clean" approach.. does anyone have any ideas how it could be done better?? thank you. regards, luis.

3 2

n-ary ordered tree template in boost ??
by Marc Cromme 06 Sep '02

06 Sep '02

Dear Boost Members I was wondering if there is an n-ary ordered tree template in boost, something like template <size_t N, class Type, class Alloc = allocator<Type > > class NTree which is a standard container, and a reversible container, ordered by level. That is, first element is root, next series of elements are childs of roots, next series are grandchilds of root , and so on. This way one could parse level by level like a reversible container. There should be the obvious member functions for inserting/deleting child nodes and subtrees. As far as I can see, there is not such a template in boost, but the question is then: 1) can it be emulated using the grah library ?? or 2) is there already an proposal/prototype for such a thing ?? or 3) would it be interesting to include such a thing in boost ?? your's sincierely, Marc Cromme

4 13

[BGL] 2D/3D layout algorithms
by Andrew Cunningham 06 Sep '02

06 Sep '02

Has anyone adapted any 2D and/or 3D graph layout algorithms to work with Boost ( i.e. minimum crossing etc...). Seems an obvious question, but I did not find anything via searching... Andrew

4 4

RE: [Boost-Users] Re: Is it possible to retrieve blank strings with boost::tokenizer?
by Sean Shubin 05 Sep '02

05 Sep '02

That did it. Thanks. -----Original Message----- From: Markus Schöpflin [mailto:markus.schoepflin@ginit-technology.com] Sent: Thursday, September 05, 2002 1:48 AM To: boost-users(a)yahoogroups.com Subject: [Boost-Users] Re: Is it possible to retrieve blank strings with boost::tokenizer? char_separator<char> sep("\t", "", keep_empty_tokens); should do the trick. HTH, Markus

1 0

Newbie boost::mem_fn problem
by Mark Snelling 05 Sep '02

05 Sep '02

Hello, My problem I think is very simple, I just can't seem to find the solution I have an STL list of object pointers and I want to call a member function on each of the pointers in the list usigng the STL for_each algorithm. This would seem easy if the list contained the actual objects rather than pointers to them, but because they are pointers I can't find a way of doing this. class Block { public: void update(int x); }; void main() { list<Block*> blocks; for_each(blocks.begin(), blocks.end(), *****); // Call update on each item in list } what should I replace the ***** with? Any help appreciated! Mark.

2 3

RE: [Boost-Users] Re: msvc-stlport build problems
by Jim.Hyslop 05 Sep '02

05 Sep '02

David Resnick (dresnick(a)mobilespear.com) wrote: > I've fixed the problem, here's my solution in case anyone > else needs it: Hi, David Would you mind taking a moment to add this to the Boost Users Wiki site? Follow this URL: http://www.crystalclearsoftware.com/cgi-bin/boost_wiki/wiki.pl?CommonBuildPr oblems and edit the page to add a brief description of the problem and how you solved it. Thanks! -- Jim

1 0

Is it possible to retrieve blank strings with boost::tokenizer?
by Sean Shubin 05 Sep '02

05 Sep '02

For example, using the code: string line = "\taaa\t\tbbb\t"; typedef tokenizer<char_separator<char> > tok; char_separator<char> sep("\t"); tok tokens(line, sep); for(tok::iterator i = tokens.begin();i!=tokens.end();++i) { cout << "\t\"" << *i << "\"" << endl; } I get this output: "aaa" "bbb" I want to get this output: "" "aaa" "" "bbb" ""

2 1

boost.python overloading to_python
by dkottowk 04 Sep '02

04 Sep '02

hi, please consider the following example class A { public: A() {} A(const A& a) {} }; class A2 : public A { public: A2() : A() {} A2(const A2& a2) : A(a2) {} }; class B { public: const A& getA(); //returns refs to A or A2 }; now, when exposing this in boost, i would always get A's when calling B::getA() from python, which i understand, since AFAIK the copy Ctors are used to instantiate the python object. to overcome this i wrote: PyObject *to_python(const A &a) { if (dynamic_cast<A2*>(&a)) { return to_python::((const A2 &)a);//done by class_builder<A2> } else { //HOW DO I CALL THE IMPL DONE BY class_builder<A> ? return 0; } } which works fine, except that i dont know how to resort to the automatic implementation for PyObject *to_python(const A& a) provided by the class builder for A. any help is greatly appreciated. greetings, daniel

1 0

thread link problem with STLport
by Sam Partington 04 Sep '02

04 Sep '02

Hi, I'm trying to link an app using boost threads and stlport on gcc 2.95(linux), and its coming back with an undefined reference: Debug/threadex.o: In function `threadex::go(void)': .../threadex.cpp:37: undefined reference to `boost::thread::thread(boost::function0<void, boost::empty_function_policy, boost::empty_function_mixin, allocator<boost::function_base> > const &)' and the offending code is new boost::thread(threadex_caller(this)); where theadex_caller looks like this struct threadex_caller { threadex_caller(threadex*); void operator()(); }; I think the problem is the allocator template parameter of boost::function0, a clue is given if I don't link against the pthread library, I get the following undefined reference (amongst others) Debug/libboost_thread.a(thread.o): In function `boost::thread::thread(boost::function0<void, boost::empty_function_policy, boost::empty_function_mixin, _STL::allocator<boost::function_base> > const &)': .../src/thread.cpp:112: undefined reference to `pthread_create' where you can see that the thread::thread function expects a function0 with an _STL::allocator<>, not a plain allocator<> I also think that the problem is that the thread.hpp isn't getting the namespace mapping function, but why? TIA Sam

1 0

boost::function map with different return types
by Julia Donawald 04 Sep '02

04 Sep '02

Hi, I have the following code: --------------- #include <boost/bind.hpp> #include <boost/function.hpp> #include <map> #include <string> class Foo { public: void Function3() const { int i =2; }; const std::string Function4() const { return ""; }; unsigned int Function5() const { return 2; }; }; std::map<std::string, boost::function<void> > functions; Foo foo; functions["Function3"] = boost::bind(&Foo::Function3, &foo); functions["Function4"] = boost::bind(&Foo::Function4, &foo); functions["Function5"] = boost::bind(&Foo::Function5, &foo); // call the functions functions["Function4"](); -------------- Calling the correct function with the entries in the map is no problem, but how can I get the return value. I tried something like that: "unsigned int i = functions["Function5"]();" but sadly it doesnt work, maybe cause in the map I declared for the return value "void"? How can I have such a map where to have functions with different return types in it and to get the return value after a call of the function through the map? Thanks in advance Julia

3 4

Re: threads and signals question
by Ken Thomases 04 Sep '02

04 Sep '02

On 29 Aug 2002, at 20:12, "lparrab" <lparrab(a)gmx.net> wrote: > I'm starting a multi-threaded program using Boost.Thread library: > it is a "server", in which a thread loops around a "select" to > accept new connections and see when a socket has data to be read, in > which case puts the fds in queue from which other threads are taking > them out to read the data. > The problem I have, is that when a "ReceiverThread" finishes reading > the data from the socket, it puts the fd back in the "all > connections list", but by this time the other ("main")thread already > copied the fds from the list and got into the next "select" loop, > which means the socket from which the "receiver" just finished > reading, will not be "selected" (monitored) untill the next select > loop, which will not be untill a) a new connection comes or b)one of > the other connected sockets gets some data that takes makes the main > thread return from select. The cleanest way to wake the main thread from select is to make one of the file descriptors in its FDSET become readable. What you need is a special file descriptor that is always in the FDSET that becomes readable when some other thread needs the main thread to come out of the select call. For this you can use a pipe. Here's a way that requires minimal changes to your existing design: When a thread needs the main thread to wake up, it writes a byte into one end of the pipe. It doesn't matter what the byte contains, its presence in the pipe is the "information" payload. When the main thread comes out of the select, if its end of the pipe was among the readable fds, then it should pull one byte out of the pipe (and discard it). If you use extreme care, you can instead flush the pipe of all bytes if you're sure you won't cause a race condition that way. (Example of race condition: Main thread comes out of select, Main thread builds new FDSET for next select call, Receiver thread writes to pipe to inform it of newly available fd, Main thread flushes pipe and blocks in select. To fix this race condition, you would reverse the order in which the Main thread flushes the pipe and builds the FDSET.) Here's the better design: Consider: why do you need the Receiver threads to wake the Main thread? To transfer responsibility for an fd from a Receiver thread to the Main thread. So, this is similar to the way the Main thread transfers fds to Receiver threads through a queue, and should be implemented in a similar fashion. Instead of using the pipe just to signal the Main thread to wake, actually use it to transfer the fds, just as you use the queue in the other direction. So now, when Receiver threads are done with an fd, they write the actual fd into the pipe. The Main thread wakes because the other end of the pipe was in its FDSET. It should read any and all fds out of the pipe and add them to its "connection list" for its next pass through select. (You can read just one fd from the pipe for each time that select indicates the pipe is readable, or you can set the pipe to be non-blocking and read until the pipe is empty.) This approach is a better design because it implements the actual transfer of responsibility and because the connection list should be private to the Main thread. This way doesn't require that the Receiver threads access the connection list, so the connection list doesn't have to be protected by a mutex. > I think having a thread waiting on each open connection is not the > right approach here, so I thought of going a kind of "production line" > approach in which a few threads (5-10) are reading the data from a > bunch of open connections (50-100), the only thing they do is read the > data, parse it and form a "Request" which they put in a queue for > further processing and then go on to service the other connections. > Then another group of threads is taking this requests and doing > something, and putting Response objects in yet another queue, which > another group of threads "Sender" is taking and sending thorugh the > socket. Just be aware that this design may mean that the client sees responses arrive out of order from their requests. If the client only submits a request after any previous requests have been satisfied, you're OK. If the client can have multiple requests outstanding at once, the Request threads or the Sender threads may finish out-of-order. Also, if the Sender threads don't write atomically or with exclusive access, then the data from multiple responses might be interleaved. This illustrates why it is cleaner to have a thread take full ownership of a connection for its lifetime as others have suggested. Of course, you're not being unreasonable in your approach, but you can marry the two somewhat. You might have each worker thread combine the Receiver, Request, and Sender jobs in sequence and then relinquish the connection back to the main thread. It's not clear to me that there is any advantage in terms of number of threads, thread workload, or throughput to splitting those jobs up. Hope that helps, Ken

1 0