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

(no subject)
by Hook.RobertA(CC) 04 Mar '02

04 Mar '02

I put the following up at source forge support, then realised that doesn't seem to be likely to get a response.... I would have put this through the yahoo group, but I'm stuck behind a mind-bogglingly repressive firewall. SO I'll scream for help here, bearing in mind the dictum that in cyberspace no-one can hear you scream. I have download 1.27.0, and am attempting to build and use Boost under HP-UX 11.00. There are two problems, and I cannot tell if they are related or not. Firstly, the install instructions along the lines of jam -sBOOST_ROOT=. -=sTOOLS="hpux" simply produce the following (after I bodgied up a hpux-tools.jam) +: line 1304: yacc stack overflow at keyword } don't know how to make all ...found 1 target... ...can't find 1 target... after smacking my head on the desk for several days, I've given up on that one and attempted to build boost_1_27_0/libs/tuple/test/io_test.cpp using aCC -c +z +p $APP_INCLUDE -o io_test.o io_test.cpp (where APP_INCLUDE is -I <yada/yada/yada> as appropriate) which results in "tuple_basic.hpp", line 290 # Cannot generate a specialization of incomplete template struct STATIC_ASSERTION_FAILURE<(bool)0>. BOOST_STATIC_ASSERT(length<cons>::value == 2); // check length = and errors cascading after that. Looking at various messages and comments, I believe that someone somewhere has managed to get boost working under HP-UX, but I'm dammned if I can. Help! Oh yes, by the way, I'm stuck behind a repressive firewall and cannot even *read* the message archive at Yahoo, let alone subscribe / post! (Puts head down on desk and begins sobbing quietly) -- Robert Hook - DCC Analyst/Programmer Polaris Group, Information Systems Branch Queensland Police Service Hook.RobertA(a)police.qld.gov.au PH: 07 3364 3807 ********************************************************************** This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify securityscanner(a)police.qld.gov.au. This footnote also confirms that this email message has been checked for the presence of computer viruses. **********************************************************************

1 0

[Fwd: compile time container type identification]
by hicks 01 Mar '02

01 Mar '02

Hello I need to identify if a container is of a certain type. Specifically, from inside the code for container template <class T> Cont, I want to identity if T is itself of type Cont<U> for some U. (The type of U does not matter.) I couldn't see anything in *_traits.hpp that looks applicable. Any advice or references would be appreciated. Craig Hicks KGK Tokyo [Non-text portions of this message have been removed]

1 0

RE: [Boost-Users] Linking an Archive
by Tom Matelich 27 Feb '02

27 Feb '02

The first getting_started1.so contains all the symbols from all of the objects linked into it. The second one only contains the symbols from getting_started.o and the symbols from libboost_python.a which satisfy unsatified symbols of getting_started1.o. Linking shared library is like linking an executable, it only takes the symbols from archives that it needs to satisfy itself. The first question is why you want libboost_python.a. If it is a valid reason, then you need to find a way to get all the code you need out of it into the shared lib. The first would be to use ar to get all the objects out, then take those and put them into the shared library. The second would be to keep adding unsatified symbols to getting_started.o until getting_started1.so has everything it needs. Tom > -----Original Message----- > From: doak87508 [mailto:jdoak@lanl.gov] > Sent: Wednesday, February 27, 2002 10:21 AM > To: Boost-Users(a)yahoogroups.com > Subject: [Boost-Users] Linking an Archive > > > I am playing around with using an archive to created a shared library > (.so file) instead of linking all the individual objects. I have > success when I link individual objects: > > % g++ -shared classes.o conversions.o errors.o extension_class.o > functions.o init_function.o module_builder.o objects.o types.o > cross_module.o getting_started1.o -o getting_started1.so > > % python > 1;31mh[1] >>> import getting_started1 > 1;31mh[1] >>> getting_started1.greet() > 'hello, world' > > Now, I create an archive: > > % ar r libboost_python.a classes.o conversions.o errors.o > extension_class.o functions.o init_function.o module_builder.o > objects.o types.o cross_module.o > > Then, I link that archive into the shared library: > > % g++ -shared libboost_python.a getting_started1.o -o > getting_started1.so > > % python > 1;31mh[1] >>> import getting_started1.so > Traceback (most recent call last): > File "<stdin>", line 1, in ? > ImportError: ./getting_started1.so: undefined symbol: > add__Q35boost6python19module_builder_basePQ45boost > 6python6detail8functionPCc > > I have also tried linking as follows with no change in result: > > % g++ -shared -L. -lboost_python getting_started1.o -o > getting_started1.so > > Does someone know how to get linking working with an archive file? > > JD > > > > 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 > http://docs.yahoo.com/info/terms/ > >

2 1

in-place set_union/difference/etc ...
by ivan_dobrianov 27 Feb '02

27 Feb '02

Does anyone know of any existing standard tool [an insert iteartor or something equivalent] that can allow one to perform union, difference, etc on sets and maps in-place. E.g. instead of using set<MyClass> s1, s2, temp; set_union( s1.begin(), s1.end(), s2.begin(), s2.end(), inserter(temp, temp.begin())); temp.swap(s2); i'd like to do, say: copy(s1.begin(), s1.end(), my_unioner(s2, s2.begin())); which will do s2 = s1 + s2 in LINEAR time. It's pretty simple to write an inserter like that, but I am just wondering if there is one already ... or something similar. Thanks!

1 0

Bug in boost_1_27_0.tar.gz threads library
by Martin Brown 27 Feb '02

27 Feb '02

Hi There seems to be a missing '}' in libs/thread/src/once.cpp when compiling on gcc-3.04 on Redhat Linux 6.2. I added one and it seems to work. Not sure if it is in the right place. Context diff: *** once.cpp Thu Feb 7 15:33:15 2002 --- once.cpp.new Wed Feb 27 15:53:02 2002 *************** *** 55,62 **** --- 55,63 ---- { once_callback* cb = reinterpret_cast<once_callback*>(pthread_getspecific(key)); (**cb)(); } + } #elif defined(BOOST_HAS_MPTASKS) void *remote_call_proxy(void *pData) { std::pair<void (*)(), boost::once_flag *> &rData(*reinterpret_cast<std::pair<void (*)(), boost::once_flag *> *>(pData)); Thanks for some great libraries! Martin __________________________________________________ Do You Yahoo!? Everything you'll ever need on one web page from News and Sport to Email and Music Charts http://uk.my.yahoo.com

2 1

Linking an Archive
by doak87508 27 Feb '02

27 Feb '02

I am playing around with using an archive to created a shared library (.so file) instead of linking all the individual objects. I have success when I link individual objects: % g++ -shared classes.o conversions.o errors.o extension_class.o functions.o init_function.o module_builder.o objects.o types.o cross_module.o getting_started1.o -o getting_started1.so % python 1;31mh[1] >>> import getting_started1 1;31mh[1] >>> getting_started1.greet() 'hello, world' Now, I create an archive: % ar r libboost_python.a classes.o conversions.o errors.o extension_class.o functions.o init_function.o module_builder.o objects.o types.o cross_module.o Then, I link that archive into the shared library: % g++ -shared libboost_python.a getting_started1.o -o getting_started1.so % python 1;31mh[1] >>> import getting_started1.so Traceback (most recent call last): File "<stdin>", line 1, in ? ImportError: ./getting_started1.so: undefined symbol: add__Q35boost6python19module_builder_basePQ45boost 6python6detail8functionPCc I have also tried linking as follows with no change in result: % g++ -shared -L. -lboost_python getting_started1.o -o getting_started1.so Does someone know how to get linking working with an archive file? JD

1 0

autoconf+make
by josefsyk 26 Feb '02

26 Feb '02

I've been struggling to port parts of the boost libraries to Linux using gnu's automake, just to hit the wall every time. Anyone out there who has done this already? The files needed are: boost/smart_ptr.hpp boost/pending/ct_if.hpp boost/random.hpp boost/bind.hpp boost/function.hpp Would be nice to have the Makefile.am for BGL too. cheers.

2 1

boost::condition question
by svanechel 26 Feb '02

26 Feb '02

Hi, I'm trying to create a mechanism that synchronizes access to shared data on a per-operation basis. The idea is that some operations (E.g. read operations) can proceed concurrently, while others (write operations) must block all other threads. I thought boost::condition would do the trick. Since I do not have that much experience in multithreaded programming it is entirely possible that I do not fully understand boost::condition or its allowable usage. I tried to put operations in an "Operation Group". All operations belonging to the same group are allowed to execute concurrently, while all others are blocked (E.g. all read operations are in the same group). It is also possible that an operation doesn't specify an Operation Group, in which case a temporary group is assigned to the operation for each thread that calls it. Write operations are constructed this way. Below is an example of what I mean (OperationLevelLockable is defined below): class SomeClass : private OperationLevelLockable { public: int ReadData() { Lock lock( this, READ_OP_GROUP ); return ( mData ); } void WriteData( int data ) { Lock lock( this ); // assign temp group to operation, // so each write // operation blocks all other accesses mData = data; } private: int mData; }; I'm developing on Windows 2000 on a multiprocessor machine. This is the code I use to test my idea: const int32 OPERATION_GROUP_1 = 1; const int32 OPERATION_GROUP_2 = 2; const int NUM_THREADS = 4; class TestOperationLevelLockable : public MTTestCase { public: TestOperationLevelLockable() : mShared( this ), mCurrentGroup( 0 ), mThreadCount( 0 ) { // Create two sets of threads, each performing a // different operation // When the TestOperationLevelLockable object is // constructed all // registered threads are created an start running RegisterThread( this, &TestOperationLevelLockable::Operation1Thread, NUM_THREADS ); RegisterThread( this, &TestOperationLevelLockable::Operation2Thread, NUM_THREADS ); } void CallFromGroup( int32 opGroup ) { boost::mutex::scoped_lock lock( mMutex ); if ( 0 == mCurrentGroup ) { mCurrentGroup = opGroup; } if ( ( mCurrentGroup != opGroup ) && ( mThreadCount < NUM_THREADS ) ) { ReportFailure( "OperationLevelLockable failed to lock out operations of different group", __FILE__, __LINE__ ); } ++mThreadCount; } protected: class MTObject : public OperationLevelLockable { public: MTObject( TestOperationLevelLockable* target ) : mTarget( target ) {} void CallFromGroup( int32 opGroup ) { mTarget->CallFromGroup( opGroup ); } private: TestOperationLevelLockable* mTarget; }; MTObject mShared; int32 mCurrentGroup; int mThreadCount; boost::mutex mMutex; void DoMTTest() { } void Operation1Thread() { MTObject::Lock lock( mShared, OPERATION_GROUP_1 ); mShared.CallFromGroup( OPERATION_GROUP_1 ); SleepThread( 30 ); // seconds, to keep the thread busy } void Operation2Thread() { MTObject::Lock lock( mShared, OPERATION_GROUP_2 ); mShared.CallFromGroup( OPERATION_GROUP_2 ); SleepThread( 30 ); // seconds, to keep the thread busy } }; And here is OperationLevelLockable: class OperationLevelLockable { class AccessCounter; public: OperationLevelLockable(){} typedef Internal::DefaultTMIntType IntType; template<typename T> struct MakeSharedType { typedef T volatile Type; }; class Lock; friend class Lock; class Lock { public: Lock( OperationLevelLockable& obj ) : mCounter( obj.mOpAccessCounter ) { mCounter.EnterOperation( Lock::TempOpGroup() ); } Lock( OperationLevelLockable& obj, int32 opGroup ) : mCounter( obj.mOpAccessCounter ) { mCounter.EnterOperation( opGroup ); } ~Lock() { mCounter.LeaveOperation(); } private: Lock( Lock const& ); Lock& operator=( Lock const& ); static IntType TempOpGroup() { static volatile IntType opGroup = Internal::kFIRST_AUTO_GROUP; IntType newGroup = OperationLevelLockable::AtomicIncrement( opGroup ); if ( newGroup >= Internal::kLAST_AUTO_GROUP ) { OperationLevelLockable::AtomicAssign( opGroup, Internal::kFIRST_AUTO_GROUP ); } return ( newGroup ); } AccessCounter& mCounter; }; static inline IntType AtomicIncrement( volatile IntType& lval ); static inline IntType AtomicDecrement( volatile IntType& lval ); static inline void AtomicAssign( IntType& lval, IntType rVal ); static inline void AtomicAssign( volatile IntType& lval, IntType rVal ); private: class AccessCounter : private boost::base_from_member<boost::recursive_mutex> { public: AccessCounter() : mLock( member ), mRefCnt( 0 ), mOpGroup( Internal::kANY_GROUP ) {} bool IsOpGroupActive( IntType opGroup ) const { return ( ( Internal::kANY_GROUP == mOpGroup ) || ( opGroup == mOpGroup ) ); } void LeaveOperation() { if ( 0 == OperationLevelLockable::AtomicDecrement( mRefCnt ) ) { SetOpGroup( Internal::kANY_GROUP ); mCanOpProceed.notify_all(); } } void EnterOperation( IntType opGroup ) { mCanOpProceed.wait( mLock, CanOpProceed( *this, opGroup ) ); SetOpGroup( opGroup ); OperationLevelLockable::AtomicIncrement( mRefCnt ); } private: AccessCounter( AccessCounter const& ); AccessCounter& operator=( AccessCounter const& ); struct CanOpProceed { CanOpProceed( AccessCounter const& counter, IntType opGroup ) : mOpGroup( opGroup ), mCounter( counter ) {} bool operator()() const { return ( mCounter.IsOpGroupActive( mOpGroup ) ); } private: CanOpProceed& operator=( CanOpProceed const& ); int32 mOpGroup; AccessCounter const& mCounter; }; void SetOpGroup( IntType opGroup ) { OperationLevelLockable::AtomicAssign( mOpGroup, opGroup ); } boost::recursive_mutex::scoped_lock mLock; boost::condition mCanOpProceed; volatile IntType mRefCnt; volatile IntType mOpGroup; }; OperationLevelLockable( OperationLevelLockable const& ); OperationLevelLockable& operator=( OperationLevelLockable const& ); AccessCounter mOpAccessCounter; }; The idea is that the first operation that is executed acquires the lock and blocks out all threads not belonging to the same Operation Group. The original blocking works fine. Only threads executing in the same Operation Group are allowed to continue. When the last thread of the group leaves the operation in OperationLevelLockable::AccessCounter::LeaveOperation it releases all waiting threads and a thread from the other group is allowed to run. However when this thread finishes no new threads are allowed to run anymore and the application hangs. Am I using boost::condition incorrectly. Should I use something else? It the test code wrong? Are there any other errors in the code that prevent it from behaving as expected. Should I use another technique to achieve the same? I hope there is someone out there who has the hart to read through the entire posting and help me. I hope I have given enough information. Thx. Sven Van Echelpoel.

2 4

Building regex with msvc6sp5 and stlport
by John Maddock 25 Feb '02

25 Feb '02

>I have succesfully built the regex library >under msvc6sp5, using the vc6-sltport.mak file. > >But I notice that the compiler gives me warnings >relating to the microsoft stl include directories: > >example: vc98\include\iterator: blah blah blah Then you aren't compiling / linking to STLport >Should it not be linking to the stlport include files? Yes. Did you either * update your INCLUDE and LIB environment variables to point to STLport. or * invoke the makefile using something like: nmake CXXFLAGS="-Ic:\STLport-4.5.1\stlport" -fvc6-stlport.mak - John Maddock http://ourworld.compuserve.com/homepages/john_maddock/

2 1

Thread support?
by John Maddock 24 Feb '02

24 Feb '02

>If I'm using Boost in a single threaded app, do I need to define >BOOST_DISABLE_THREADS to avoid Boost using locks, etc. internally? You don't need to, but it may remove some redundent code, if your app is single threaded *and* links to a multithreaded runtime. - John Maddock http://ourworld.compuserve.com/homepages/john_maddock/

4 4