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

How to link boost library DLLs?
by Anand Arumugam 14 Feb '21

14 Feb '21

Hello all, How to properly link boost library DLLs? I have #define-d the following: #define BOOST_ALL_DYN_LINK #define BOOST_DYN_LINK #define BOOST_THREAD_USE_DLL #define BOOST_THREAD_PROVIDES_FUTURE_CONTINUATION #define BOOST_THREAD_VERSION 4 And I still get the following problem: --------------------------- testprogs.exe - System Error --------------------------- The code execution cannot proceed because boost_thread-vc142-mt-gd-x64-1_75.dll was not found. Reinstalling the program may fix this problem. --------------------------- OK --------------------------- I get the same error when I try to run the program using DLLs that I built and while running using the prebuilt binaries. Not sure what step I'm missing. The program runs fine if I build using static libs. The problem is only with DLLs. Any help is much appreciated. Thanks, Anand.

3 3

Boost.Asio as "thread-pool": How to apply back-pressure?
by Dominique Devienne 10 Feb '21

10 Feb '21

Hi, I'm using Boost.ASIO, not for networking, but simply to parallelize work items, with a simple graph of "processing nodes" as detailed below. It's working fine, but uses too much memory. I'd like insights on limiting memory use via "throttling", or what I've also seen called "back-pressure". At a high level, I process two files (A and B) composed of several "entries" each, extracting a subset or a transformation of those entries, that I then write into an output file (C). (those 3 files reach into the many GBs, thus the need for parallelism, and limiting memory use). Extracting entries from A and B are independent operations, implemented single-threaded, producing independent work items (for subsetting or transforming each item), A#1...A#n, and B#1..B#m. That's the "fan-out" part, with each work-item (task) scheduled on any thread of the ASIO pool, since independent. Writing to C is also single-threaded, and needs to "fan-in" the work posted by the A#n and B#m functors, and I serialize that via posting to a C-specific strand (still on any thread, as long as serialized, doesn't matter). Lets call all those tasks writing to C the C#n+m tasks, which are posted to the strand via the A#s and B#s. My issue is that Boost.Asio seems to schedule an awful lots of A# and B# tasks, before getting to the C# tasks, which results in accumulating in memory too many C# tasks, and thus using too much memory. I don't see a way to force more C "downstream" tasks to be scheduled, before processing so much A and B tasks, and accumulating pending C tasks in the work queue, thus using too much memory again. Could someone please recommend a way to have a more balanced flow of tasks in the graph? Or alternative designs even, if what I do above is not ideal. Is Boost.Asio even suitable in this case? Thanks, --DD

7 17

[boost-process] why does my child process not get EOF on stdin
by harald 01 Feb '21

01 Feb '21

Dear community, dear experts! I am trying to use boost::process to call weasyprint to convert HTML to a picture. To do that I wrote a test program (further down in the mail). This worked fine and did what i expected. But it stopped working as soon as I tried to convert two files at the same time by changing new renderer_t(io_context, "<head></head><body></body>"); in main to new renderer_t(io_context, "<head></head><body></body>"); new renderer_t(io_context, "<head></head><body></body>"); Now weasyprint hangs. I changed the call to weasyprint to weasyprint2, which is a short bash script for testing purposes: --------------------------------------------------------------------------------- #! /bin/bash cat > /dev/null echo "This should be a PNG picture" --------------------------------------------------------------------------------- It reads stdin until it gets EOF and then writes something to stdout before terminating. Same result. It seems that my program does not close stdin of the child, so it never gets EOF on stdin. But only when more than one child process is running at the same time. So obviously I am doing something wrong, but I have no idea what! Does anybody have a hint for me? I am on xubuntu 20.10 compiler is g++ (Ubuntu 10.2.0-13ubuntu1) 10.2.0, boost version 1.75.0 compiler command line (from my IDE): g++ -c main.cpp -g -O0 -Wall -Wextra -std=c++17 -o build-Debug/main.cpp.o -I. -I. --------------------------------------------------------------------------------- #include <boost/smart_ptr/intrusive_ptr.hpp> #include <boost/smart_ptr/intrusive_ref_counter.hpp> #include <boost/process.hpp> #include <boost/asio/write.hpp> #include <iostream> namespace bp = boost::process; bool is_eof( std::error_code error) { boost::system::error_code eof = boost::asio::error::eof; return error == eof; } struct renderer_t : public boost::intrusive_ref_counter<renderer_t> { using pointer = boost::intrusive_ptr<renderer_t>; boost::asio::io_context &io_context; const std::string html; bp::child child; bp::async_pipe in; bp::async_pipe out; bp::async_pipe err; std::array<unsigned char, 256> outbuf; std::array<char, 256> errbuf; std::vector<unsigned char> result; std::error_code result_error; std::string errmsg; ~renderer_t() { std::cout << "renderer_t::~renderer_t\n"; if (!errmsg.empty()) std::cerr << "renderer_t::~renderer_t: stderr: " << errmsg << "\n"; } renderer_t( boost::asio::io_context &io_context, const std::string &html) : io_context(io_context), html(html), in(io_context), out(io_context), err(io_context) { post( io_context, [self = pointer(this)]() { self->start(); }); } void start() { std::error_code ec; std::cout << "renderer_t::start\n"; bp::child c( "weasyprint2 -m screen -p -f png - /dev/stdout", bp::std_in < in, bp::std_out > out, bp::std_err > err, bp::on_exit=[self = pointer(this)]( int exit, std::error_code error) { self->on_exit(error, exit); }, bp::start_dir="/data/build/", io_context, ec); if (ec) { std::cerr << "renderer_t::start: creating child failed: " << ec.message() << "\n"; if (!result_error) result_error = ec; return; } child = std::move(c); boost::asio::async_write( in, boost::asio::buffer(html), [self = pointer(this)]( std::error_code error, std::size_t bytes) { self->on_write(error, bytes); }); out.async_read_some( boost::asio::buffer(outbuf), [self = pointer(this)]( std::error_code error, std::size_t bytes) { self->on_read_out(error, bytes); }); err.async_read_some( boost::asio::buffer(errbuf), [self = pointer(this)]( std::error_code error, std::size_t bytes) { self->on_read_err(error, bytes); }); } void on_write( std::error_code error, std::size_t bytes) { std::cout << "renderer_t::on_write\n"; if (error) { std::cerr << "renderer_t::on_write: failed: " << error.message() << "\n"; if (!result_error) result_error = error; return; } if (bytes != html.size()) { std::cerr << "renderer_t::on_write: short write\n"; } boost::system::error_code ec; in.close(ec); if (ec) { std::cerr << "renderer_t::on_write: close failed: " << error.message() << "\n"; } } void on_read_out( std::error_code error, std::size_t bytes) { std::cout << "renderer_t::on_read_out\n"; if (is_eof(error)) { std::cout << "renderer_t::on_read_out: EOF\n"; return; } else if (error) { std::cerr << "renderer_t::on_read_out: failed: " << error.message() << "\n"; if (!result_error) result_error = error; return; } else { result.insert( end(result), begin(outbuf), begin(outbuf) + bytes); out.async_read_some( boost::asio::buffer(outbuf), [self = pointer(this)]( std::error_code error, std::size_t bytes) { self->on_read_out(error, bytes); }); } } void on_read_err( std::error_code error, std::size_t bytes) { std::cout << "renderer_t::on_read_err\n"; if (is_eof(error)) { std::cout << "renderer_t::on_read_err: EOF\n"; return; } else if (error) { std::cerr << "renderer_t::on_read_err: failed: " << error.message() << "\n"; if (!result_error) result_error = error; return; } else { std::string e(begin(errbuf), begin(errbuf) + bytes); std::cerr << "renderer_t::on_read_err: from stderr: " << e << "\n"; errmsg.insert( end(errmsg), begin(errbuf), begin(errbuf) + bytes); err.async_read_some( boost::asio::buffer(errbuf), [self = pointer(this)]( std::error_code error, std::size_t bytes) { self->on_read_err(error, bytes); }); } } void on_exit( std::error_code error, int exit) { std::cout << "renderer_t::on_exit\n"; if (error) { std::cerr << "renderer_t::on_exit: error: " << error.message() << "\n"; if (!result_error) result_error = error; return; } if (exit) { std::cerr << "renderer_t::on_exit: weasyprint returned " << exit << "\n"; if (!result_error) result_error = make_error_code(std::errc::no_such_process); } } }; int main() { boost::asio::io_context io_context; bp::async_pipe in(io_context); new renderer_t(io_context, "<head></head><body></body>"); io_context.run(); return 0; } --------------------------------------------------------------------------------- Regards Harald

1 0

TOMS748 question
by Shriramana Sharma 29 Jan '21

29 Jan '21

Namaste. This is wrt the TOMS748 math root finding implementation: Given that `if(fa == 0)` is handled at: https://github.com/boostorg/math/blob/develop/include/boost/math/tools/toms… … and the function returns in case of the above condition, it is not clear to me what the need is for the `if (fa != 0)` at: https://github.com/boostorg/math/blob/develop/include/boost/math/tools/toms… Note that I do not find the condition being applied at the SciPy implementation of this algorithm at: https://github.com/scipy/scipy/blob/v1.6.0/scipy/optimize/zeros.py#L1220 Now I also note similar checks at: https://github.com/boostorg/math/blob/develop/include/boost/math/tools/toms… https://github.com/boostorg/math/blob/develop/include/boost/math/tools/toms… … but regarding those lines, while I can understand the need for checking there after more bracketing work is done, it is not clear to me why `(fa == 0)` is alone checked for, and not `(fb == 0)` as well. I would have thought the algorithm is not biased towards a or b. Indeed, in the last steps from: https://github.com/boostorg/math/blob/develop/include/boost/math/tools/toms…, … both the conditions are checked for, which is what I would expect even in the intervening steps. Now the lines seem to have originated in the very first commit implementing this algorithm: https://github.com/boostorg/math/commit/79a8199f9c26ba353fb52ec4543f75ffc2a… … but I do not see a way to contact the author jzmaddock other than posting to this list, so I hope someone/they see here and reply. Thanks! -- Shriramana Sharma ஶ்ரீரமணஶர்மா श्रीरमणशर्मा 𑀰𑁆𑀭𑀻𑀭𑀫𑀡𑀰𑀭𑁆𑀫𑀸

2 7

Correct way to handle error in boost::filesystem::recursive_directory_iterator::increment()?
by Dominique Pellé 23 Jan '21

23 Jan '21

Hi I see code like this in an application to iterate over files in a directory with boost::filesystem (v1.70) without using exceptions: boost::filesystem::path dir("somedir"); boost::system::error_code error; boost::filesystem::directory_iterator node(dir, error), end; if (error) { return ...; } for (; node != end; node.increment(error)) { const boost::filesystem::path& path = node->path(); // ... } It appears to work... or mostly work since a crash was reported and the stack points to this area of the code. It's a rare crash that I could not reproduce myself. I suspect that there is a bug in the above code and it should instead be: boost::filesystem::path dir("somedir"); boost::system::error_code error; boost::filesystem::directory_iterator node(dir, error), end; for (; !error && node != end; node.increment(error)) { const boost::filesystem::path& path = node->path(); // ... } In other words, I suspect that we need to check that node.increment(error) in the for loop did not give an error before checking the condition "node != end", or else we may enter the loop in case of error and then it may access node->path() with an invalid node iterator. I could not find in the doc at ... https://www.boost.org/doc/libs/1_75_0/libs/filesystem/doc/reference.html ... what exactly happens when node.increment(error) gives an error. Perhaps the doc deserves clarification or an example illustrating how to reliably iterate over files without using exceptions would be welcome. Thanks Dominique

1 1

Serialization and static Libraries
by Wolfgang Meyerle 20 Jan '21

20 Jan '21

Hi, probably has been asked already thousand of times but I'm stuck currently with some code: Assuming I have the following files (The Base class Action is serialized in the same way): File: Ping.h #include"Action.h" classPing:publicAction{ public: voidoperator()(); Ping(); Ping*clone()const; ~Ping(); template<typenameArchive> voidserialize(Archive&ar,constunsignedint); }; BOOST_CLASS_EXPORT_KEY2(Ping,"Ping"); File: Ping.cpp #include"Ping.h" #include<iostream> #include<boost/serialization/base_object.hpp> #include<boost/serialization/export.hpp> #include<boost/serialization/tracking.hpp> #include"Action.h" #include"system.h" #include"pong.h" #include"serializerdeserializer.h" #include"system.h" #include"connection.h" #include"communication.h" voidPing::operator()(){ } Ping::Ping() { } Ping*Ping::clone()const{ Ping*ping=newPing(*this); returnping; } template<typenameArchive> voidPing::serialize(Archive&ar,constunsignedint){ ar&BOOST_SERIALIZATION_BASE_OBJECT_NVP(Action); } Ping::~Ping(){} BOOST_CLASS_EXPORT_IMPLEMENT(Ping); BOOST_CLASS_TRACKING(Ping,boost::serialization::track_never); If I now compile the project into a library. And link this library to a different project. If I try to serialize the class now it doesn't work. I get the following error: unregistered class - derived class not registered or exported According to the boost documentation I need to use BOOST_CLASS_EXPORT_KEY in the header file and BOOST_CLASS_EXPORT_IMPLEMENT in the corresponding cpp file. What did I miss? Regards, Wolfgang

1 0

[Boost-MPI][Boost-ODEINT] Running odeint within MPI-context without using mpi_state
by Roland Richter 19 Jan '21

19 Jan '21

Hei, I am currently using the odeint-solver for calculating the right hand side of a non-linear equation. For speeding up the calculation of the right hand side, I decided to test the MPI-implementation. Now, after I would prefer to keep the code backwards compatible, I did not want to replace state_type with mpi_state, and instead split up the initial vector by hand before calling odeint without telling odeint that I am operating within an MPI context. I take care of all communications which involve the vectors, and therefore I assumed that this approach should work (especially after it worked correctly in small test programs). Nevertheless, I encountered the issue that after a certain amount of steps the step size between the different threads starts to differ by a small amount, even though I am in the same iteration. As example: step number rank step size n 1 0.00025 n 0 0.00025 //Ok n + 1 1 0.00051152 n + 1 0 0.000511523 //Not ok I checked my program with valgrind for possible memory leaks/corruption which could overwrite something, but nothing relevant came up. The problem is repeatable. Therefore, is the general idea correct at all, or am I doing something wrong here by neglecting a possible communication between the different threads within odeint? Thanks! Regards, Roland

2 2

Conversion numeric::array in numpy::ndarray
by philippe DALET 17 Jan '21

17 Jan '21

Hello, It's my first mail on this maillist. I started to upgrade a package for python 3.7, 3.8, 3.9 (linux and windows): vpython-wx. One file gives me some problems num_util.hpp My context: ubuntu 20.04: boost 1.71 python3.7 windows 10: boost 1.75 python3.7, 3.8, 3.9 I want to create a double_array class: Original version --------------------- namespace cvisual { namespace python { using boost::python::numeric::array; class double_array : public array { public: explicit double_array( const boost::python::handle<>& h ) : array(h) {} }; array makeNum(npy_intp n, NPY_TYPES t =NPY_DOUBLE); New version ----------------- namespace cvisual { namespace python { using boost::python::numpy::ndarray; class double_array : public ndarray { public: explicit double_array( const boost::python::handle<>& h ) : ndarray(h) {} }; ndarray makeNum(npy_intp n, NPY_TYPES t =NPY_DOUBLE); error message with VC2019 community error conversion 'const boost::python::handle<PyObject>' in 'boost::python::detail::borrowed_reference' error message with gcc 9.3 no matching function for call to ‘boost::python::numpy::ndarray::ndarray(const boost::python::handle<>&)’ Thanks a lot for your help. Philippe DALET Laboratoire de Physique Lyp Champollion, Figeac FRANCE

1 0

Re: [Boost-users] boost::regex: regex meta data or AST for regex?
by Frank Bergemann 16 Jan '21

16 Jan '21

On 16.12.20 18:05, Frank Bergemann via Boost-users wrote: > in concrete: > Is it possible to not only get the #no of groups (sub-expressions) via > basic_regex::mark_count(), but also their position (begin() / end() > iterators) in the given regex string (c'tor argument)? > > more general: > Can an AST be created for a the given regex string (c'tor argument)? > E.g with boost::spirit? I am busy with #2 parsers: 1) RegexParser: Intended to be a full regular expression parser https://github.com/F-Bergemann/RegexParser 2) RegexSplitter: to just break up regular expression into substrings for top level regex-group & non-regex-group substrings: https://github.com/F-Bergemann/RegexSplitter It's just a start and so much wrong within, i know. But i have one concrete question for the RegexSplitter: How can i avoid in RegexSplitter.cpp to use #157 ( tok_element >> *tok_element ) [ qi::_val = phx::new_<ASTNode> (qi::_1, qi::_2, ASTNode::COLLECTION_c) ]; I want to pass a SINGLE std::vector<ATSNode *> only to the receiving #51 // COLLECTION_c #52 ASTNode::ASTNode( #53 ASTNode const * first, #54 std::vector<ASTNode *> & others, #55 Type const type) without the SEPARATE parameter 'first'. I.e. i need to "merge" qi::_1 and qi::_2.

1 0

Re: [Boost-users] boost ptr_map memory access violation
by roland kapl 08 Jan '21

08 Jan '21

Dear All! After finding some inspiration on https://stackoverflow.com/questions/4088387/how-to-store-pointers-in-map , I've succeeded storing the property tree pointers in a std::map... std::string TreeizeRelD::writeTree(const std::vector<std::vector<std::string>> &control, const std::vector<std::vector<std::vector<std::string>>> &data, pt::ptree & ptTree) { // now the map is local... typedef std::map<std::string, pt::ptree*> treeMap; std::map<std::string, treeMap> parentTableTrees; .... std::string subRootNode = parentNode + (subnodeOfParent != "" ? "." + subnodeOfParent : ""); std::string tableLookup = subRootNode + "." + rootElemRec; // simply add records if no foreign key name exists (usually the top (root) table) if (foreignKeyName == "") { for (std::pair<std::string, pt::ptree> && keyPair : tables[tableLookup]) { parentTableTrees[subRootNode].insert(std::make_pair(keyPair.first, &ptTree.add_child(subRootNode, keyPair.second))); ... // iterate through all parent key record collections referred to by subtables parentNode (rootnode + optional subnode) for (std::pair<std::string, pt::ptree*> && parentKeysPair : parentTableTrees.find(lookupNode)->second) { // iterate through all records (ptrees) in key collection for (pt::ptree::iterator parentRecPair = parentKeysPair.second->begin(); parentRecPair != parentKeysPair.second->end(); ++parentRecPair) { // important here: assign a reference to parentRecPtree (pointer!), otherwise a copy is created! pt::ptree &parentRecPtree = parentRecPair->second; ... parentTableTrees[subRootNode].insert(std::make_pair(rowsFK, &parentRecPair->second.get_child(subnodeFRec))); ... parentTableTrees[subRootNode].insert(std::make_pair(rowsFK, &parentRecPtree.put_child(subnodeOfParent, foreignRecordset))); The complete source of this: https://github.com/rkapl123/OreControl/blob/master/OreMgr/TreeizeRelD/treei… -regards, Roland ---------------------------------------------------------------------- Message: 1 Date: Wed, 6 Jan 2021 11:49:07 +0000 From: roland kapl <roland.kapl(a)hotmail.com<mailto:roland.kapl@hotmail.com>> To: "boost-users(a)lists.boost.org<mailto:boost-users@lists.boost.org>" <boost-users(a)lists.boost.org<mailto:boost-users@lists.boost.org>> Subject: [Boost-users] boost ptr_map memory access violation (stack/heap problem) Message-ID: <AM0PR03MB424285F7832EDC3C73C96860EDD00(a)AM0PR03MB4242.eurprd03.prod.outlook.com<mailto:AM0PR03MB424285F7832EDC3C73C96860EDD00@AM0PR03MB4242.eurprd03.prod.outlook.com>> Content-Type: text/plain; charset="us-ascii" Dear Boost-Experts! I'm trying to use a ptr_map along with boost property trees. I need to store pointers to specific parts of a built up property tree to access these parts later on for easy insertion there. However, when the ptr_map is cleared/destroyed, I get memory access violations/stack overflow messages for my test cases. The ptr_map (typedef treeMap) is globally defined within a std::map to a ptr_map to property trees as follows (I know this isn't good practice, but passing it as a reference didn't make a difference, so I left it that way): namespace pt = boost::property_tree; typedef boost::ptr_map<std::string, pt::ptree> treeMap; std::map<std::string, treeMap> parentTableTrees; The property tree is defined as follows in this function (this builds a property tree from tabular data in the passed table variable "data" using definitions form table variable "control"): std::string TreeizeRelD::writeTreeAndCreateXML(const std::vector<std::vector<std::string>>& control, const std::vector<std::vector<std::vector<std::string>>>& data, int *result) { std::string resultString; pt::ptree &propTree = pt::ptree(); std::string returnStr = TreeizeRelD::writeTree(control, data, propTree); *result = 0; if (returnStr != "") { *result = 1; return returnStr; } returnStr = TreeizeRelD::createXML(propTree, resultString); if (returnStr != "") { *result = 1; return returnStr; } // here the memory access violation occurs: parentTableTrees.clear(); return resultString; } In function writeTree the propTree (being passed by ref as ptTree) is populated in following three lines: std::string TreeizeRelD::writeTree(const std::vector<std::vector<std::string>> &control, const std::vector<std::vector<std::vector<std::string>>> &data, pt::ptree & ptTree) ... // FIRST LINE parentTableTrees[subRootNode].insert(keyPair.first, &ptTree.add_child(subRootNode, keyPair.second)); ... // iterate through all parent key record collections referred to by subtables parentNode (rootnode + optional subnode) for (treeMap::value_type && parentKeysPair : parentTableTrees.find(lookupNode)->second) { // iterate through all records (ptrees) in key collection for (pt::ptree::iterator parentRecPair = parentKeysPair.second->begin(); parentRecPair != parentKeysPair.second->end(); ++parentRecPair) { pt::ptree &parentRecPtree = parentRecPair->second; ... // SECOND LINE // parentRecPair is an iterator variable from the ptr_map: parentTableTrees[subRootNode].insert(rowsFK, &parentRecPair->second.get_child(subnodeFRec)); ... // THIRD LINE // parentRecPtree is the referenced property tree itself: parentTableTrees[subRootNode].insert(rowsFK, &parentRecPtree.put_child(subnodeOfParent, foreignRecordset)); I already have the suspicion that the memory access violation comes from ptr_map taking ownership of the pointers (references) passed to it in the above three lines and that these are all created on the stack and not the heap, where ptr_map expects them to be. When the ptr_map is finally destroyed, it tries to access memory, that's impossible to access... But I didn't manage to successfully create heap objects that also do NOT copy the property tree objects, which is important as I only need references to the property trees. Any help is appreciated. -regards, Roland

1 0