July 2011 - Boost-users - lists.boost.org

[Serialization] Segmentation fault in split load
by Aditya Gadre 31 Jul '11

31 Jul '11

Hi, I am using Boost::Serialization to save and load a custom class using binary archive. I have split save and load functions as I need to do some initialization during loading process. However, I am getting a segmentation fault while loading and I am not very sure of the reason. The segmentation fault occurs while loading a boolean. I would appreciate any help in this regard. I have attached a full back trace and also the header file for the custom class. Thanks, - Aditya

2 2

[units] quantity to/from string / setting value component apart from the units
by Michael Powell 31 Jul '11

31 Jul '11

Units developers/user, Preferably, I would like to get after a quantity's value and/or its units in string format, plumb that through to the .NET Managed environment. The managed plumbing is straightforward enough to research, no problem. Can I use boost::units::to_string(q), where q is declared something like quantity<si::length,float>? And then, what can we expect after calling it? For instance, if q is set to 2.5 * meters, then can we expect "2.5 m"? Also, can I set a quantity value apart from its unit? In other words, if I've got a quantity<si::length,float> q again, I'd like to do something like this: q.value(2.5); or: q = 2.5; Or some way that preserves the current quantity units. Thank you... Regards, Michael

1 0

DateTime from String
by Kraus Philipp 31 Jul '11

31 Jul '11

Hello, I read a string with a date / time format, like: Fri, 29 Jul 2011 13:48:57 +0000 Can I use Boost.DateTime to convert the string into a std::time_t value? How can I do this? Thanks Phil

3 2

asio problem with detecting http header end
by Kraus Philipp 31 Jul '11

31 Jul '11

Hello, I use Boost.ASIO for receiving Twitter JSON data. I use this code: boost::asio::write( p_socket, l_request ); boost::asio::streambuf l_response; boost::asio::read_until(p_socket, l_response, "\r\n\r\n"); So in my optional the l_response variable should be read the full HTTP header. But if I write it to the std::cout I get: HTTP/1.1 200 OK Date: Fri, 29 Jul 2011 20:08:37 GMT Server: hi Cache-Control: max-age=15, must-revalidate, max-age=300 Expires: Fri, 29 Jul 2011 20:13:37 GMT Content-Type: application/json;charset=utf-8 Content-Length: 9763 Vary: Accept-Encoding X-Varnish: 1045274856 Age: 0 Via: 1.1 varnish X-Cache-Svr: smf1-aba-35-sr2.prod.twitter.com X-Cache: MISS Connection: close {"completed_in":0.025,"max_id":97033383951081472,"max_id_str":"97033383951081472","next_page":"?page=2&max_id=9703338395108147 I can change the "until read parameter" to \n, Connection: close or something else, but the read_until method does not stop the reading at the correct chars. Same code on a Wikipedia call runs without problems. Can anyone help me please to receive the correct header? Thanks Phil

4 3

i earned 297.11 today on the internet...
by Lars Schouw 31 Jul '11

31 Jul '11

Hey you...if u would you like to take in 297.11 from your computer try this #bizope# it's very easy

1 0

[units] build up a runtime
by Michael Powell 31 Jul '11

31 Jul '11

Units developers/users, I need some advice. I am fairly confident boost::units is the way for us to go and support a design- or compile-time unit safe calculations environment. However, now for the tricky business of exposing that functionality to the rest of our application. I am in the midst of trying to do this by exposing some Managed C++ classes, plausibly representing each kind of dimension; to get started with I implemented a base Quantity class, and I am extending that to a Length. I would like to include a quantity<si::length> member, for starters, but I am receiving a build error: Error 2 error C4368: cannot define 'si_l' as a member of managed 'CS::Boost::Units::Dimensions::Length': mixed types are not supported c:\source\safetyvalve2011\cs.calculations.managed\units\Length.h 41 CS.Calculations.Managed This might extend to include a quantity<us::length>, for instance; and so on for other unit systems. Whatever .NET class we expose, this will act like the proxy between the application layer and the calculations layer; we might have Dimension-aware UI controls that work on Length, or Quantity, for instance. That sort of thing. Anyone encountered anything like this before? Or have suggestions how we might overcome this hurdle? I'm puzzling this one through in the meantime... Thank you... Regards, Michael

1 0

Using boost interprocess file_lock.timed_lock
by Bill Moo 30 Jul '11

30 Jul '11

Hello List, Could someone please give me a working example of how to use this please. I am obviously doing something wrong as depending on how much I write to the std::ofstream I either end up with a proper file or a file containing a single byte! I would really appreciate your assistance with this as it's driving me to distraction. Here is the approach I am taking. std::ofstream os ; os.open(m_FileName.c_str(), std::ios_base::out | std::ios_base::binary | std::ios_base::app) ; try { if (os.is_open()) { os.setf(std::ios::fixed, std::ios::floatfield) ; os.precision(3) ; if (fileLock.timed_lock(pt += pt::seconds(5))) { try { for (int o = 0; o < 48; o++) { os << "Write my data here....." ; for (int i = 0; i < 51) { os << i ; // But if I output this simple line and I get a 1 byte file and NO errors or exceptions! } } fileLock.unlock() ; } catch (std::ofstream::failure e) { qDebug() << e.what() ; } } else { qDebug() << "No lock on file!" ; } } else { qDebug() << "File not opened!" ; } } catch (ip::interprocess_exception& e) { qDebug() << e.what() ; } catch (std::exception& e) { qDebug() << e.what() ; } os.close(); -- Bill

4 6

vxworks and pre-built boost libraries
by Brian Vandenberg 29 Jul '11

29 Jul '11

I'm attempting to use boost in vxworks 6.5 and I'm looking for information on how to get the pre-built libraries to build (those that can be). Some amount of support for vxworks was added back in 1.38, but vxworks isn't on the list of available/supported target OSs. Are there any resources out there detailing what can be made to work/does work/doesn't work, and what needs to be done to get the workable stuff to build? Thanks for your time. -Brian

2 5

[units] runtime conversion approach
by Michael Powell 29 Jul '11

29 Jul '11

Units developers/user, I need some help understanding unit conversions using boost::units. I see there are conversions.hpp header files. Does this mean that there are built-in unit conversions for fundamental units, or even mainstream derived ones by simply including the conversions and performing the appropriate operation? Can I do something like this, for instance? quantity<si::length> test_m = 1.0 * meter; quantity<us::inch> test_in = test_m * inch / meter; What is the best way to go about setting up conversions, or for that matter, defining a unit system; for instance, the us (english) unit system does not appear to be available? I've read a couple of the old mailing list entries, but I'm not sure I fully understand the examples that I could follow them. Thank you... Regards, Michael

2 1

noncopyable and move semantics
by Nathan Ridge 29 Jul '11

29 Jul '11

Hello, The current implementation of boost::noncopyable prevents derived classes from declaring a defaulted move constructor: #include <boost/noncopyable.hpp> struct S : boost::noncopyable { S(S&&) = default; }; S f(); int main() { S s = f(); } Errors (with gcc trunk): test3.cpp: In function 'int main()': test3.cpp:12:13: error: use of deleted function 'S::S(S&&)' test3.cpp:5:5: error: 'S::S(S&&)' is implicitly deleted because the default definition would be ill-formed: /usr/include/boost/noncopyable.hpp:27:7: error: 'boost::noncopyable_::noncopyable::noncopyable(const boost::noncopyable_::noncopyable&)' is private test3.cpp:5:5: error: within this context The problem is that the default move constructor for S tries to move the noncopyable base subobject, but noncopyable does not have a move constructor (one isn't generated implicitly because a user-defined copy constructor is present). Then it tries to fall back to the noncopyable copy constructor, but that of course is private. If we give the definition of the S move constructor explicitly, all is well because no one is trying to move or copy the noncopyable subobject: struct S : boost::noncopyable { S(S&&) {} }; However, having to define the move constructor explicitly is annoying and a maintenance hazard (e.g. if you add a data member, you have to add it to the move constructor too). If we modify noncopyable to give it a move constructor, the defaulted move constructor now works in derived classes. I think we should either give noncopyable a move constructor, or add a similar class ("noncopyable_but_movable"? - I'm bad with names) that is like noncopyable but also has a move constructor. Thoughts? Regards, Nate.

3 10