nvcc throws
/usr/include/boost/concept/detail/has_constraints.hpp:29: error: ‘&’ cannot appear in a constant-expression
/usr/include/boost/concept/detail/has_constraints.hpp:29: error: template argument 2 is invalid
the has_constraints.hpp already has some suspicious code in it:
#if BOOST_WORKAROUND(__SUNPRO_CC, <= 0x580)
// Work around the following bogus error in Sun Studio 11, by
// turning off the has_constraints function entirely:
// Error: complex expression not allowed in dependent template
// argument expression
inline no has_constraints_(...);
#else
template <class Model>
inline yes has_constraints_(Model*, wrap_constraints<Model,&Model::constraints>* = 0);
inline no has_constraints_(...);
#endif
Question: Is this a nvcc - boost incompatibility, or could there be something wrong with my code?
Why do you have the "&" before "Model::constraints"? I would think that's the problem. (// it's not apparently)
EDIT:
http://forums.nvidia.com/index.php?showtopic=182890 talk about this issue and have some hack work arounds
http://forums.nvidia.com/index.php?showtopic=150025
EDIT2:
Well, after running around this for a while, this is what I'm sticking with:
http://forums.nvidia.com/index.php?showtopic=215470 "There is a known compatibility issue with boost and nvcc. A work around is to split the sources such that you compile the cuda code with nvcc and the boost code with the host compiler." by Justin Luitjens in the Group: NVIDIA Employees
Try it out, but if you can't make it work, compile separately, and then link them.
I have a patch here http://code.google.com/p/asadchev/source/browse/trunk/projects/boost/boost-1.46.0.nvcc.patch
Perhaps you can see what is changed and fix your code likewise
Be aware that the source is seen by both, gcc-like and egg-like compiler parts.
What we did at the end was to turn has_constraints_ off as in the example quoted in the question for Sun Studio 11
Related
Here is my code:
#include <iostream>
#include <iomanip>
#include <complex>
#include <cmath>
int main()
{
using namespace std::complex_literals;
std::cout<< std::fixed<<std::setprecision(1);
std::complex<double> z1= 1i *1i; //imaginary unit squared
std::cout<<"i * i= " <<z1 <<'\n';
std::complex<double> z2=std::pow(1i, 2); //imaginary unit squared
std::cout <<"pow(i,2) =" <<z2 <<'\n';
}
I compile with gcc hello.cpp -lstdc++ -o hello.o
Basically it won't let me double the powers of a complex number when i do std::complex z2=std::pow(1i, 2);..
I get the following error
error: no matching function for call to 'pow(complex int, int)'
std::complex z2=std::pow(1i, 2);
However, if i remove the complex number and do std::complex z2=std::pow(2, 2);
it returns 4, the correct answer..
There are many more lines of compile errors, but i made it brief
This answer follows up the comments to the original question:
you have to force the compiler to use the c++14 standard with the -std=c++14 option because the literal operator""i is part of the C++14 spec.
GCC uses c++14 by default since version 6.1. Check your compiler version with gcc -v and refer to this link for GCC standard support.
EDIT:
I was able to reproduce the compiling issue with GCC 6.3 through the link provided by Mr Richard Critten in the comments to the original question, who was the first to point to the correct answer. My apology because I totally overlooked the reference to the C++14 standard.
Anyway, for the sake of clarity, I'm editing this answer, because I've found something that may be interesting to share.
The reason why compiling with GCC 6.3 fails is the fact that the reference standard has been changed in December 2016 from C++14 to GNU++14, see here.
GNU++14 is an extension to the C++ standard, that, among other things, provides additional functions overload for standard APIs.
I've found that with GNU++14 SFINAE fails in finding a proper overload for the std::pow() functions unless the type is explicitly set in the template call like in the snipped below:
std::complex<double> z2=std::pow<double>(1i, 2);
The GNU++14 includes changes to the cmath and complex header files, that I believe are the cause of the issue.
Turning on the C++14 flag, that is not the default anymore, fixes the problem.
I don't know why it is not compiling on your system, it runs fine on mine.
I think something is wrong with your compiler. And, not the version. yet you could try std::cout << __cplusplus. See what this prints.
C++11 doesn't recognize std::complex_literals, so doubt that is the case, but yet the compiler still couldn't find the function.
I honestly don't understand why it is searching for (complex int, int). The compiler is either corrupt or an old beta version.
You should either download a fresh version of the compiler or run it online somewhere. Try adding -std=c++14, but I doubt that would help.
Try your code here(it works):
https://www.jdoodle.com/online-compiler-c++14
So, I installed Microsoft Visual Studio.. I am using the clang compiler and my program ran smoothly with no errors!
Thanks for all the help guys, but i think clang is a better c++ compiler
I'm trying to use shared_ptr from std:: in c++ 11 so I'm including which apparently uses typeinfo. The problem is that I'm getting the error bellow.
In file included from /usr/include/c++/4.8/memory:71:0,
from /home/dev/openclTest/modules/Quant.cpp:10:
/usr/include/c++/4.8/typeinfo:39:37: error: expected ‘}’ before end of line
#pragma GCC visibility push(default)
From what it sounds like, this is a known bug listed here:
https://lists.debian.org/debian-gcc/2007/09/msg00497.html
Anyway, I was hoping I didn't have to use boosts library, so if there's a quick fix anyone knows about, please let me know. I'm using cmake and g++ to build the project.
Thanks!
I just had a similar issue with GCC 4.9.2. It seems to be the common error message in case you forget some braces in a file included prior to typeinfo.
I did not find the code to exactly reproduce the message but got something very similar using this minimal test:
test.cpp:
{
#include <typeinfo>
int main() {}
I guess it is likely the same error as yours as it happens in very different GCC versions. In case it really is the same issue, check your other includes.
So in my CSE course we are given a header file to use right now for our programs that we're writing.
Unfortunately I can't get terminal to compile using that header, it gives quite a few errors (compiling with just 'g++'). Also, when I'm at my university and I'm using PuTTY I get the same errors while using this header. However, I don't get the errors when I compile with 'g++ -std=c++14'.
I've tried compiling with this command on terminal on my mac, but it says it doesn't recognize the c++14 part.
dhcp-10-202-147-243:hw1pr1 Admin$ g++ -std=c++14 hw1pr1.cpp
error: invalid value 'c++14' in '-std=c++14'
Any help on how I could get this to work would be greatly appreciated. Hopefully this all made some sort of sense.
Here's the error I get when I compile with the header file I'm talking about in terminal with just g++.
/Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/../lib/c++/v1/ext/hash_map:212:5: warning:
Use of the header <ext/hash_map> is deprecated. Migrate to <unordered_map>
[-W#warnings]
# warning Use of the header <ext/hash_map> is deprecated. Migrate to ...
^
In file included from read_first_name.cpp:1:
./std_lib_facilities_4.h:43:20: error: no matching function for call to object
of type 'hash<char *>'
return hash<char*>()(s.c_str());
^~~~~~~~~~~~~
/Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/../lib/c++/v1/ext/__hash:39:12: note:
candidate function not viable: 1st argument ('const value_type *'
(aka 'const char *')) would lose const qualifier
size_t operator()(char *__c) const _NOEXCEPT
^
In file included from read_first_name.cpp:1:
./std_lib_facilities_4.h:112:8: warning: comparison of unsigned expression < 0
is always false [-Wtautological-compare]
if (i<0||size()<=i) throw Range_error(i);
~^~
./std_lib_facilities_4.h:118:8: warning: comparison of unsigned expression < 0
is always false [-Wtautological-compare]
if (i<0||size()<=i) throw Range_error(i);
~^~
3 warnings and 1 error generated.
This error doesn't happen and the program will compile fully when I use PuTTY and 'g++ std=c++14'
There's lots of change between C++ standards, so what is valid in one revision need not be in another.
g++ defaults to -std=gnu++98 for C++, which is the decades old C++98-standard enhanced with GNU extensions (most of which are conformant).
Choose the proper revision: -std=c++1y -pedantic is a very close approximation to C++14.
What changes introduced in C++14 can potentially break a program written in C++11?
Looking at what you say you're having to use and the name format of that .cpp file, I think I'm in the same class. A year later, looks like, but here's my solution for archive's sake:
The std_lib_facilities.h header comes with the Bjarne Stroustrup textbook, "Programming: Principles and Practices Using C++". For those unaware, Bjarne Stroustrup invented C++ (he has a pretty good idea what he's talking about). Incidentally, the book is a fantastic way to learn C++, if one takes the time to actually read it. The std_lib_facilities.h header is just a convenient header file for beginners in C++, containing links to all the major standard libraries used in the textbook, as well as some helper functions that help account for potential mistakes or errors, or are just convenient for learning (such as an error() function that handles simple exception throwing for the student, or adding an "out of bounds" check for vectors). It's ultimately just a way to allow students to hop right into code without having to learn specifics about the header.
Stroustrup keeps updated with C++ and thus includes several libraries that require the c++11 standard. The CSCE department wants its students (at least in this early class) to connect to the department's Unix system and compile from there, in order to avoid confusion with downloading and updating compilers.
I happened to already have had a couple C++ classes beforehand, and thus already had g++ set up on my Ubuntu laptop. I avoided including the std_lib_facilities for as long as possible since I was getting the same error as Topic Creator Joe, where g++ didn't recognize the "c++11" part (manually including the required libraries worked fine until we had to use a class from the textbook that used one of the header's helper functions) . Eventually, I found a help topic online that advised me simply to update my g++ compiler to 4.7 or higher, since 4.6 and lower doesn't have support for C++11 (or, of course, C++14). It was oddly rather involved compared to updates one might be used to on Mac or Windows, and I doubt the exact process would apply, but that is (was?) likely the problem: it's just an older version of g++, and it needs an update to compile C++11 and later. I recommend searching for ways to update g++/gcc for Mac.
Should, y'know, anyone else with this problem stumble upon this and not have their problem solved yet.
I was wondering if someone could help me with this.
I've retrieved the source code for the gnuplot-iostream interface from http://www.stahlke.org/dan/gnuplot-iostream/. However, when I attempt to compile the code using the command:
]$ cmake .; make
I get the following compiler error
/.../gnuplot-iostream.h: In constructor ‘Gnuplot::Gnuplot(const std::string&)’:
/.../gnuplot-iostream.h:427: error: ‘never_close_handle’ is not a member of ‘boost::iostreams’
I'm using Scientific Linux 6.2 (kernal 2.6.32-220.23.1.el6.x86_64), g++ 4.4.6, and have boost libraries installed (/usr/include/boost/iostreams/ exists).
Any assistance would be very much appreciated.
D
enum file_descriptor_flags was added in boost::iostreams only in 1.44.0.
enum file_descriptor_flags
{
never_close_handle = 0,
close_handle = 3
};
So, the solution is simply update the boost library (thanks ForEveR).
If however, like me, you do not have the access to update the libraries on your system you should note that I was able to force compilation and obtain basic functionality by simply replacing the two occurrences of boost::iostreams::never_close_handle in the gnuplot-iostream.h file with 0.
I'm currently writing a very simple game engine for an assignment and to make the code a lot nicer I've decided to use a vector math library. One of my lecturers showed me the Sony Vector Math library which is used in the Bullet Physics engine and it's great as far as I can see. I've got it working on Linux nicely but I'm having problems porting it to work on OS X (intel, Snow Leopard). I have included the files correctly in my project but the C++ version of the library doesn't seem to compile. I can get the C version of the library working but it has a fairly nasty API compared to the C++ version and the whole reason of using this library was to neaten the code in the first place.
http://glosx.blogspot.com/2008/07/sony-vector-math-library.html
This blog post that I've stumbled upon seems to suggest something's up with the compiler? It's fairly short so I couldn't take a lot of information from it.
When I try to use the C++ version I get the following errors (expanded view of each error):
/usr/include/vectormath/cpp/../SSE/cpp/vectormath_aos.h:156:0
/usr/include/vectormath/cpp/../SSE/cpp/vectormath_aos.h:156:
error: '__forceinline' does not name a type
second error:
/Developer/apps/gl test/main.cpp:7:0 In file included from /Developer/apps/gl test/main.cpp
/usr/include/vectormath/cpp/vectormath_aos.h:38:0 In file included from
/usr/include/vectormath/cpp/vectormath_aos.h
/usr/include/vectormath/cpp/../SSE/cpp/vectormath_aos.h:330:0 In file included from
/usr/include/vectormath/cpp/../SSE/cpp/vectormath_aos.h
/usr/include/vectormath/cpp/../SSE/cpp/vecidx_aos.h:45:0 Expected constructor, destructor,
or type conversion before '(' token in /usr/include/vectormath/cpp/../SSE/cpp/vecidx_aos.h
Finally two errors at the end of the main.cpp file:
Expected '}' at the end of input
Expected '}' at the end of input
I've Googled my heart out but I can't seem to find any answers or anything to point me in the right direction so any help will be greatly received.
Thanks,
__forceinline is a reserved word that is supported by only a couple compilers. Clearly, your compiler does not support the __forceinline keyword and the code in question is non-portable.
A very poor workaround would be to pass a new define to your compiler that gives the keyword the correct meaning. E.g.: -D__forceinline=inline or -D__forceinline=__attribute__((always_inline)) (Thanks Paul!)
The SSE version was assumed to be only for Microsoft Visual Studio. For other platforms (Mac etc) you can use the scalar version.
Bullet\Extras\vectormathlibrary\include\vectormath\scalar\cpp
It looks like someone's fixed this up and posted a patched version in response to this very issue.
Now GCC compliant.
Which compiler are you using on OS X ? There are 4 to choose from in the standard Xcode 3.2 install and the default is gcc 4.2. You might be better off trying gcc 4.0.