Functional `using namespace std` causing compilation error with icpc -std=c++0x

[anonymous]

The following two files, boost/functional/hash/detail/float_functions.hpp boost/functional/hash/detail/hash_float.hpp

Contains using namespace std, which can cause problems when compiled with icpc -std=c++11. For example,

#include <boost/thread.hpp>
int main () {}

which will use and instance float_hash_value in boost/functional/hash/detail/hash_float.hpp will emit an overload error about fpclassify is ambitious, as both std::fpclassify and ::fpclassify are perfect match. (tested on Ubuntu 12.04 with icpc 13). This is due to the fact that intel compilers supply its own math.h header and use its own optimized math functions instead of system headers. Change the directive to using std::fpclassify solve the issue .

[Marshall Clow]

(In [81122]) Narrower 'using'; Refs #7470

[Daniel James]

That'll break any libraries which don't have std::fpclassify. It also won't work as it's only one of the two headers, and if the other one is changed it will break the mechanism for detecting which floating point functions are available. It was written that way for a reason.

[Daniel James]

(In [81124]) Hash: Revert r81122. Refs #7470.

I'll get back to this later, it probably requires compiler specific changes.

[Christian Godenschwager <christian.godenschwager@…>]

Any news on this one? A Boost 1.53.0 with icpc 13.0 & C++11 support would be highly appreciated.

[Daniel James]

This should have been fixed in trunk in r81677 and release in r81920.