Agument to constexpr functions

[helios]

I'm not aware of that reason, either.

[JLBorges]

For example, if it were allowed, this would be well-formed program on some implementations, and ill-formed on some others.

template< double > struct X {} ;

void function( X< 1.0 > ) {}

void function( X< 1.00000000001 > ) {}

Even if, for argument's sake, this were to be allowed, it would make writing cross-compilers very hard (A cross-compiler would have to accurately emulate the target platform's floating point environment at compile time.)

A design goal is that entities in namespace std should, to the extent possible, have the same meaning across implementations; code using the standard library should be portable code.