Mixing complex numbers and ints

[Gabriel Dos Reis]

Peter Bienstman <Peter.Bienstman@rug.ac.be> writes:

| Hi,
| 
| The following program
| 
| #include <complex>
|  
| int main()
| {
|   std::complex<double> a(1,1);
|   a + 1;

That is bogus. It should read a + 1.0

| } 
| 
| produces 
| 
| c.cpp: In function `int main()':
| c.cpp:6: no match for `std::complex<double>& + int' operator  
| 
| This worked in 2.95.

The library shipped with 2.95 isn't conforming.

| ... Using doubles instead of ints does work.

Yes, by design.

| Is this a true regression or just a correct implementation of stricter
| promotion rules imposed by the standard?

That is not a regression.  It is just a correct implementation of the
Standard which says:

  template<typename T> complex<T> operator+(const complex<T>&, const T&)

So use of that operator triggers template-argument deduction which
doesn't take most standard conversions into account.  You need to be
explicit about type issues.

-- Gaby
CodeSourcery, LLC                       http://www.codesourcery.com