[Bug c++/31780] [4.2/4.3 regression] ICE with incompatible types for ?: with "complex type" conversion

[mmitchel at gcc dot gnu dot org]

------- Comment #7 from mmitchel at gcc dot gnu dot org  2007-07-07 22:21 -------
I've attached a patch which fixes this bug in an obvious way.

Since complex types are arithmetic types in GNU C++, we should allow standard
conversions to them from integers, just as we do for all other arithmetic
types.

However, this runs into problems with libstdc++.  In particular,
std::complex<double> has a constructor from double and also a constructor from
__complex__ double.  Making the change in this patch makes that conversion
ambiguous because now "std::complex<double>(1)" can go via either the
"__complex__ double" constructor or the plain "double" constructor.

I'm pretty sure that we did indeed discuss this at some point in the past,
although I couldn't find a link and I don't remember what we decided, if
anything.  

The easiest way to fix this is probably to add more constructors to
std::complex<double> to match all of the arithmetic types directly, e.g., add
"std::complex<double>::complex(int, int = 0)".  Then, that will be the best
match. 

I think we need input from the libstdc++ maintainers before trying to do
anything in the front end.  Paolo?


-- 

mmitchel at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pcarlini at suse dot de
             Status|ASSIGNED                    |WAITING


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31780