Hello.

While reading the operator overloading chapter of Thinking in C++ I came 
across the author saying that if there exist two separate definitions 
for type conversion between two user-defined types, one being an 
operator and the other a constructor, the compiler gives an error. But 
when I tried this, I found that the compiler (at least GCC) does not 
raise an error and prefers the constructor to the operator.

Please peruse the attached code. You need to compile with -DALLOWCONS to 
enable the constructor:

g++ -o ambiguous ambiguous-type-conversion.cpp -DALLOWCONS
./ambiguous-type-conversion
g++ -o ambiguous ambiguous-type-conversion.cpp
./ambiguous-type-conversion

You observe that the constructor is used when present, in preference to 
the operator. Is there a reason for this? I mean I would expect at least 
a warning in such a case, like I get when I initialize a variable with 
the extern keyword.

Shriramana Sharma.

P.S: I had to resort to somewhat ugly hacks to actually get the two 
user-defined classes to define the constructor and operator -- the 
example provided in the book is incomplete. One is the forward 
declaration of the second class before the first class. Two is the 
friend declaration of the second class within the first class. Three is 
the definition of the operator of the first class *after* the second class.