https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118282
Bug ID: 118282
Summary: GCC fails to identify feasible type conversion scheme
between operands of conditional expression
Product: gcc
Version: 15.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: wangbopku15 at gmail dot com
Target Milestone: ---
Consider the following code:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
struct A
{
explicit A(int);
operator void*() const;
};
void foo(const A& x) {
auto res = 0 ? x : 0;
}
int main(){
A a{5};
foo(a);
}
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The conditional expression inside 'foo' seems feasible. Although conversion
from type 'int' to type 'A' is forbidden due to the explicit declaration of A's
constructor, the conversion from type 'A' to type 'void *' using the overloaded
type conversion operator of 'A' is still fluent when the last operand '0' is
treated as a pointer.
GCC seems not to realize this by giving the following diagnostic:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
<source>: In function 'void foo(const A&)':
<source>:8:19: error: operands to '?:' have different types 'const A' and 'int'
8 | auto res = 0 ? x : 0;
| ~~^~~~~~~
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Clang, ICC, and EDG accept this. Note that clang will normally reject it if the
overloaded 'operator void*()' is removed.
Please see https://godbolt.org/z/v8ecGKPz1
