Richard Smith wrote:
I was recently bitten by bug in some code using boost::optional. A simple test case is:
#include <boost/optional.hpp> #include <cassert>
int main() { boost::optional<int> i = 0; assert( i == 0 ); }
Somebody had written code like this, presumably assuming the second argument to the comparison operator would convert implicitly to a boost::optional<int> and the code would behave as if it read:
assert( i && i.get() == 0 );
OK, deep comparison operators have been requested before, but all this time I remained unconvinced there were really needed, until now. As you said, optional<> provides value-based semantics in other places, so given the kind of bug you discovered, I think is time to add these too. Best Fernando Cacciola