std::numeric_limits<bool>::traps should be true

[Alcaro]

#include <limits>

bool bool_trap() { return std::numeric_limits<bool>::traps; }
bool char_trap() { return std::numeric_limits<char>::traps; }
bool int_trap() { return std::numeric_limits<int>::traps; }

Result: false, true, true https://godbolt.org/z/q94jhqjsT

Expected: true, true, true. true/false traps just as much as 1/0 (if not optimized out - it is UB).

You can argue that division by false promotes to int, therefore it's not an operation on bool that traps, but if so, char is wrong.

Alternatively, one can argue that the entire traps member should be deprecated and removed, since every trapping operation is UB, and reasoning about the behavior of undefined behavior makes no sense.

See also microsoft/STL#5816, where MS-STL asks what traps should do.

[aviralgarg05]

I would like to solve this issue

[frederick-vs-ja]

I'm not sure whether this is invalid because the standard requires std::numeric_limits<bool>::traps to be false ([numeric.special]/3). Perhaps an LWG issue is wanted.

[frederick-vs-ja]

Per the rationale in LWG554, I think std::numeric_limits<char>::traps is wrong. We should make the value false for all non-promoted integer types.