`<iomanip>`: consider removing `operator>>` for `istream >> setfill(c)`

[achabense]

According to the standard(this version), std::setfill is only for ostreams. Therefore, istream >> setfill(c) has no well-defined meaning and can lead to confusion.

The following piece of code should get removed.

STL/stl/inc/iomanip

Lines 28 to 35 in a621095

	template <class _Elem2, class _Traits>
	friend basic_istream<_Elem2, _Traits>& operator>>(basic_istream<_Elem2, _Traits>& _Istr, const _Fillobj& _Manip) {
	// set fill character in input stream
	static_assert(is_same_v<_Elem, _Elem2>, "wrong character type for setfill");
	_Istr.fill(_Manip._Fill);
	return _Istr;
	}

[fsb4000]

Thanks. I agree.

I noticed that LLVM libc++ doesn't have the operator. But GCC libstdc++ has the operator.

https://gcc.godbolt.org/z/a6xd65xnY

Will you report this bug to GCC developers?

Or should I do that?

[achabense]

Igor Zhukov (@fsb4000) I'm not famaliar with GCC, you may do the report.

[fsb4000]

Sure. I created: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109922

[achabense]

It seems also better to define separate classes as return type for get_* and put_*, and give only matching operator<< or >> for each class, to make the following cases directly unavailable.

#include<iostream>
#include<iomanip>

int main() {
	double _{};
	tm __{};

	//compilable, but comfusing and also not specified by the standard.
	std::cout << std::get_money(_);
	std::cout << std::get_time(&__, "");

	//not compilable due to constness. the constraint is too indirect.
	//std::cin >> std::put_money(_);
	//std::cin >> std::put_time(&__, "");
}