<xlocale>: boolalpha extraction should be case-sensitive

[StephanTLavavej]

Reported by @b7f7 in ea1aaf7#commitcomment-45101146 . Also tracked by Microsoft-internal VSO-1259300 / AB#1259300 .

This is a regression in VS 2019 16.8 caused by GH-1168. I developed a complete test case:

C:\Temp>type meow.cpp

#include <cassert>
#include <ios>
#include <sstream>
using namespace std;

enum class Parse { Success, Failure };

struct TestCase {
    const char* str;
    ios_base::fmtflags flags;
    Parse expected;
    bool result;
};

// clang-format off
constexpr TestCase test_cases[] = {
    {"0",     ios_base::fmtflags{}, Parse::Success, false},
    {"1",     ios_base::fmtflags{}, Parse::Success, true },
    {"2",     ios_base::fmtflags{}, Parse::Failure, true }, // N4868 [facet.num.get.virtuals]/6
    {"WOOF",  ios_base::fmtflags{}, Parse::Failure, false}, // N4868 [facet.num.get.virtuals]/3.6
    {"false", ios_base::boolalpha,  Parse::Success, false},
    {"true",  ios_base::boolalpha,  Parse::Success, true },
    {"WOOF",  ios_base::boolalpha,  Parse::Failure, false}, // N4868 [facet.num.get.virtuals]/7
    {"FALSE", ios_base::boolalpha,  Parse::Failure, false}, // regressed by GH-1168
    {"TRUE",  ios_base::boolalpha,  Parse::Failure, false}, // regressed by GH-1168
};
// clang-format on

int main() {
    for (const auto& test : test_cases) {
        bool val = !test.result;
        istringstream iss(test.str);
        iss.setf(test.flags);
        iss >> val;
        assert(iss.fail() == (test.expected == Parse::Failure));
        assert(val == test.result);
    }
}

C:\Temp>cl /EHsc /nologo /W4 meow.cpp && meow
meow.cpp
Assertion failed: iss.fail() == (test.expected == Parse::Failure), file meow.cpp, line 35

The "FALSE" and "TRUE" cases are asserting because the Standard says that the extraction should fail, but the extraction is succeeding after GH-1168 modified _Getloctxt() to perform case-insensitive comparisons:

STL/stl/inc/xlocale

Line 3274 in 04369fc

} else if (_First == _Last || _CType.tolower(_Ptr[_Off]) != _CType.tolower(static_cast<_Elem>(*_First))) {

[fsb4000]

developer community issue: DevCom-1289669

[StephanTLavavej]

As this is a runtime behavior regression and has been encountered by 2-3 users, I am marking this as high priority.