QString::toUtf8() not returning correct UTF-8 representation

QString::toUtf8() is not returning the correct UTF-8 code when the QString's input is not UTF-8. On Windows, the default charset is Windows-1252/cp1252, but QString::toUtf8() should still return a correct UTF-8 representation.

The following snippet can be used to reproduce the issue:

#include <QTextStream>

int main()
{
    QTextStream out(stdout, QIODevice::WriteOnly);
    QTextStream in(stdin, QIODevice::ReadOnly);
    out << "Enter unicode char: " << flush;
    QString line = in.readLine();
    out << "Displays correctly: " << line << ", but not the correct utf-8 code: " << line.toUtf8().toHex() << endl;

    return 0;
}

Unless the QTextStream codec is changed, this will read a character in cp1252 and output it in the same encoding. In between, it should be converted to a UTF-16 QString and output a correct UTF-8 byte sequence.

Example input:

ä

Expected output: 

Displays correctly: ä, but not the correct utf-8 code: c3a4

Actual output on Msys2 / mintty:

Displays correctly: ä, but not the correct utf-8 code: c383c2a4

This is probably correct assuming mintty uses UTF-8 as encoding, but Qt expected cp1252. However, compiled natively and run on the PowerShell, I get

Displays correctly: ä, but not the correct utf-8 code: e2809e

Which makes no sense at all to me.

In fact, even when bypassing the QTextStream magic and reading the raw bytes via QFile, I cannot find a way to properly decode my string:

QFile in;
in.open(stdin, QIODevice::ReadOnly);

// produces ä and e2809e0a
QString line = QString::fromLocal8Bit(in.readLine());

// produces ? and efbfbd0a
// QString line = QString::fromUtf8(in.readLine());

// produces ä and c2840a
// QString line = QString::fromLatin1(in.readLine());

out << line << " " << line.toUtf8().toHex();