I would like opinions on what I suspect is a potential compiler bug, when using a postfix increment operator in a if statement.

Minimal code below:

using IntVector = std::vector<int>;


std::pair<IntVector, IntVector> my_unweave(const IntVector& xs)
{
    std::pair<IntVector, IntVector> result;
    std::size_t counter = 0;
    for (const auto& x : xs) 
    {
        if (counter++ % 2 == 0)           // This line will fail with MSVC 2022 Preview for ARM
        {                                // It will output OOODD / EEEVEN / OOODD / EEEVEN
            printf("EEEVENn");          // (the postfix ++ fails...)
            result.first.push_back(x);
        }
        else 
        {
            printf("OOODDn");        
            result.second.push_back(x);
        }

        // Placing ++counter here instead of in the if test would solve the issue
        // ++counter;
    }
    return result;
}

void foo(const IntVector& xs)
{
    // This version works however
    std::size_t counter = 0;
    for (const auto& x : xs)
    {
        if (counter++ % 2 == 0)
            printf("Evenn");
        else
            printf("Oddn");
    }
}


int main()
{
    foo(IntVector({ 0, 1, 2, 3 }));

     auto u = my_unweave(IntVector({ 0, 1, 2, 3 }));

}

will print:

Even                // foo() works
Odd
Even
Odd
OOODD                // but in my_unweave() we have an inversion
EEEVEN               // (it starts with "OOODD")
OOODD
EEEVEN

I suspect that the line

if (counter++% 2 == 0)

is where it fails.

This is a very rare occurrence and it fails only with *Visual Studio 2022 ARM Preview” (Version 17.4.0 Preview 5.0). What’s more it fails only in release builds and not in debug builds.

So my question is: is that a compiler bug or is this use of the postfix increment operator likely to be a UB?