Ban mistyped wrapping arithmetic?

[thomastanck]

I think %- (mod with negated RHS) with no space in between should be a parse error, since the user most likely intended -% and there is a chance (albeit small) that it can parse and run.

This came from a question posted on #zig-help on the Discord community:

    error: negation of type 'u8'
const new = old %- amount;

The poster intended -% but the transposed %- parses as well, and instead the poster got an error about illegally negating amount.

I also managed to come up with a test that causes the unintuitive behaviour to parse and run:

pub fn comptimeWrappingSub(comptime a: comptime_int, comptime b: comptime_int) comptime_int {
    return a %- b;
}

pub fn main() u8 {
    return comptimeWrappingSub(1, -2);
}
// returns 1 instead of 3?

It's surprisingly hard to hit this case, because if a and b were signed, you get an error about needing to use @rem or @mod instead. But here it passes through fine since both a and b are comptime_int.

[jedisct1]

Isn't zig fmt immediately going to turn %- into % -?

With proper highlighting, both symbols are even of different color.

[ifreund]

There's a precedent for disallowing confusing (lack of) whitespace like this: #6823.

[Vexu]

This now results in:

error: binary operator `%` has whitespace on one side, but not the other.
    return a %- b;
             ^