sign extending in shift operators in rust

I am reading rust programming language using book Programming Rust 2nd edition. Here in context of right shift operators and left shift operators mentioned as below

> Bit shifting is always sign-extending on signed integer types and zero-extending on unsigned integer types

My question is what is "always sign-extending"? If possible, provide an example. How overflowing and underflowing is handled in shift operators in Rust.

Thanks in advance for your time and help.

Signed:
  64 (0b01000000) >> 2 = 16 (0b00010000) // filled with 0, because sign is positive
  -64 (0b11000000) >> 2 = -16 (0b11110000) // filled with 1, because sign is negative

Unsigned:
  64 (0b01000000) >> 2 = 16 (0b00010000) // filled with 0, because unsigned
  192 (0b11000000) >> 2 = 48 (0b00110000) // filled with 0, because unsigned

With that way of padding, it is always true that a >> 1 is equal to a / 2, even for negative numbers. Or in our example, a right shift of 2 is equal to a division through 4.

Left shifts (<<) are always padded with 0. This again causes << 1 to be equal to * 2, no special case for negative numbers is needed here.

For more information about how integers are represented in binary, read about the two's complement.

Source:

fn main() {
    println!("Signed:");
    println!(
        "  {} ({:#010b}) >> 2 = {} ({:#010b}) // filled with 0, because sign is positive",
        64i8,
        64i8,
        64i8 >> 2,
        64i8 >> 2
    );
    println!(
        "  {} ({:#010b}) >> 2 = {} ({:#010b}) // filled with 1, because sign is negative",
        -64i8,
        -64i8,
        -64i8 >> 2,
        -64i8 >> 2
    );

    println!();
    println!("Unsigned:");
    println!(
        "  {} ({:#010b}) >> 2 = {} ({:#010b}) // filled with 0, because unsigned",
        64u8,
        64u8,
        64u8 >> 2,
        64u8 >> 2
    );
    println!(
        "  {} ({:#010b}) >> 2 = {} ({:#010b}) // filled with 0, because unsigned",
        (-64i8) as u8,
        (-64i8) as u8,
        (-64i8) as u8 >> 2,
        (-64i8) as u8 >> 2
    );
}