Currently default values that are zero are not displayed in usage.

Value interfaces whose String() string method doesn't output "0", "", "", or "false" for their uninitialized value aren't considered zero valued.

There are two simple ways that I see to solve this. One is to use Flag.DefValue instead of Flag.Value.String() in the nested switch statement in the default case. This appears to be the most consistent way to program this as all of the other cases are comparing Flag.DefValue with their type's zero value. Of course Flag.DefValue is set from Flag.Value.String() when the flag is first initialized, so this should have no effect on existing code.

I personally feel that the use of Flag.Value.String() in the nested switch statement is incorrect code, but it's a bit subjective.

Using Flag.DefValue instead is a small change, but it will allow a user to override Flag.DefValue to the empty string if they wish to hide the "(default %v)" in the usage description for the flag.

The other way, more obtrusive but also more flexible way, to achieve this is by adding an explicit option to the Flag struct that will hide the default value and skip the zero value check altogether.

Something like

type Flag struct {
...
	HideDefaultValue bool
}
...
func (f *FlagSet) FlagUsagesWrapped(cols int) string {
...
		if !flag.HideDefaultValue && !flag.defaultIsZeroValue() {
			if flag.Value.Type() == "string" {
				line += fmt.Sprintf(" (default %q)", flag.DefValue)
			} else {
				line += fmt.Sprintf(" (default %s)", flag.DefValue)
			}
		}