This isn't a bug per se, but it's a bug pattern that tools could help with.

Given this code

type T struct {
     F *int
}

Let's say you change this exported field F into a private field f plus an exported
accessor, like so:

type T struct {
    f *int
}
func (t *T) F() *int { return t.f }

Now you have to rename all references to T.F into T.F().  Mostly you can rely on
compiler errors to help you find what you need to change... but not always.  Consider:

  if t.F == nil { ... }

Thanks to the new feature whereby t.f evaluates to a method closure, this statement is
still valid after the change, but is now a provably false condition, so the statement
has no effect.  This is not what the user intended.

It seems to me that a comparison of f==nil, where f is the name of a method or function,
or an expression that creates a bound function, is almost surely a mistake.

We could make it a compiler error, or we could make go vet report it.