Confusing Core Lint error

Summary

When working on #17478 (adding a Core pass to maximise shadowing, to sanity-check ghc), I ran across a rather confusingly-worded lint error message. I had a broken pass, which (more-or-less) renamed all uniques to coincide, and ran that pass and Lint over the program

app :: (a -> b) -> a -> b
app g xs = g xs

This generated the following error - notice how lint tells us the Var: xs, but what it is really complaining about is the occurrence of g, which is bound by xs, since they have the same unique (rB7 here). This is rather confusing, since, looking at the name, there is nothing wrong with how xs is bound and used!

*** Core Lint errors : in result of Core plugin:  Shadowing ***
<no location info>: warning:
    [RHS of app{v rB7}[lidx] :: forall @a{tv aDK}[tv] @b{tv aDL}[tv].
                                  (a{tv aDK}[tv] -> b{tv aDL}[tv])
                                  -> a{tv aDK}[tv] -> b{tv aDL}[tv]]
    [in body of lambda with binder a{tv rB7}[tv] :: TYPE{(w) tc 32Q}
                                                      'LiftedRep{(w) d 63A}]
    [in body of lambda with binder b{tv rB7}[tv] :: TYPE{(w) tc 32Q}
                                                      'LiftedRep{(w) d 63A}]
    [in body of lambda with binder g{v rB7}[lid] :: a{tv rB7}[tv]
                                                    -> b{tv rB7}[tv]]
    [in body of lambda with binder xs{v rB7}[lid] :: a{tv rB7}[tv]]
    In the expression: g{v rB7}[lid] xs{v rB7}[lid]
    In the expression: g{v rB7}[lid] xs{v rB7}[lid]
    Substitution: [TCvSubst
                     In scope: InScope {a{tv XBa} b{tv XBb} $trModule{v raK} xs{v rB7}}
                     Type env: [rB7 :-> b{tv XBb}[tv]]
                     Co env: []]
    Mismatch in type between binder and occurrence
    Var: xs{v rB7}[lid]
    Binder type: b{tv XBb}[tv]
    Occurrence type: b{tv XBb}[tv] -> b{tv XBb}[tv]
      Before subst: a{tv rB7}[tv] -> b{tv rB7}[tv]
*** Offending Program ***
app{v rB7}[lidx]
app{v rB7}[lidx]
  = \ @ a{tv rB7}[tv] @ b{tv rB7}[tv] g{v rB7}[lid] xs{v rB7}[lid] ->
      g{v rB7}[lid] xs{v rB7}[lid]

I have a fork of 8.6.5 https://gitlab.haskell.org/brprice/ghc/tree/WIP/better-lint-var-type-msg which changes the error message to explicitly print both the problematic binder and occurrence, which I think makes the error much more obvious:

*** Core Lint errors : in result of Core plugin:  Shadowing ***
<no location info>: warning:
    [RHS of app{v rdF}[lidx] :: forall @a{tv afq}[tv] @b{tv afr}[tv].
                                  (a{tv afq}[tv] -> b{tv afr}[tv])
                                  -> a{tv afq}[tv] -> b{tv afr}[tv]]
    [in body of lambda with binder a{tv rdF}[tv] :: TYPE{(w) tc 32Q}
                                                      'LiftedRep{(w) d 63A}]
    [in body of lambda with binder b{tv rdF}[tv] :: TYPE{(w) tc 32Q}
                                                      'LiftedRep{(w) d 63A}]
    [in body of lambda with binder g{v rdF}[lid] :: a{tv rdF}[tv]
                                                    -> b{tv rdF}[tv]]
    [in body of lambda with binder xs{v rdF}[lid] :: a{tv rdF}[tv]]
    In the expression: g{v rdF}[lid] xs{v rdF}[lid]
    In the expression: g{v rdF}[lid] xs{v rdF}[lid]
    Substitution: [TCvSubst
                     In scope: InScope {a{tv XdI} b{tv XdJ} $trModule{v r7D} xs{v rdF}}
                     Type env: [rdF :-> b{tv XdJ}[tv]]
                     Co env: []]
    Mismatch in type between binder and occurrence
    Binder: xs{v rdF}[lid]
    Occurrence: g{v rdF}[lid]
    Binder type: b{tv XdJ}[tv]
    Occurrence type: b{tv XdJ}[tv] -> b{tv XdJ}[tv]
      Before subst: a{tv rdF}[tv] -> b{tv rdF}[tv]
*** Offending Program ***
app{v rdF}[lidx]
app{v rdF}[lidx]
  = \ @ a{tv rdF}[tv] @ b{tv rdF}[tv] g{v rdF}[lid] xs{v rdF}[lid] ->
      g{v rdF}[lid] xs{v rdF}[lid]

I'm happy to make a merge request if this change is wanted (modulo bikeshedding) - though it is such a trivial change that it may be easier to not round trip via me!

Steps to reproduce

I don't know of a decent way to reproduce without a broken plugin. See comment #17803 (comment 256576).

Environment

• GHC version used: 8.6.5