Refactors across comments yield invalid code

[ysangkok]

Unsure whether this is an hlint or apply-refact bug, so I am reporting it here, extracted from #1642.

It looks related to

• Comment is repeated. mpickering/apply-refact#136

Using apply-refact from git, and hlint from git, on GHC 9.12.2 with a Hackage index from today.

Main.hs:

main = id $
  -- Hello world
  pure ()

% hlint --serialise Main.hs | tee refactors
[( "Main.hs:(1,8)-(3,9): Warning: Redundant id\nFound:\n  id $ pure ()\nPerhaps:\n  pure ()\n"
 , [Replace
       { rtype = Expr
       , pos = SrcSpan
          { startLine = 1
          , startCol = 8
          , endLine = 3
          , endCol = 10 }
       , subts = [("x",SrcSpan {startLine = 3, startCol = 3, endLine = 3, endCol = 10})]
       , orig = "x"}])]

% refactor --refact-file refactors Main.hs
Applying 1 hints
main =
  -- Hello world pure ()

This is an invalid program now, since there is no expression on the rhs.