x ← ¯p; v ← ret〈¯x, ξ〉; ¯z ← ¯r ′ ] η ′ ⊢⊣ [¯x ← ¯p; u ← ret〈¯x, ⊤〉; v ← ret〈fst(u), snd(u) ∧ ξ〉; ¯z ← ¯r ′ ] η ′ where ¯r ′ = ¯r[snd(v)/y, fst(v)/¯x], η ′ = η[snd(v)/y, fst(v)/¯x]. Since do ¯x ← ¯p; ret〈¯x, ξ ↔ χ〉 = do ¯x ← ¯p; ret〈¯x, ⊤〉, we can continu (0)