d69a8eff3f
This PR fixes elaboration of constants in the `rewrite` tactic. previously, `rw [eq_self]` would elaborate `eq_self` twice, and add it to the infotree twice. This would lead to the "Expected type" being delaborated with an unknown universe metavariable. I added a test to show this error during delaboration of the "Expected type". This was reported on Zulip as a panic message during delaboration: [#mathlib4 > Crash in `sup`/`inf` / `max`/`min` delaborators](https://leanprover.zulipchat.com/#narrow/channel/287929-mathlib4/topic/Crash.20in.20.60sup.60.2F.60inf.60.20.2F.20.60max.60.2F.60min.60.20delaborators/with/515946714)
21 lines
768 B
Text
21 lines
768 B
Text
import Lean
|
|
|
|
/-!
|
|
Previously, `rw [my_lemma]` would elaborate `my_lemma` twice, both times generating new universe metavariables.
|
|
This caused the "Expected type" to contain a universe metavariable that wasn't in the metavariable context.
|
|
|
|
This test verifies that the generated universe level is in the metavariable context.
|
|
-/
|
|
|
|
open Lean PrettyPrinter Delaborator SubExpr
|
|
/-- No-op delaborator that checks that the universe level is in the metavariable context -/
|
|
@[delab app.Eq]
|
|
def checkUniv : Delab := do
|
|
let .const _ [.mvar u] := (← getExpr).getAppFn | failure
|
|
_ ← u.getLevel -- if `u` isn't in the metavariable context, this throws an error during elaboration
|
|
failure
|
|
|
|
example : True := by
|
|
try rw [eq_self]
|
|
--^ $/lean/plainTermGoal
|
|
trivial
|