c7d3401417
This PR fixes an issue where go-to-definition would jump to the wrong location in presence of async theorems. While the elaborator does not explicitly depend on `FVar`s not being reused between declarations, the language server turned out to do so. As we would have to split the name generator in any case as soon as we add any parallelism within proofs, we now do so for any async code in order to uphold this invariant again. --------- Co-authored-by: mhuisi <mhuisi@protonmail.com>
31 lines
598 B
Text
31 lines
598 B
Text
import Lean.Elab.Tactic.Basic
|
|
|
|
/-!
|
|
Kernel errors should not lead to follow-up errors but should be detectable using `#print axioms`.
|
|
-/
|
|
|
|
/-- error: (kernel) declaration has metavariables 'bad' -/
|
|
#guard_msgs in
|
|
def bad : Empty := by
|
|
run_tac do Lean.Elab.Tactic.popMainGoal
|
|
|
|
theorem zero_eq_one : 0 = 1 := bad.elim
|
|
|
|
/--
|
|
info: def bad : Empty :=
|
|
?_uniq.4
|
|
-/
|
|
#guard_msgs in
|
|
set_option pp.raw true in
|
|
#print bad
|
|
|
|
/--
|
|
info: theorem zero_eq_one : 0 = 1 :=
|
|
Empty.elim bad
|
|
-/
|
|
#guard_msgs in
|
|
#print zero_eq_one
|
|
|
|
/-- info: 'zero_eq_one' depends on axioms: [bad] -/
|
|
#guard_msgs in
|
|
#print axioms zero_eq_one
|