expr.pis produces a type incorrect expr out of a type correct one

[cipher1024]

Prerequisites

• Put an X between the brackets on this line if you have done all of the following:
  • Checked that your issue isn't already filed.
    • Specifically, check out the wishlist, open RFCs,
      or feature requests.
  • Reduced the issue to a self-contained, reproducible test case.

Description

Using expr.pis on a very simple expression resulted in the following error (which I think means some of the intermediate representation for the parser are not removed).

1731:3: unknown declaration '2'
state:
x y : ℕ
⊢ false

Steps to Reproduce

Put the following in a Lean file:

open tactic
example {x y : ℕ}
: false :=
begin
  (do x ← get_local `x,
      p ← to_expr ```(x ≤ y),
      type_check (p.pis [x]),
      trace "success")
end

Expected behavior: [What you expect to happen]

"success" should get printed.

Actual behavior: [What actually happens]

Error:

1731:3: unknown declaration '2'
state:
x y : ℕ
⊢ false

Reproduces how often: [What percentage of the time does it reproduce?]

100%

Versions

You can get this information from copy and pasting the output of lean --version,
please include the OS and what version of the OS you're running.

Lean (version 3.3.1, commit 316d67c3be31, RELEASE)

Mac OS X v 10.13.2

Additional Information

Any additional information, configuration or data that might be necessary to reproduce the issue.

[Kha]

This is not a bug but a feature request for a missing set of APIs that can work with these "special" local constants (called local_decl_refs in the C++ code) derived from a local context in contrast to "regular" locals constructed manually. For reference, here is a simple implementation of pis for this local kind:

namespace tactic
open expr
meta def pis : list expr → expr → tactic expr
| (l::ls) e :=
do local_const uniq pp info _ ← pure l,
   t ← infer_type l,
   e ← pis ls e,
   pure $ pi pp info t (abstract_local e uniq)
| [] e := pure e
end tactic

While this function works with regular locals just as well currently, it is not clear if this is desirable: https://github.com/leanprover/lean/blob/e6a98ffe9c437c157e8b5acb2b9f1e79de66ff1a/src/library/type_context.cpp#L1191. It's probably better to clearly separate these two local kinds.