Adding a field ci_cstr_nargs to case_info and mind_consnrealargs to

one_inductive_body so that when eta-expanding at "match" printing time
we know if a let is part of the expected signature or part of the
body.

This is an easy fix for bugs like #3293. Another fix could be to
enforce, as an invariant, or better syntactically, that
"match"/"Case"'s have the body of their branches expanded.

1 files changed, 2 insertions, 1 deletions

diff --git a/kernel/constr.mli b/kernel/constr.mli
index f2d9ac9f5..82a2de094 100644
--- a/kernel/constr.mli
+++ b/kernel/constr.mli
@@ -25,7 +25,8 @@ type case_printing =
 type case_info =
   { ci_ind        : inductive;
     ci_npar       : int;
-    ci_cstr_ndecls : int array; (** number of real args of each constructor *)
+    ci_cstr_ndecls : int array; (* number of pattern vars of each constructor (with let's)*)
+    ci_cstr_nargs : int array; (* number of pattern vars of each constructor (w/o let's) *)
     ci_pp_info    : case_printing (** not interpreted by the kernel *)
   }