(+\)/2 in call/1 confused predicate calls

[bakaq]

?- [user].
:- use_module(library(lambda)).

p.

?- Goal = (X +\ p), call(Goal).
   error(existence_error(procedure,p/0),p/0).
?- Goal = (X +\ (user:p)), call(Goal).
   Goal = X+\(user:p).

This seems to be the ultimate cause for aarroyoc/scryer-playground#20. I couldn't reduce this for a custom predicate instead of (+\)/2, so I guess the problem is there, maybe in no_hat_call/1?

[bakaq]

Also, this works for some reason:

?- call(X +\ p).
   true.

[infradig]

$ scryer-prolog
?- [user]. :- use_module(library(lambda)). p.

?- Goal = (X +\ user:p), write_canonical(Goal), nl.
:(+(_74543,user),p)
Goal = X+\user:p.
?-

[bakaq]

That doesn't seem to be the problem:

?- Goal = (X +\ (user:p)), write_canonical(Goal), nl, call(Goal).
+\(_43,:(user,p))
   Goal = X+\(user:p).

[UWN]

Works in SICStus. If at it, the existence error produced by Scryer does not reveal the actual module where the error occurs, which makes diagnosing difficult.

[hurufu]

I think this is still not fixed fully:

?- use_module(library(lambda)).
   true.
?- [user].
p.
same(A,A).

?- X = (_+\p), same(X, Y), call(Y).
   error(existence_error(procedure,p/0),p/0). % unexpected
?- X = (_+\p), call(X).
    X = _A+\p.   % expected
?- 

UPDT: I've found this behavior during learning semantics of goal expansion for #2532

[UWN]

Same problem when all is dynamic:

?- asserta((t2:-X=(_+\p),call(X))).
   true.
?- t2.
   error(existence_error(procedure,p/0),p/0).
?- p.
   true.

[UWN]

Even

?- asserta((t3:-call(_+\p))).
   true.
?- t3.
   error(existence_error(procedure,p/0),p/0).

[bakaq]

What in tarnation!!????

?- [user].
:- use_module(library(lambda)).

p.

a(\p).

?- G = (\p), call(G).
   G = \p.
?- a(G), call(G).
   error(existence_error(procedure,p/0),p/0).
?- a(G0), G = (\p), G0 = G, call(G).
   G0 = \p, G = \p.
?- a(G0), G = (\p), G0 = G, call(G0).
   % G0 and G unified, why does one work and the
   % other give an error!!?? 
   error(existence_error(procedure,p/0),p/0).