Issue when chaining parametricity translation with definition about metacoq

[CohenCyril]

@drouhling I just realized we should continue the minimization a bit further... replacing the first call of tTranslate by a simpler function... (maybe a single tmMkDefinition or tmMkInductive is enough?)

[CohenCyril]

Also, maybe the backtracking comes from typeclass resolution, we should try forcing the Translation to be a fixed one: I just noticed Translation is a class without parameters, so it is extremely dangerous to keep it implicit.

[SimonBoulier]

I encountered the same problem!
I will try to fix it as soon as I can.
Say me if you make some progress.

[drouhling]

I focused on these lines in tTranslate (lines 91 to 93 in translation_utils.v):

        monad_fold_left (fun _ e => tmMkInductive' e) decls tt ;;
        print_nf  (id ++ " has been translated as " ++ id') ;;
        ret (add_global_decl (InductiveDecl kn d) (fst ΣE), E ++ snd ΣE)%list

since they are the last lines to be executed when translating for instance unit.

In the case of the translation of unit chained with a definition:

Notation pb :=
     (tInd {| inductive_mind := "Top.unitᵗ"; inductive_ind := 0 |} []).
Fail Run TemplateProgram (tTranslate' emptyTC "unit" ;; tmMkDefinition "toto" pb).

What is executed at the end of the call to tTranslate corresponds to:

Notation bt :=
 ({|
 ind_npars := 0;
 ind_bodies := [{|
                ind_name := "unitᵗ";
                ind_type := tProd nAnon
                              (tInd
                                 {|
                                 inductive_mind := "Coq.Init.Datatypes.unit";
                                 inductive_ind := 0 |} []) (tSort [(Level.lSet, false)]);
                ind_kelim := [InProp; InSet; InType];
                ind_ctors := [("ttᵗ",
                              tApp (tRel 0)
                                [tConstruct
                                   {|
                                   inductive_mind := "Coq.Init.Datatypes.unit";
                                   inductive_ind := 0 |} 0 []], 0)];
                ind_projs := [] |}];
 ind_universes := Monomorphic_ctx
                    ([],
                    {| ConstraintSet.this := [];
                       ConstraintSet.is_ok := ConstraintSet.Raw.empty_ok |}) |}).

Notation d :=
  ({|
ind_npars := 0;
ind_bodies := [{|
               ind_name := "unit";
               ind_type := tSort [(Level.lSet, false)];
               ind_kelim := [InProp; InSet; InType];
               ind_ctors := [("tt", tRel 0, 0)];
               ind_projs := [] |}];
ind_universes := Monomorphic_ctx (UContext.make [] ConstraintSet.empty) |}).

Notation E := [(IndRef {| inductive_mind := "Coq.Init.Datatypes.unit"; inductive_ind := 0 |},
pb); (ConstructRef {| inductive_mind := "Coq.Init.Datatypes.unit"; inductive_ind := 0 |} 0,
tConstruct {| inductive_mind := "Top.unitᵗ"; inductive_ind := 0 |} 0 [])].

Run TemplateProgram (tmMkInductive' bt ;; ret (add_global_decl
(InductiveDecl "Coq.Init.Datatypes.unit" d) (fst emptyTC), E)%list).

If we run after this the command

Run TemplateProgram (tmMkDefinition "toto" pb).

everything is ok, but if we chain both commands:

Run TemplateProgram (tmMkInductive' bt ;; ret (add_global_decl
(InductiveDecl "Coq.Init.Datatypes.unit" d) (fst emptyTC), E)%list ;;
tmMkDefinition "toto" pb).

we get Error: Anomaly "Uncaught exception Not_found." Please report at http://coq.inria.fr/bugs/.

The ret part does not seem to matter much since

Run TemplateProgram (tmMkInductive' bt ;; tmMkDefinition "toto" pb).

also raises the exception.

[SimonBoulier]

I don't remember exactly what was the problem but I fixed it in #89.
(Moreover the translation plugin becomes up to date).
Say me if you encounter any other problem.

[CohenCyril]

@SimonBoulier since

Run TemplateProgram (tmMkInductive' bt ;; ret (add_global_decl
(InductiveDecl "Coq.Init.Datatypes.unit" d) (fst emptyTC), E)%list ;;
tmMkDefinition "toto" pb).

raises an anomaly, it is not about the translation plugin.

[SimonBoulier]

Yes but I also corrected other stuffs in this branch.

[CohenCyril]

Hi @mattam82, any idea about @drouhling's bug?
@SimonBoulier, in #89, why are you merging in a branch that is not master (or any coq release)?

[mattam82]

I think it was fixed by properly threading the updated global environment in the monadic interpretation function (e.g. Not_found was raised because tmMkDefinition did not see the declaration of the inductive before).

[mattam82]

#89 was merged into 8.8 through another PR

[mattam82]

Does this fix the original issue?

[drouhling]

Yes. I'm closing the issue.