'Undeclared variable' error on use of declared constructor

[toinehartman]

Describe the bug
Using MStatus::ignored() throws an Undeclared variable error, while using MStatus::tpl_uptodate() does not.

To Reproduce

rascal>import lang::rascalcore::check::CheckerCommon;
ok
rascal>ignored();
|prompt:///|(0,7,<1,0>,<1,7>): Undeclared variable: ignored
Advice: |https://www.rascal-mpl.org/docs/Rascal/Errors/CompileTimeErrors/UndeclaredVariable|
rascal>MStatus::ignored();
|prompt:///|(0,16,<1,0>,<1,16>): Undeclared variable: MStatus::ignored
Advice: |https://www.rascal-mpl.org/docs/Rascal/Errors/CompileTimeErrors/UndeclaredVariable|
rascal>tpl_saved();
MStatus: tpl_saved()

Expected behavior
No error.

Desktop (please complete the following information):

• Context: Eclipse plugin, Commandline REPL
• Rascal Version: 0.41.0-RC76

Additional context:
Relevant declaration in lang::rascalcore::check::CheckerCommon:

data MStatus =
      rsc_not_found()
    | tpl_not_found()
    | rsc_changed()
    | parsed()
    | parse_error()
    | module_dependencies_extracted()
    | checked()
    | check_error()
    | code_generated()
    | code_generation_error()
    | tpl_uptodate()
    | tpl_saved()
    | ignored()
    | bom_update_needed()
    ;

[jurgenvinju]

Hmmmm it's ironic that "ignored" is ignored. Just to be sure we should first check that the actual name "ignored" is not a factor.

[jurgenvinju]

@DavyLandman this topic intersects with the recent fixes and optimizations; so better have a look before the release.

[DavyLandman]

I went back to 0.41.0-RC73 and it has the same behavior.

Looks like MStatus is not missing that alternative.

rascal>#MStatus
type[MStatus]: type(
  adt(
    "MStatus",
    []),
  (adt(
      "MStatus",
      []):choice(
      adt(
        "MStatus",
        []),
      {
  ...
        cons(
          label(
            "ignored",
            adt(
              "MStatus",
              [])),
          [],
          [],
          {}),
 ...
      })))

[DavyLandman]

Aha:

rascal>lang::rascalcore::check::CheckerCommon::ignored()
|prompt:///|(0,47,<1,0>,<1,47>): A value of type set[str] is not something you can call like a function, a constructor or a closure.
Advice: |https://www.rascal-mpl.org/docs/Rascal/Errors/CompileTimeErrors/UnsupportedOperation|

I've looked at the places where there is a global define of ignored, and the only place the set[str] ignored in analys::typepal::Collector. It's a global private variable. But because we extend, this gets our own private variable.

So could it be this is an error where after evaluating the extend we loose the ability to reach the ignored in the adt (but that private field ignored is not visible to the REPL?)

[jurgenvinju]

Definitely something to look into, especially since it's only "ignored" that had the problem.

1. If the private variable came into scope during extend, then "can not call as a function" should be the error, not the UndeclaredVariable. Although it is more common to promote local definitions over extended ones. If that were implemented it should just work.
2. With the MStatus prefix, the code should just work. Regardless of other definitions in scope.

Looks like we have two to three causes of trouble at the same time.

[DavyLandman]

I mean than from the repl's view, ignored is a private variable of the CheckerCommon module, why we're even getting that message is confusing, as if we can access private globals?

hmm, yes, we can

rascal>lang::rascalcore::check::CheckerCommon::traceTPL
bool: false
rascal>traceTPL
|prompt:///|(0,8,<1,0>,<1,8>): Undeclared variable: traceTPL
Advice: |https://www.rascal-mpl.org/docs/Rascal/Errors/CompileTimeErrors/UndeclaredVariable|

that's not good. (I'm still on RC73)

[jurgenvinju]

I think that accessing private globals is only possible in the REPL for inspection purposes, but this may be old info.

[jurgenvinju]

Now we have four misbehaviors. We should check them one by one.