what should R() do if R has some generic fields with defaults?

[mppf]

Summary of Problem

I am expecting that R() will run the type constructor for a generic record R, and that the type constructor has a default formal for each generic field with a default value. But, this does not seem to be happening for types that are generic but have defaults for some fields.

Steps to Reproduce

Source Code:

record R {
  type t;
  param rank: int = 2;
}

proc printType(type t) {
  writeln(t:string);
  if t.t == ? then
    writeln(".t is uninstantiated");
  else
    writeln(".t: ", t.t:string);
  if t.rank == ? then
    writeln(".rank is uninstantiated");
  else
    writeln(".rank: ", t.rank:string);
}

printType(R);
printType(R());

R
.t is uninstantiated
.rank is uninstantiated
R
.t is uninstantiated
.rank is uninstantiated

I would expect that it output

R
.t is uninstantiated
.rank is uninstantiated
R
.t is uninstantiated
.rank: 2

Associated Future Test(s):

It's not technically a future, but PR #22718 added a warning for this case & a test of that warning

test/types/type_variables/ferguson/some-defaults-ignored.chpl

[bradcray]

@mppf: This issue slipped past me when it was filed. Is it still relevant? What labels should be attached to it? (e.g., bug vs. 2.0?)

[mppf]

It's still relevant and it's related to #18214. I'm not sure what the program shown should do in terms of language design (so I'm not even sure if it's a bug).

However, it is not an issue for Chapel 2.0, because the compiler warns of unstable behavior in this case.