[BUG] Parameter named after contextual keyword is lost

[JohelEGP]

Title: Parameter named after contextual keyword is lost.

Such a Cpp2 parameter is lowered to Cpp1 as no parameter at all.

Minimal reproducer (https://cpp2.godbolt.org/z/1exj84Wb4):

t: type = {
  f: (in) = { }
}
g: (final) = { }
// h: (out, x) = { } // Error!
main: () = { }

Commands:

cppfront -clean-cpp1 main.cpp2
clang++17 -std=c++2b -stdlib=libc++ -lc++abi -I . main.cpp

Expected result:
A diagnostic,
or the equivalent using y as an identifier.

Actual result and error:

#include "cpp2util.h"

class t;
  
class t {
  public: static auto f() -> void;

  public: t() = default;
  public: t(t const&) = delete;
  public: auto operator=(t const&) -> void = delete;
};
auto g() -> void;
// h: (out, x) = { } // Error!
auto main() -> int;

  auto t::f() -> void{}

auto g() -> void{}

auto main() -> int{}

[JohelEGP]

See also #387 (comment).

I believe (in: T) should be a function type with an by-in parameter of type T.

Visualized

See https://highlight-cpp2.godbolt.org/z/K9nvoqKdh.

That may be relevant in deciding the semantic meaning of (𝘵𝘩𝘪𝘴-𝘴𝘱𝘦𝘤𝘪𝘧𝘪𝘦𝘳) as a parameter-declaration-list.

[hsutter]

Thanks, this is a good example.

The linked PR #505 would allow this code:

t: type = {
  f: (in) -> _ = in;
}
g: (final) -> _ = final.first;
h: (out, x) -> _ = out(x);

I'm inclined to say that these should be interpreted as a missing parameter name, and should get a better diagnostic and be rejected at compile time.

I don't think I want to allow these. Are they ever necessary? They seem confusing (at least visually) and allowing them creates new things to teach and learn that would not be necessary if we disallowed parameters named the same as a parameter qualifier (parameter-direction or this-specifier).

[JohelEGP]

That makes sense.
I wasn't sure before, but now I'm convinced.
I'll rebase #505 to diagnose as you mentioned.

[hsutter]

It's okay, I just made the change and am about to check it in. Thanks again!