Re-considering package generic boundary

[dalance]

#1223 and #1294 introduced generic package basename as generic boundary, but there seems to be a problem.
For example, the function signature of func is not determined in the following code.
So deep nested foo_pkg causes long and complex type error like C++'s template.

package foo_pkg::<T: type> {
  function func (x: input T) -> T {
  }
}

This is the reason why proto module was introduced.
So I think making clear what members are visible through generic variables using proto package
is better even though some boilerplate code is necessary.

proto package proto_foo {
  const ADDRESS_WIDTH: u32;
  const DATA_WIDTH:    u32;
  type address_t;
  type data_t;
}

package foo_pkg::<AW: const, DW: const> for proto_foo {
  const ADDRESS_WIDTH: u32 = AW;
  const DATA_WIDTH:    u32 = DW;
  type address_t = logic<ADDRESS_WIDTH>;
  type data_t    = logic<DATA_WIDTH>;
}

module foo_top::<PKG: proto_foo> (
  i_address: input PKG::address_t,
  i_data:    input PKG::data_t,
) {
  import PKG::*;

  var address: address_t;
}

By this approach, controlling member visibility is enabled.

proto package proto_foo {
  type address_t;
  type data_t;
}

package foo_pkg::<AW: const, DW: const> for proto_foo {
  // The following const are not visible through generic variables
  const ADDRESS_WIDTH: u32 = AW;
  const DATA_WIDTH:    u32 = DW;

  type address_t = logic<ADDRESS_WIDTH>;
  type data_t    = logic<DATA_WIDTH>;
}

@taichi-ishitani How do you think about it?

[taichi-ishitani]

Do we need to speicfy types of parameters and/or result of a proto function?

[dalance]

Do we need to speicfy types of parameters and/or result of a proto function?

I think it is more safety because it protects unexpected type errors caused by wrong usage.

[taichi-ishitani]

OK, I understand.
I will work on this after closing #1312.

[dalance]

I'm preparing to release v0.14.0. How do you think whether this issue and #1312 should be included in v0.14.0?
I can wait to complete all of them, or release v0.14.0 by disabling #1223 temporarily.

[taichi-ishitani]

I think it will take 1 week to fix package boundary issues.

[dalance]

OK.I'll release v0.14.0 without #1223 because it already contains enough updates.

[taichi-ishitani]

definitions of enum, struct and union in a proto package are same with actual definitions.
It would be nice to copy these definitions from proto package from actual package.

[dalance]

It would be nice to copy these definitions from proto package from actual package.

I think so too. But completly automatic copy may be confusable about where these types are defined.
I prefer small explicit syntax. It may be related #1212.