frontend: Issue with type checker in alias register

[BadGraphixD]

Creating a register alias for accessing a single bit of a register causes the typechecker to break. The resulting type of the alias register is incorrect. It should be Bits<1> instead of RegFormat:

instruction set architecture ISSUE = {

  register REG : RegFormat

  // both do not work
  alias register REGa = REG.a
  // alias register REGa : Bits<1> = REG.a

  format RegFormat : Bits<8> =
  { a : Bits<1> , b : Bits<7> }

  format InstrFormat : Bits<8> =
  { dummy : Bits<8> }

  constant c : Bits<1> = 0

  instruction instr : InstrFormat = if REGa = c then {}
  encoding instr = { dummy = 0 }
  assembly instr = "instr"
}

Running openvadl check issue.vadl yields:

error: Type Mismatch
     ╭──[code/open-vadl/sys/issue.vadl:17:40]
     │
  17 │   instruction instr : InstrFormat = if REGa = c then {}
     │                                        ^^^^^^^^ note: The left type is RegFormat while right is Bits<1>
     │
     Both types on the left and right side of an binary operation should be equal.

Using the second (commented out) alias definition yields:

error: Type Mismatch
     ╭──[code/open-vadl/sys/issue.vadl:7:35]
     │
   7 │   alias register REGa : Bits<1> = REG.a
     │                                   ^^^^^ Expected `Bits<1>` but got `RegFormat`.
     │

[flofriday]

Unfortunately the alias implementation is far from clean/stable and there seem to be quite a few bugs.

One current workaround would be:

instruction set architecture ISSUE = {

  register REG : Bits<8>

  // both do not work
  alias register REGa: Bits<1> = REG(0 as Bits<3>)

  format RegFormat : Bits<8> =
  { a : Bits<1> , b : Bits<7> }

  format InstrFormat : Bits<8> =
  { dummy : Bits<8> }

  constant c : Bits<1> = 0

  instruction instr : InstrFormat = if REGa as Bits<1> = c then {}
  encoding instr = { dummy = 0 }
  assembly instr = "instr"
}

[BadGraphixD]

I have implemented #832 using the second way mentioned in #777, so:

[ <anno name> : sf, priv ]
register MSR: MSRFormat

At this point, the first way, which requires this issue to be solved, is good to have, but not necessary. The important thing would be, that I can read the BitSlice the alias accesses (so in the example above, that I can read the slice [1] somewhere).

Unless it takes little time so solve this issue, I would close it as "not planned".