[SR-8811] Diagnosing Implicit Accessors for Uninhabited Types Should Be Smarter

[CodaFi]

Previous ID	SR-8811
Radar	None
Original Reporter	@CodaFi
Type	Improvement
Status	Reopened
Resolution

Additional Detail from JIRA

Votes	0
Component/s	Compiler
Labels	Improvement, SILDiagnostics, StarterBug
Assignee	None
Priority	Medium

md5: fe8546f9296464e8c547c7c953e07a35

Issue Description:

Try to emit SIL for the following code

enum Empty {}

class Foo {
  var x: Empty // function with uninhabited return type 'Empty' is missing call to another never-returning function on all paths

  init(_ x: Empty) {
    self.x = x
  }
}

It's good that we error, it's dismaying that we're diagnosing an implicit declaration. In particular, we're diagnosing the getter. We should check for accessors and emit something specialized here.

[belkadan]

Alternately, we could change SILGen to emit a trap as the body of the function.

…actually, we should probably just change Sema to disallow a stored property with uninhabited type, since the rest of the language won't stop you from trying to call the getter. You can already see how silly it is in the initializer, which tries to set it.

[Azoy]

This is resolved here: #19992

[jckarter]

Making it an error to define a variable with uninhabited types is not the correct fix. The diagnostic in the getter is a bug; since it should already be unreachable to have a struct or class instance with an uninhabited stored property, the getter is already impossible to call, and we should not emit any reachability diagnostics for it. Maybe this could be a warning, but it should not be an error to declare a variable or property with uninhabited type.