Compiler goes into an infinite loop with invalid @type

[filterfish]

Elixir and Erlang/OTP versions

Erlang/OTP 27 [erts-15.1.2] [source] [64-bit] [smp:12:12] [ds:12:12:10] [async-threads:1] [jit:ns]

Elixir 1.17.3 (compiled with Erlang/OTP 27)

Operating system

Debian Linux (sid)

Current behaviour

I've been playing around with types and I made error in the @type. I added t().String instead of String.t() (from this example) and the compiler went into an infinite loop. The full code fragment is:

defmodule Foo do                    
  @typedoc "A product"    
  @type product() :: t().String   
end

Other observations:

• t.String also fails
• T.String compiles
• t.String followed by T.String fails with the following message:

== Compilation error in file lib/foo.ex ==
** (Kernel.TypespecError) lib/foo.ex:10: type product/0 is already defined in lib/foo.ex:9
    (elixir 1.17.3) lib/kernel/typespec.ex:969: Kernel.Typespec.compile_error/2
    (stdlib 6.1.2) lists.erl:2151: :lists.foldl_1/3
    (elixir 1.17.3) lib/kernel/typespec.ex:226: Kernel.Typespec.translate_typespecs_for_module/2

Expected behaviour

The compiler should report an error.

[sabiwara]

The infinite loop seems to be:

Kernel.Typespec.typespec({:__aliases__, [line: 4], [{:t, [line: 4], []}, :String]}, [], env, state)

because

Kernel.Typespec.expand_remote({:__aliases__, [line: 4], [{:t, [line: 4], []}, :String]}, ...)
# returns itself
{:__aliases__, [line: 4], [{:t, [line: 4], []}, :String]}

[josevalim]

Good find. We probably need to check it returns an atom and raise otherwise.

[sabiwara]

Good find. We probably need to check it returns an atom and raise otherwise.

Actually we were already checking it... just instead of raising otherwise we kept looping 😄
#14155