Avoid false positives when diffing bitstrings, closes #15204

Author: josevalim

lib/ex_unit/lib/ex_unit/diff.ex

@@ -157,9 +157,7 @@ defmodule ExUnit.Diff do
 
   ## diff_value
 
-  defp diff_value(literal, literal, env)
-       when is_atom(literal) or is_number(literal) or is_reference(literal) or
-              is_pid(literal) or is_function(literal) do
+  defp diff_value(literal, literal, env) do
     {%__MODULE__{equivalent?: true, left: literal, right: literal}, env}
   end
 
@@ -195,6 +193,9 @@ defmodule ExUnit.Diff do
   end
 
   defp diff_value(left, right, env) when is_binary(left) and is_binary(right) do
+    # We only differ over binaries because it is hard to diff over bitstrings
+    # cause we don't know where the misaligned bit is. We could compare bit by bit
+    # but it would be too verbose
     diff_string(left, right, ?", env)
   end
 

lib/ex_unit/test/ex_unit/diff_test.exs

@@ -1034,7 +1034,13 @@ defmodule ExUnit.DiffTest do
     )
   end
 
-  test "strings" do
+  test "bitstrings" do
+    assert_diff(<<1::1>> == <<1::1>>, [])
+    refute_diff(<<1::1>> == <<1::2>>, "-<<1::size(1)>>", "+<<1::size(2)>>+")
+    refute_diff(<<1::1>> == <<0::1>>, "-<<1::size(1)>>", "+<<0::size(1)>>+")
+  end
+
+  test "binaries" do
     assert_diff("" = "", [])
     assert_diff("fox hops over the dog" = "fox hops over the dog", [])