diff --git a/core/src/main/java/org/jruby/RubyKernel.java b/core/src/main/java/org/jruby/RubyKernel.java
index 5146f40a73e6..f69b5a46fd3d 100644
--- a/core/src/main/java/org/jruby/RubyKernel.java
+++ b/core/src/main/java/org/jruby/RubyKernel.java
@@ -2028,16 +2028,6 @@ public static IRubyObject fork19(ThreadContext context, IRubyObject recv, Block
         return fork(context, recv, block);
     }
 
-    @JRubyMethod(module = true)
-    public static IRubyObject tap(ThreadContext context, IRubyObject recv, Block block) {
-        if (block.getProcObject() != null) {
-            block.getProcObject().call(context, recv);
-        } else {
-            block.yield(context, recv);
-        }
-        return recv;
-    }
-
     @JRubyMethod(name = {"to_enum", "enum_for"}, rest = true, keywords = true)
     public static IRubyObject obj_to_enum(final ThreadContext context, IRubyObject self, IRubyObject[] args, final Block block) {
         // to_enum is a bit strange in that it will propagate the arguments it passes to each element it calls.  We are determining
@@ -2644,4 +2634,15 @@ public static IRubyObject method(IRubyObject self, IRubyObject symbol) {
     public static IRubyObject method19(IRubyObject self, IRubyObject symbol) {
         return method(self, symbol);
     }
+
+    // defined in Ruby now but left here for backward compat
+    @Deprecated
+    public static IRubyObject tap(ThreadContext context, IRubyObject recv, Block block) {
+        if (block.getProcObject() != null) {
+            block.getProcObject().call(context, recv);
+        } else {
+            block.yield(context, recv);
+        }
+        return recv;
+    }
 }
diff --git a/core/src/main/ruby/jruby/kernel/kernel.rb b/core/src/main/ruby/jruby/kernel/kernel.rb
index b48a586b21e1..13c4ab473df7 100644
--- a/core/src/main/ruby/jruby/kernel/kernel.rb
+++ b/core/src/main/ruby/jruby/kernel/kernel.rb
@@ -1,16 +1,22 @@
 module Kernel
-  module_function
-  def exec(*args)
+  module_function def exec(*args)
     _exec_internal(*JRuby::ProcessUtil.exec_args(args))
   end
 
   # Replaces Java version for better caching
-  def initialize_dup(original)
+  module_function def initialize_dup(original)
     initialize_copy(original)
   end
 
   # Replaces Java version for better caching
-  def initialize_clone(original, freeze: false)
+  module_function def initialize_clone(original, freeze: false)
     initialize_copy(original)
   end
+
+  def tap(&b)
+    # workaround for yield to lambda not triggering arity error
+    # see https://bugs.ruby-lang.org/issues/12705
+    b.lambda? ? b.call(self) : yield(self)
+    self
+  end
 end