Why do I get “stack level too deep” from method_missing in irb 1.9.3?

Question

Scenario:

-bash-3.2$ irb -f
ruby-1.9.3-p0 :001 > @v = {}
 => {} 
ruby-1.9.3-p0 :002 > def method_missing(sym, *args); @v[sym]; end
 => nil 
ruby-1.9.3-p0 :003 > a
(irb):2: stack level too deep (SystemStackError)
-bash-3.2$ 

I ran with -f to avoid loading any irbrc stuff. I'm expecting to get nil when I input a. What's going on, and is there a workaround? I tried wrapping a with a begin/rescue Exception block but that didn't do anything.

This also happens with 1.9.2, but not 1.9.1.

More strange behavior:

-bash-3.2$ irb -f
irb(main):001:0> @v = {}
=> {}
irb(main):002:0> def method_missing(sym, *args); @v[sym]; end; 5.times { p a }
nil
nil
nil
nil
nil
=> 5
irb(main):003:0> a
(irb):2: stack level too deep (SystemStackError)
-bash-3.2$ 

This tells me that there's a bug in irb, or that some obscure bug in ruby is being triggered by irb. Also, after defining method_missing, even methods that exist like local_variables or eval cause the error.

Answer 1

Looks like defining it as a singleton method works:

def self.method_missing(sym, *args); @v[sym]; end

Defining it as a top-level method replaces BasicObject#method_missing, which probably affected some irb internals like Phrogz said.