# What does this warning actually mean?

**URL:** <https://www.ruby-forum.com/t/what-does-this-warning-actually-mean/111089>\
**Category:** Ruby\
**Created:** [August 2, 2007, 2:48pm UTC](https://www.ruby-forum.com/t/what-does-this-warning-actually-mean/111089 "2007-08-02T14:48:00Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![Michael\_TSRichter](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/m/e5b9ba/32.png) [@Michael\_TSRichter](https://www.ruby-forum.com/u/Michael_TSRichter)\
**Post date:** [August 2, 2007, 2:48pm UTC](https://www.ruby-forum.com/t/what-does-this-warning-actually-mean/111089/1 "2007-08-02T14:48:00Z")

</div>

So, I’m trying my hand at metaprogramming and am finding this the result  
of executing my code with -w enabled:

warning: class variable access from toplevel singleton method

(This is spammed ad infinitum anywhere I use the overridden gets, but  
it’s always the same message and always the same cause: gets.)

The code it’s pointing to looks like this:

def override\_methods(io\_in)  
puts io\_in  
class \<\< io\_in  
alias :old\_gets :gets  
@@override\_buffer=[]  
def gets separator = $/  
raise ArgumentError, “cannot use non-standard  
separator” unless separator == $/

```
                    unless line = @@override_buffer.shift
                            return old_gets
                    else
                            return line
                    end
            end
            def find_end
                    for a in 1..20
                            line = self.old_gets
                            if line == " __END__ \n"
                                    @@override_buffer = []
                                    break
                            else
                                    @@override_buffer << line
                            end
                    end
                    self
            end
    end
    io_in

```

end

def find\_block\_begin io\_in  
io\_in = override\_methods(io\_in)  
io\_in.find\_end  
end

To me the culprit is likely that @@override\_buffer thing, but I can’t  
for the life of me figure out how to get around it. My modified gets  
needs access to that (shared) buffer that the added find\_end method  
generates while peeking ahead. @override\_buffer and override\_buffer  
don’t work (the latter for reasons obvious to me, the former … not so  
obvious).

Now, the output of the code is exactly what I expect it to be, so this  
isn’t a show-stopper. I am, however, in the habit of making my code -w  
clean if it’s at all possible. How would I do it with that snippet?

---

<div class="post-metadata">

**Author:** ![Michael\_TSRichter](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/m/e5b9ba/32.png) [@Michael\_TSRichter](https://www.ruby-forum.com/u/Michael_TSRichter)\
**Post date:** [August 2, 2007, 3:06pm UTC](https://www.ruby-forum.com/t/what-does-this-warning-actually-mean/111089/2 "2007-08-02T15:06:02Z")

</div>

On 8/2/07, Michael T. Richter [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) wrote:

Michael I got no clue what you are trying to do here, but maybe the  
following is hint enouh, BTW, just do not use class variables ;).  
522/23 \> cat test1.rb && ruby test1.rb

class \<\< self  
@@a=42  
def a  
@@a +=1  
end  
end  
puts a  
test1.rb:3: warning: class variable access from toplevel singleton  
method  
test1.rb:5: warning: class variable access from toplevel singleton  
method  
test1.rb:5: warning: class variable access from toplevel singleton  
method  
43

* * *

523/24 \> cat test2.rb && ruby test2.rb

class \<\< self  
def a  
@a||=42  
@a+=1  
end  
end  
puts a

43

HTH  
Robert

---

<div class="post-metadata">

**Author:** ![Michael\_TSRichter](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/m/e5b9ba/32.png) [@Michael\_TSRichter](https://www.ruby-forum.com/u/Michael_TSRichter)\
**Post date:** [September 25, 2007, 11:11pm UTC](https://www.ruby-forum.com/t/what-does-this-warning-actually-mean/111089/3 "2007-09-25T23:11:10Z")

</div>

On 8/2/07, Michael T. Richter [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) wrote:

> 

Michael please post in raw text format, this reads terribly and can  
probably  
not go to the newsgroup.

This much I figured out for myself, Robert. [image: ;)]

> 

If you did you should know exactly how to solve the problem, use class  
instance variables instead of class variables. A sound advice in all  
cases  
🙂

The problem is that I need to share a variable across _two_ functions –  
a

> buffer, to be precise. One function fills the buffer while the other  
> consumes from it. The functions are to be singleton methods on an  
> instantiated IO object. I can’t see how to make them share an instance  
> variable without doing funky things like calling instance\_variable\_set on  
> the object to provide it with said instance variable before adding the  
> singletons. Using a class variable was a desperate test after running  
> headlong into a brick wall trying to share them.

You see we are not talking instance variables here, actually we are, the  
important thing is instance variable of what, of the class object, or in  
your case singleton class object, that will perfectly do the trick.

As it is now, I’ve got it working with an additional call to

> instance\_variable\_set before defining the singletons, but it looks kind of  
> clunky and I was hoping there was a more elegant approach available.

sure does, sure is.

Robert

–