Backport #1151

Aliased methods change super logic when retrieved with Object#method

Added by headius (Charles Nutter) over 3 years ago. Updated about 1 year ago.

[ruby-core:22040]
Status:Assigned Start date:02/13/2009
Priority:Normal Due date:
Assignee:wyhaines (Kirk Haines) % Done:

100%
Category:-
Target version:-

Description

This is a peculiar case I don't believe I've reported before. It seems that "method" can change the super behavior of an alias:

 # Test weird likely-a-bug where method() will repurpose where super goes to
 class Foo222
   def a; 'a'; end
   def b; 'b'; end
 end

 class Bar222 < Foo222
   def a; super; end
   alias b a
 end

 puts('a' == Bar222.new.b) # true
 puts('a' == Bar222.new.method(:b).call) # false

Ruby 1.9 behaves as you would expect, calling the "a" super method in both cases. We changed our behavior in JRuby 1.1.2 to match Ruby 1.8.6, but I still believe this is a bug. The JRuby bug report is here: http://jira.codehaus.org/browse/JRUBY-1192 and I reported it to ruby-core here: http://blade.nagaokaut.ac.jp/cgi-bin/scat.rb/ruby/ruby-core/11600 and a patch was proposed here: http://blade.nagaokaut.ac.jp/cgi-bin/scat.rb/ruby/ruby-core/11603. Since it behaves as I expect in 1.9, I assume the 1.8.x behavior is incorrect.

Associated revisions

Revision 25693
Added by marcandre over 2 years ago

* eval.c (method_inspect, method_name, mnew): Bug fix when method created from an alias. Based on a patch by Peter Vanbroekhoven [ruby-core:22040]

History

#1
Updated by headius (Charles Nutter) about 3 years ago

Any comments on this one? I've run into another case where I could fix this and improve other JRuby behavior, but I need confirmation that the Ruby 1.8 behavior is incorrect.

#2
Updated by headius (Charles Nutter) about 3 years ago

A bit more background...

I am adding inline caching to super calls. The "buggy" behavior makes this more complicated, since depending on whether the "a" child method or "b" alias was called, the super path will be different, even though the call site is the same and the method containing it is the same.

I am temporarily adding logic to my cache that will check if the name has changed, but I still find this behavior counter-intuitive, and I believe it is a bug.

#3
Updated by al2o3cr (Matt Jones) over 2 years ago

Not sure if this is related or not, but I've run across another very similar issue regarding method(:name).call not being equivalent to calling the method.

Example is here: http://gist.github.com/221585

The net result is more dramatic than in this example; rather than simply calling the wrong method, the call to super fails with 'superclass not found'. But the essential problem is the same - calling B.new.foo produces a different result than B.new.method(:foo).call.

@Charles - can you try the Gist out on JRuby? Thanks!

#4
Updated by headius (Charles Nutter) over 2 years ago

On Fri, Oct 30, 2009 at 9:53 AM, Matt Jones <redmine@ruby-lang.org> wrote:
> Not sure if this is related or not, but I've run across another very similar issue regarding method(:name).call not being equivalent to calling the method.
>
> Example is here: http://gist.github.com/221585
>
> The net result is more dramatic than in this example; rather than simply calling the wrong method, the call to super fails with 'superclass not found'. But the essential problem is the same - calling B.new.foo produces a different result than B.new.method(:foo).call.

In JRuby, it produces the expected output in both cases:

~/projects/jruby ➔ jruby foo.rb
Overridden foo
Overridden foo

(or at least, I believe this is the behavior you would expect to see)

Ruby versions all seem to do it differently, with only 1.8.6 matching
JRuby's output:

~/projects/jruby ➔ ruby foo.rb
Overridden foo
Overridden foo

~/projects/jruby ➔ ruby1.8.7 foo.rb
Overridden foo
foo.rb:5:in `foo': super: no superclass method `foo' (NoMethodError)
	from foo.rb:28:in `call'
	from foo.rb:28

~/projects/jruby ➔ ruby1.9 foo.rb
module foo
module foo

The odd results don't make sense to me.

- Charlie

#5
Updated by marcandre (Marc-Andre Lafortune) over 2 years ago

  • Status changed from Open to Closed
  • % Done changed from 0 to 100
This issue was solved with changeset r25693.
Charles, thank you for reporting this issue.
Your contribution to Ruby is greatly appreciated.
May Ruby be with you.

#6
Updated by shyouhei (Shyouhei Urabe) over 2 years ago

  • Status changed from Closed to Assigned
  • Assignee set to wyhaines (Kirk Haines)

Also available in: Atom PDF