Bug #6239: super Does Not Pass Modified Rest Args When Originally Empty

• Status: Rejected
• Assignee: -
• Target version: -
• ruby -v: 1.8.7-p358

[mudge (Paul Mucur)]

Given the following code:

class A
  def a(*args)
    args
  end
end

class B < A
  def a(*args)
    args << "foo"

    super
  end
end

1.8.7-p358> B.new.a("bar")
=> ["bar", "foo"]
1.8.7-p358> B.new.a
=> []

But:

1.9.2-p290> B.new.a("bar")
=> ["bar", "foo"]
1.9.2-p290> B.new.a
=> ["foo"]

The 1.9.2 behaviour is more predictable than 1.8.7, should this be backported or is this expected behaviour?

[mudge (Paul Mucur)]

As a side note, both JRuby 1.6.6 and Rubinius 1.2.4 in 1.8 compatibility mode follow the behaviour of 1.9.2.

James Hunt and I have filed a pull request to the RubySpec project to document this issue at: https://github.com/rubyspec/rubyspec/pull/130

[matz (Yukihiro Matsumoto)]

• Status changed from Open to Rejected

It's behavior change in 1.9, and we are not going to change the behavior in 1.8 (other than fixing bugs) any more.

Matz.

[marcandre (Marc-Andre Lafortune)]

Hi Matz

matz (Yukihiro Matsumoto) wrote:

It's behavior change in 1.9, and we are not going to change the behavior in 1.8 (other than fixing bugs) any more.

Oh, it is?

So you mean that it was a known fact that there was an exceptional case with super and modified rest argument only when that rest argument was empty? And that it was a "feature"? Was there any rationale for that behavior?

I'm not saying the bug must be fixed, as it can be easily circumvented using explicit arguments, but it definitely looks like a bug to me.

[matz (Yukihiro Matsumoto)]

Hi,

It has not been considered as a bug. The old document described this
behavior. Ko1 asked me to allow behavioral change in 1.9, and I admit.

						matz.

In message "Re: [ruby-core:44023] [Ruby 1.8 - Bug #6239] super Does Not Pass Modified Rest Args When Originally Empty"
on Sun, 1 Apr 2012 03:20:18 +0900, "marcandre (Marc-Andre Lafortune)" [e-mail] writes:

|matz (Yukihiro Matsumoto) wrote:
|> It's behavior change in 1.9, and we are not going to change the behavior in 1.8 (other than fixing bugs) any more.
|
|Oh, it is?
|
|So you mean that it was a known fact that there was an exceptional case with super and modified rest argument only when that rest argument was empty? And that it was a "feature"? Was there any rationale for that behavior?
|
|I'm not saying the bug must be fixed, as it can be easily circumvented using explicit arguments, but it definitely looks like a bug to me.

[mudge (Paul Mucur)]

Thank you, Matz.

We have now covered this behaviour in RubySpec as of https://github.com/rubyspec/rubyspec/commit/ffb634f3899dfc0d8a494afa8c754992041ddb2e