From: Tanaka Akira <akr@...>
Date: 2009-08-07T02:28:24+09:00
Subject: [ruby-core:24804] Re: [Bug #1852] Enumerable's #hash Raises 	ArgumentError When Recursive Values are Present

In article <67e307490908022024p6f7b753ej85cd46c61413523e@mail.gmail.com>,
  Run Paint Run Run <runrun@runpaint.org> writes:

> What's more, the bug report that seemingly triggered this change was
> not derived from a "real application". So why the uneven burden of
> proof?

As [ruby-core:24771] and [ruby-core:24773], Ruby 1.9.2 may
break such application anyway even if #hash is not changed
or #hash returns a constant because eql? is changed.

There are incompatibility here anyway.

> Again, all I'm asking for is an explanation of this change. It's hard
> enough to spec trunk Ruby without seemingly arbitrary changes being
> made "just because". :-/

1.9.1's #hash is broken according to 1.9.2's #eql?.

The exception "fix" it.  Since cyclic data structure in hash
key is rare, I think exception is not big problem.

I guess the fix mentioned in [ruby-core:24761] may have some
performance impact for usual non-cyclic hash key since all
top-level #hash method call needs catch().  If the
performance impact is small enough, no problem for the fix.
-- 
Tanaka Akira