From: Caleb Tennis <caleb@...>
Date: 2005-12-07T20:52:20+09:00
Subject: Re: 1.8.4 and respond_to

>
> How broken?
>

Object initialization was done like this:

     if (rb_respond_to(result, rb_intern("initialize")) != 0) {
	rb_obj_call_init(result, argc, argv);
     }

So, suddenly, objects in Ruby 1.8.4 weren't being initialized.


>> So, I guess what I'm asking is if there's a good reason for  
>> altering this
>> behavior between minor releases?  It seems much more suited for  
>> breaking
>> between a larger release, in my humble opinion.
>
> The reason was:
> 1) it returned true even if a global method was matched,
> 2) it didn't conform to the default behavior of Object#respond_to?.

Right, I agree that it deserved to be fixed.  I think, though, that  
there is some general expectation of people who are familiar with  
open source software that when a program goes from version X.Y.Z to  
X.Y.(Z+1) that it's "binary compatible" with other releases.  This  
seems like too big of change of behavior.

My opinion is that it would be better to change rb_respond_to back to  
the way it was previously, and add a warning about it being  
deprecated in favor of rb_obj_respond_to.

Thanks,
Caleb