From: "Vít Ondruch" <v.ondruch@...>
Date: 2013-07-09T14:02:46+09:00
Subject: [ruby-core:55873] Re: ruby 1.9.3 p448 breaks ABI

Dne 9.7.2013 2:09, KOSAKI Motohiro napsal(a):
>> Given that not all are native English speakers, I'd like to be 100% sure I understand what is meant by
>> "...it is out of our scope" with regards to ABI breakage by a point release.
>>
>> I'm aware of the following wiki page that briefly mentions ABI
>>
>>    https://bugs.ruby-lang.org/projects/ruby/wiki/ReleaseEngineering
>>
>> but I'm not aware of any ruby-core doco describing what is "in scope" and "out of scope" regarding ABI compliance.
>>
>> Does "...it is out of our scope" mean that ruby-core views point releases as "in scope" when the point releases
>> are "best efforts ABI compatible", but "100% (strict) ABI compatibility" for point releases is "out of scope",
>> or something else?
>>
>> For example, for any 1.9.3pXYZ point release, does ruby-core view it as "out of scope" to be 100% 1.9.1 ABI compatible,
>> but "in scope" to be best efforts 1.9.1 ABI compatible?
> Right. It's case by case thing. Example, when we find a security
> issue, we don't hesitate to fix
> it even if it can't be avoided ABI breakage.

That is fine, but it should be know breakage and it should be acknowledged.

>   Another example, we
> strongly want to keep zero gem breakage, but we don't mind internal
> API even though the checker think it's exported symbol and a part of
> ABI. In the other words, our stable tree focuses to maintain stable
> ruby eco-system and it rarely mismatch strict ABI compatibility.

We have the same focus.

>   If
> anyone dislikes it, feel free to make and maintain your own
> maintenance tree.

We actually doing so and every other distribution is doing it, but we 
try to avoid this situation as much as possible. This is not good for 
distribution users nor for maintainers.

>
> In this case, it's on the border line because rb_f_lambda is
> deprecated.

We have to disagree on this point. I am pretty sure this was 
unintentional change. This should not happen.

>   If I was the maintainer, I didn't make the ABI breakage.
> But please be aware, I and you are not the maintainer's employer. They
> have a right to make and keep their branch maintenance policy. If you
> want to help them, please do. But it's unfair if anyone who help
> nothing blames them. They always did their best.
>

Sorry, I don't meant to blame anybody and I am trying to help as much as 
I can, given my limited resources.


V�t