From: Michael Klishin <michael.s.klishin@...>
Date: 2011-09-09T16:41:18+09:00
Subject: [ruby-core:39419] Re: RubySpec vs CRuby's test/...

Urabe Shyouhei:

> and you never know WHY IT SHOULD.  I don't believe this is a proper
> "standard" behaviour that all CRuby committers can agree.  Giving
> negative numbers for %u a programmer should have intended to cast it
> to an unsigned integer.  This spec does describe the status quo of
> our implementation but not for the "ideal" figure.

test/* tests do not explain that either. And no documentation does. It is not surprising for a project that describes some existing implementation to only claim how things work, not why.

Nobody says that RubySpec is "ideal". Specs that explain *why* things work in certain ways take forever to write and only can be written in retrospect after years of trial and error, take a look at HTML5 and related specifications for example.

I think the point of this thread originally was to encourage MRI developers to *use* RubySpec more (it needs some work on Windows, but other than that, it is a giant body of work that is already done) and contribute to it when behavior of the language changes, not claim that RubySpec is ideal or an absolute of some kind.


MK