[ruby-talk:8954] Re: String confusion

Robert asks:
> On Wed, 10 Jan 2001, Yukihiro Matsumoto wrote:
> > It's a bug.  Here's the patch.  Thank you.
> > 
> Dave&Andy: Would Rubicon have caught this? How complete is it?

I just tested with latest 1.7.0 and Rubicon 0.1.4. I'm pretty sure this one
would have been missed. Here's the tests for String#reverse!:

  def test_reverse!
    a="beta"
    assert_equal("ateb",a.reverse!)
    assert_equal("ateb",a)
  end

As you see, the problem was the side-effect, that reverse! didn't only
change string pointed by self, but the other (dupped) string too.

Rubicon has very few tests for mixed call sequences (or side-effects in
general), and is very pragmatic (testing only how things should work, not
how they could break). I'd guess Rubicon gets more comprehensive with time. 

OTOH, some sources recommend testing object oriented code *at least* in
alpha-omega cycle, which we're not even aiming at. But I doubt simple
alpha-omega test would have captured this one.

Checks for this class of errors (namely missed rb_str_modify) could be added
for succ!, next!, upcase!, downcase!, capitalize!, swapcase!, concat, <<,
sub, sub!, gsub, gsub!, strip!, chop!, chomp!, tr!, tr_s!, delete!,
squeeze!, slice!, and global functions sub, sub!, gsub, gsub!, chop!,
chomp!.

Nevertheless I got couple of interesting warnings or bugs:

- TestProc.rb#test_arity, Proc.new{||}.arity is 0 
  instead of expected -1
- TestKernel.rb uses NotImplementError which is not defined
- Regexp.last_match core dumps:

$ ruby -v 
ruby 1.7.0 (2001-01-09) [i686-linux]
$ ruby -e'm = Regexp.last_match'
-e:1: [BUG] Segmentation fault
ruby 1.7.0 (2001-01-09) [i686-linux]
Aborted (core dumped)

So Rubicon is already doing it's job :).

	- Aleksi