On Wed, Jan 05, 2011 at 04:27:35PM -0500, Benjamin R. Haskell wrote:
> Maybe it makes sense, but it seems weird that the tests are conditioned 
> on running Ruby scripts rather than looking at `ri --version`.
ri --version reports something weird.  I don't know since when
things were changed, checking for modules that have been removed
or added seems to be fair enough.

Are you thinking there's too much (and too direct) ruby
invocations?

> And how does this fare against different Ruby implementations?  Would 
> JRuby 1.9.2 necessarily use the same 'ri' conventions as MRI Ruby 1.9.2?
Yes.  Latest JRuby supports both 1.8 and 1.9 Ruby versions and
have copies of standard runtime libraries (both versions) which
includes RI.

> And more generally, since it's looking in directories that would only 
> contain Ruby documentation, would it hurt to include both .yaml and .ri 
> for versions that support .ri?
I thought about it, yes.  Newer RI doesn't read .yaml
documentation.  It wouldn't hurt, but it would complete something
that can't be shown.


This patch will un-break ri completion for latest stable MRI Ruby
version.  I wrote this completion and I am responsible for that.
However, I am not satisfied with current implementation, I'm
working on it.

-- 
Regards,    --
Sir Raorn.   --- http://thousandsofhate.blogspot.com/