From: Trans Date: 2007-12-23T03:12:06+09:00 Subject: Re: Status of two issues On Dec 22, 8:58 am, Yukihiro Matsumoto wrote: > Hi, > > In message "Re: Status of two issues" > on Sat, 22 Dec 2007 22:47:57 +0900, Trans writes: > > |> Could you submit URL or mail ID [ruby-core:xxxx] along with your > |> questions, please? > | > |http://rubyforge.org/tracker/index.php?func=detail&aid=15213&group_id... > | > |http://rubyforge.org/tracker/index.php?func=detail&aid=14947&group_id... > > Thank you. > > Regarding the former, autoload is still done by direct C function > callm but it is perhaps solved by RubyGems integration in 1.9, but I > have not yet confirmed. Anyone? > > I don't consider the latter a bug, rather it's only an application of > DuckTyping (or ChikenTyping one might say). Thanks for looking at these, Matz. For the autoload, the problem arises in that I'm using my own special system for loading files. It's not just a Gems issue per se. My system works great, but there is no way to override how autoload loads files. B/c of this I can't use my system on anything that uses autoload -- which is a serious show-stopper. For the other, I take it you did not follow the conversion about this on ruby-talk. The problem is that if someone defines an Object#meta or Kernel#meta extension method it will cause open-uri to bomb. Besides that, open_uri is simply using the #respond_to? to test to see if its Meta module has been included, so changing it to is_a? is actually more precise to the authors intention. T.