From: Charles Oliver Nutter <headius@...>
Date: 2011-03-10T06:35:55+09:00
Subject: [ruby-core:35473] Re: MRI silently handles .so files as Ruby sources

On Wed, Mar 9, 2011 at 2:59 PM, Simone Carletti <weppos@weppos.net> wrote:
> Hello,
> today I noticed that if you have a file called filename.so.rb and you try to
> require it without specifying the extension, MRI silently understands that
> it isn't a .so file, will load the content and treat it as Ruby source.
>>
>> # filename.so.rb
>> puts __FILE__
>> # somefile.rb
>> require 'filename.so'
>> # => /../../filename.so.rb
>
> Other interpreters handle the case in a different way.
> For example, JRuby won't load the file because it looks for a Ruby extension
> in a file called filename.so.
> See��http://jira.codehaus.org/browse/JRUBY-5569
> I had a small chat with��Charles Oliver Nutter (you can read the comments in
> the JIRA ticket) and I'm curious to know if the way MRI works should be
> considered a bug, a coincidence or a feature.
> Thanks,

Just when I thought I knew all load-searching oddities :)

If you look at the JRuby bug you'll see that it's a very specific case
and I show several scenarios that either work or don't work...

* Only fails over for "extension" extensions like .so or
.bundle...does not fail over for things like foo.rb.rb or foo.rb.so
* Appears to do a search for both .rb and <extension extension> if the
original filename doesn't work
* Does not fail over to "extension" extensions that are not
appropriate for current platform, like .so.so on OS X

I think it's confusing and unpredictable and would just as soon have
the logic be "if it ends with an extension extension, only search
extension extensions."

- Charlie