[MacRuby] #1249: Webrick escaping is wrong

[MacRuby]

#1249: Webrick escaping is wrong
--------------------------------+-------------------------------------------
 Reporter:  cyberfox@…          |       Owner:  lsansonetti@…        
     Type:  defect              |      Status:  new                  
 Priority:  major               |   Milestone:                       
Component:  MacRuby             |    Keywords:  regexp, escaping     
--------------------------------+-------------------------------------------
 Greetings,
 Something's wacky with HTTPUtils escaping, it escapes 'v' and 'z', when it
 shouldn't.

 {{{
 $ macirb
 irb(main):001:0> require 'webrick'
 => true
 irb(main):002:0>
 WEBrick::HTTPUtils.escape('http://abcdefghijklmnopqrstuvwxyz.com')
 => "http://abcdefghijklmnopqrstu%76wxy%7A.com"
 }}}

 versus 1.9.2

 {{{
 $ irb
 ruby-1.9.2-p180 :001 > require 'webrick'
  => true
 ruby-1.9.2-p180 :002 >
 WEBrick::HTTPUtils.escape('http://abcdefghijklmnopqrstuvwxyz.com')
  => "http://abcdefghijklmnopqrstuvwxyz.com"
 }}}

 There's a FIXME in webrick/httputils.rb about non-ascii regexps, which is
 why the 'z' is broken, but the 'v' is a magical problem with the code that
 generates the control character regex. :(

 I've fixed it in my branch by replacing some of the constants as follows:

 {{{
     UNESCAPED = /([\x00-\x20\x7F\<\>\#%"'{}|\\^\[\]`\x80-\xFF])/
     UNESCAPED_FORM =
 /([\x00-\x1F;\/?:@&=+\$,\<\>\#%"'{}|\\^\[\]`\x7F\x80-\xFF])/
     NONASCII  = /([\x80-\xFF])/
 }}}

 By using ranged regexes instead of individual character regexes, it
 doesn't run into the uncomfortable confusion of \v being vertical tab, but
 also being interpreted as a 'v' character, and hopefully avoids whatever
 issues was being 'FIXME'ed for the NONASCII case.

 My fix may not be the best one, but the WEBrick HTTPUtils escape method is
 broken without some kind of a fix, though, which is a pretty serious bug.

 --  Morgan

-- 
Ticket URL: <http://www.macruby.org/trac/ticket/1249>
MacRuby <http://macruby.org/>