Issue #11632 has been updated by Hannes Georg.


I've created a small script to demonstrate the behavior. The following script should print "test" but it prints some gibberish bytes in ruby 2.1.7 and 2.2.3.
~~~
require 'resolv'

msg = Resolv::DNS::Message.new
# this generates ~ 17000 bytes
500.times.each do |i|
  msg.add_answer("abcdefghijklmn%04d." % i,300,Resolv::DNS::Resource::IN::A.new("0.0.0.0"))
end
msg.add_answer("test.",300,Resolv::DNS::Resource::IN::A.new("0.0.0.0"))
msg.add_answer("test.",300,Resolv::DNS::Resource::IN::A.new("0.0.0.0"))
encoded = msg.encode
decoded = Resolv::DNS::Message.decode(encoded)
puts decoded.answer.last[0].to_s.inspect
~~~

----------------------------------------
Bug #11632: Resolv::DNS::Message.encode fails to encode messages larger than 16383 byte
https://bugs.ruby-lang.org/issues/11632#change-54627

* Author: Hannes Georg
* Status: Open
* Priority: Normal
* Assignee: 
* ruby -v: 
* Backport: 2.0.0: UNKNOWN, 2.1: UNKNOWN, 2.2: UNKNOWN
----------------------------------------
Hello Rubyists

The ruby dns message encoder automatically remembers _all_ label addresses to be reused later ( see https://github.com/ruby/ruby/blob/v2_2_3/lib/resolv.rb#L1470 ). The address field however is limited to 14 bits ( see https://tools.ietf.org/html/rfc1035#section-4.1.4 ). If the message gets larger than 16383 bytes the addresses of new labels won't fit anymore. The encoder takes this into account and truncates the addresses on write which makes them invalid ( see https://github.com/ruby/ruby/blob/v2_2_3/lib/resolv.rb#L1467 ).

My suggested solution is to not store addresses larger than 16383. Is a github pr okay for that?



-- 
https://bugs.ruby-lang.org/