From: Mark Probert <probertm@...>
Date: 2005-06-02T05:19:33+09:00
Subject: net::telnet and broken pipe


Hi ..

I am getting a consistent broken pipe error on my threaded telnet application 
and I am not sure of why.  

The error is 

 .. backtrace:
 /usr/local/lib/ruby/1.8/net/telnet.rb:601:
  in `syswrite'./bin/sercoll.rb:249:
  in `join'./bin/sercoll.rb:249:
  in `run_t'./bin/sercoll.rb:249:
  in `each'./bin/sercoll.rb:249:
  in `run_t'./bin/sercoll.rb:328:
  in `run'./bin/sercoll.rb:565

Which is pointing to the following code in telnet.rb

    def write(string)
      length = string.length
      while 0 < length
        IO::select(nil, [@sock])
        @dumplog.log_dump('>', string[-length..-1]) \
            if @options.has_key? ("Dump_log")
        length -= @sock.syswrite(string[-length..-1])  ### line 601
      end
    end

and the following in my code:

    def run_t
        threads = []
        @nodes.each do |ip_addr|
            threads << Thread.new(ip_addr) do |nd|
                nn, ip, usr, pwd = nd.split(/:/)
                GC.start
                ### wrapper around telnet                
                collect(usr, pwd, ip, nn, trc, log)
            end
        end
        threads.each { |thr| thr.join }   ### line 249
    end

The problem seems to be related to the number of threads that I am running 
(there is logic to limit threads to 20 and to do the threading in batches.. 
this problem only occurs if this logic is implemented, however, there is 
nothing to indicate that that the logic is broken, just the broken pipe in 
the thread join).

$ ruby -v
ruby 1.8.2 (2004-12-25) [sparc-solaris2.8]

Anyone have any thoughts or ideas on how to proceed?  

-- 
-mark.  (probertm at acm dot org)