Forum: Ruby DRb Problems with Mac OS X 10.5.3

It looks like the Mac OS X 10.5.3 upgrade breaks DRb when it's trying to
open a TCPSocket.

Looking into it now, but be warned:

/opt/local/lib/ruby/1.8/drb/drb.rb:865:in `initialize': getaddrinfo:
nodename nor servname provided, or not known (SocketError)
from /opt/local/lib/ruby/1.8/drb/drb.rb:865:in `open'
from /opt/local/lib/ruby/1.8/drb/drb.rb:865:in `open_server'

-Kurt

I'm running into the same problem too having just upgraded.  Is there a 
solution?


Eric

Kurt Schrader wrote:
> It looks like the Mac OS X 10.5.3 upgrade breaks DRb when it's trying to
> open a TCPSocket.
> 
> Looking into it now, but be warned:
> 
> /opt/local/lib/ruby/1.8/drb/drb.rb:865:in `initialize': getaddrinfo:
> nodename nor servname provided, or not known (SocketError)
> from /opt/local/lib/ruby/1.8/drb/drb.rb:865:in `open'
> from /opt/local/lib/ruby/1.8/drb/drb.rb:865:in `open_server'
> 
> -Kurt

No solution yet, but it looks like something has changed in some 
underlying networking library, as base socket handing methods seem to be 
broken:

[kschrader@einstein:~]$ irb
>> require 'socket'
=> true
>> Socket.getaddrinfo(Socket.gethostname, 0, Socket::AF_INET, Socket::SOCK_STREAM, nil, Socket::AI_CANONNAME)
SocketError: getaddrinfo: nodename nor servname provided, or not known
  from (irb):2:in `getaddrinfo'
  from (irb):2
>> 


Eric Ly wrote:
> I'm running into the same problem too having just upgraded.  Is there a 
> solution?
> 
> 
> Eric
> 
> Kurt Schrader wrote:
>> It looks like the Mac OS X 10.5.3 upgrade breaks DRb when it's trying to
>> open a TCPSocket.
>> 
>> Looking into it now, but be warned:
>> 
>> /opt/local/lib/ruby/1.8/drb/drb.rb:865:in `initialize': getaddrinfo:
>> nodename nor servname provided, or not known (SocketError)
>> from /opt/local/lib/ruby/1.8/drb/drb.rb:865:in `open'
>> from /opt/local/lib/ruby/1.8/drb/drb.rb:865:in `open_server'
>> 
>> -Kurt

Kurt Schrader wrote:
> No solution yet, but it looks like something has changed in some 
> underlying networking library, as base socket handing methods seem to be 
> broken:
> 


Damn, I need a fix bad -- Either that or I'm re-installing my entire OS 
X machine to get rid of 10.5.3.

I'll help anyway I can, but I'm not good with the underlyings.

Please let us know as soon as you find something.

I've not done too much Ruby socket programming, but it seems to be that 
it doesn't like 0 as a service number, because it is interpreting it as 
the real service 0, instead of the empty service... if you use nil 
instead it works fine... you might be able to patch dRb to use that as a 
work around.


irb(main):001:0> require 'socket'
=> true
irb(main):011:0> Socket.getaddrinfo(Socket.gethostname, 0, 
Socket::AF_INET, Socket::SOCK_STREAM, nil, Socket::AI_CANONNAME)
SocketError: getaddrinfo: nodename nor servname provided, or not known
  from (irb):11:in `getaddrinfo'
  from (irb):11
  from :0
irb(main):012:0> Socket.getaddrinfo(Socket.gethostname, nil, 
Socket::AF_INET, Socket::SOCK_STREAM, nil, Socket::AI_CANONNAME)
=> [["AF_INET", 0, "192.168.1.95", "192.168.1.95", 2, 1, 6]]


Kurt Schrader wrote:
> No solution yet, but it looks like something has changed in some 
> underlying networking library, as base socket handing methods seem to be 
> broken:
> 
> [kschrader@einstein:~]$ irb
>>> require 'socket'
> => true
>>> Socket.getaddrinfo(Socket.gethostname, 0, Socket::AF_INET, Socket::SOCK_STREAM, nil, Socket::AI_CANONNAME)
> SocketError: getaddrinfo: nodename nor servname provided, or not known
>   from (irb):2:in `getaddrinfo'
>   from (irb):2
>>> 
> 
> 
> Eric Ly wrote:
>> I'm running into the same problem too having just upgraded.  Is there a 
>> solution?
>> 
>> 
>> Eric
>> 
>> Kurt Schrader wrote:
>>> It looks like the Mac OS X 10.5.3 upgrade breaks DRb when it's trying to
>>> open a TCPSocket.
>>> 
>>> Looking into it now, but be warned:
>>> 
>>> /opt/local/lib/ruby/1.8/drb/drb.rb:865:in `initialize': getaddrinfo:
>>> nodename nor servname provided, or not known (SocketError)
>>> from /opt/local/lib/ruby/1.8/drb/drb.rb:865:in `open'
>>> from /opt/local/lib/ruby/1.8/drb/drb.rb:865:in `open_server'
>>> 
>>> -Kurt

Thanks for the report, we are of course very sorry about that. Ruby
didn't change in Mac OS X 10.5.3, so we are currently investigating
possible regressions in the underlying frameworks. We will let you
know.

Following Andy's idea, what about the following as a temporary fix?
Seems to work for DRb servers with a port of 0 (which seems to be the
default). Apparently DRb servers which explicitly bind to a non 0 port
are not affected by this regression.

class DRb::DRbTCPSocket
  class << self
    alias parse_uri_orig parse_uri
    def parse_uri(*args)
      ary = parse_uri_orig(*args)
      ary[1] = nil if ary[1] == 0
      ary
    end
  end
end

Laurent

Laurent,

That seems to work for now.  Thanks for looking into it.

-Kurt

Laurent Sansonetti wrote:
> class DRb::DRbTCPSocket
>   class << self
>     alias parse_uri_orig parse_uri
>     def parse_uri(*args)
>       ary = parse_uri_orig(*args)
>       ary[1] = nil if ary[1] == 0
>       ary
>     end
>   end
> end

Where does this wonderful piece of code go?

I put this just before "Rails::Initializer.run(:set_load_path)" in 
config/boot.rb

=======
  # HACK
  require 'drb'
  class DRb::DRbTCPSocket
    class << self
      alias parse_uri_orig parse_uri
      def parse_uri(*args)
        ary = parse_uri_orig(*args)
        ary[1] = nil if ary[1] == 0
        ary
      end
    end
  end
=======

Blaz Rubi wrote:
> I put this just before "Rails::Initializer.run(:set_load_path)" in 
> config/boot.rb
> 
> =======
>   # HACK
>   require 'drb'
>   class DRb::DRbTCPSocket
>     class << self
>       alias parse_uri_orig parse_uri
>       def parse_uri(*args)
>         ary = parse_uri_orig(*args)
>         ary[1] = nil if ary[1] == 0
>         ary
>       end
>     end
>   end
> =======

Anyone else get errors from inserting this code?

if you place it in the LINE before 
"Rails::Initializer.run(:set_load_path)" I got a syntax error

script/server:2:in `require': ./script/../config/boot.rb:41: class 
definition in method body (SyntaxError)
  from script/server:2

which makes sense.


if you place it outside of the Boot Class you get a SystemStackError

./script/../config/boot.rb:43:in `parse_uri_orig': stack level too deep 
(SystemStackError)
  from ./script/../config/boot.rb:43:in `parse_uri_orig'
  from /Users/lorgio/Sites/gawkk/config/boot.rb:43:in `parse_uri'
  from 
/System/Library/Frameworks/Ruby.framework/Versions/1.8/usr/lib/ruby/1.8/drb/drb.rb:874:in 
`uri_option'
  from 
/System/Library/Frameworks/Ruby.framework/Versions/1.8/usr/lib/ruby/1.8/drb/drb.rb:780:in 
`uri_option'
  from 
/System/Library/Frameworks/Ruby.framework/Versions/1.8/usr/lib/ruby/1.8/drb/drb.rb:778:in 
`each'
  from 
/System/Library/Frameworks/Ruby.framework/Versions/1.8/usr/lib/ruby/1.8/drb/drb.rb:778:in 
`uri_option'
  from 
/System/Library/Frameworks/Ruby.framework/Versions/1.8/usr/lib/ruby/1.8/drb/drb.rb:1044:in 
`initialize'
  from 
/Library/Ruby/Gems/1.8/gems/actionpack-2.0.2/lib/action_controller/session/drb_store.rb:8:in 
`new'
   ... 51 levels...
  from /Library/Ruby/Gems/1.8/gems/rails-2.0.2/lib/commands/server.rb:39
  from /Library/Ruby/Site/1.8/rubygems/custom_require.rb:27:in 
`gem_original_require'
  from /Library/Ruby/Site/1.8/rubygems/custom_require.rb:27:in `require'
  from script/server:3


Anyone get this working?
Am I placing this in the wrong place? or is there something I missed?

Blaz Rubi wrote:
> I put this just before "Rails::Initializer.run(:set_load_path)" in 
> config/boot.rb
> 
> =======
>   # HACK
>   require 'drb'
>   class DRb::DRbTCPSocket
>     class << self
>       alias parse_uri_orig parse_uri
>       def parse_uri(*args)
>         ary = parse_uri_orig(*args)
>         ary[1] = nil if ary[1] == 0
>         ary
>       end
>     end
>   end
> =======

Incase you get errors and cannot get the hack above, try putting it in 
the environment.rb file INSIDE the config block.

Lorgio Jimenez wrote:
> 
> Incase you get errors and cannot get the hack above, try putting it in 
> the environment.rb file INSIDE the config block.

Yeah that doesn't work either.

A) INSIDE config block of environment.rb,

B) Before the "Rails::Initializer.run(:set_load_path)" line in boot.rb.

Neither A or B work.

Any other ideas?

./script/../config/boot.rb:48: uninitialized constant DRb (NameError)
  from script/server:2:in `require'
  from script/server:2

Jeff Sidlosky wrote:
> ./script/../config/boot.rb:48: uninitialized constant DRb (NameError)
>   from script/server:2:in `require'
>   from script/server:2

You forgot to require DRb.
Add require 'drb' right above the code.

I had the same problems and this fix makes it possible for mongrel to 
start and I thought everything was fine.

But, I can't seem to run any background tasks using backgroundrb.

I get

The error occurred while evaluating nil.object - (NoMethodError)
20080604-13:35:57 (4356) 
vendor/plugins/backgroundrb/server/lib/backgroundrb/middleman.rb:396:in 
`worker'

Line 396 is:
      worker = ex { @jobs[key].object }


Anyone else having trouble?

Davis Kitchel wrote:

> Anyone else having trouble?

Yes.  I'm having problems with backgroundrb too, even with the hack.

I'm running Ruby out of MacPorts and it looks like a lot of others in 
this thread are too.

Is this a problem for those running the version of Ruby included with 
Mac OS X as well?  Looks like Lorgio Jimenez was running into some 
problems using Mac OS X's Ruby.

Does anyone know if/when this is going to be fixed?

Ryan Duryea wrote:

Well, I went and got some answers to my own questions!

> Is this a problem for those running the version of Ruby included with 
> Mac OS X as well?  

Sure seems like it.  Check this out:

% /usr/bin/ruby.orig -v
ruby 1.8.6 (2007-09-24 patchlevel 111) [universal-darwin9.0]

% /usr/bin/ruby.orig -r socket -e "TCPServer.open('localhost', 0)"
-e:1:in `initialize': getaddrinfo: nodename nor servname provided, or 
not known (SocketError)
  from -e:1:in `open'
  from -e:1

> Does anyone know if/when this is going to be fixed?

Keep in mind that I have almost no experience with socket programming, 
but I have submitted a patch for MacPorts under this ticket.  Use at 
your own peril!  It seems to work for me, but your mileage may vary!

https://trac.macports.org/ticket/15528

I just thought I'd tell people what I ended up doing to solve the 
problem so that they might avoid some of the problems I ran into.

I downgraded to 10.5.0 via my leopard install cd using the preserve 
perferences option.

I was hoping it'd leave everything the same, just downgrade the system, 
so all my filesystems would be intact (/usr/local/src, bin etc etc.)

Wrong assumption.  They were scattered all over a folder that was 
created called "Previous System"

At first I thought I could piece it all together, but no way!.  It was 
like a bomb went off and scattered my development environment all over 
the place with no discernible logic.

So

I rebuilt my whole environment from scratch - gems, mysql, ruby, rails, 
subversion etc etc etc.

Total nightmare.

If you can find a way to patch the problem until apple fixes it - do.

Jeff Sidlosky wrote:
> Lorgio Jimenez wrote:
>> 
>> Incase you get errors and cannot get the hack above, try putting it in 
>> the environment.rb file INSIDE the config block.
> 
> Yeah that doesn't work either.
> 
> A) INSIDE config block of environment.rb,
> 
> B) Before the "Rails::Initializer.run(:set_load_path)" line in boot.rb.
> 
> Neither A or B work.
> 
> Any other ideas?
> 
> ./script/../config/boot.rb:48: uninitialized constant DRb (NameError)
>   from script/server:2:in `require'
>   from script/server:2

I put it in environment.rb _outside_ of the config block and it works 
like a champ.

What a drag -- why didn't I hear about this at RailsConf?

Wes

Andy Keep wrote:
> I've not done too much Ruby socket programming, but it seems to be that 
> it doesn't like 0 as a service number, because it is interpreting it as 
> the real service 0, instead of the empty service... if you use nil 
> instead it works fine... you might be able to patch dRb to use that as a 
> work around.


Looks to me like this might not be the reason, as I was able to get it 
to work by using 127.0.0.1 instead of localhost:

>> require 'drb'
=> true
>> DRb.start_service('druby://localhost:0')
SocketError: getaddrinfo: nodename nor servname provided, or not known
.....
>> DRb.start_service('druby://127.0.0.1:0')
=> #<DRb::DRbServer:0x383a24 @grp=#<ThreadGroup:0x3839e8>,
.....

Having lots of problems with this bug. Tried some of the fixes suggested 
in this thread but with no luck. Any idea when we might be getting a 
fix?

Apple fixed the DRb problem in the 10.5.4 update, which was released at 
the
end of June.

http://www.macruby.org/post/47/

Craig

Looks like this has re-surfaced in 10.6 (Snow Leopard). Fresh install 
and I can't start drb with localhost. Using the hack above fixes the 
problem for now, but just though I'd mention it.