IPv6 server entries are not handled correctly

[verbit]

trafficstars

Describe the bug

When using an IPv6 as a server entry, the hostname/port is not parsed correctly (see rex -d output below). For example, with 2001:db8::1 it uses 2001 as the host and db8 as the port.

Expected behavior

I expect it to use 2001:db8::1 as the address and 23 as the port (default).

How to reproduce it

A code sample to reproduce the issue is attached.

Code example

use Rex -feature => [ '1.4', 'exec_autodie' ];

group myservers => '2001:db8::1';

task 'uptime', group => 'myservers', sub { run 'uptime'; };

Additional context

rex -d logs

me@mymachine rextest % rex -d uptime
[2023-06-03 19:03:42] DEBUG - This is Rex version: 1.14.2
[2023-06-03 19:03:42] DEBUG - Command Line Parameters
[2023-06-03 19:03:42] DEBUG - 	d = 1
[2023-06-03 19:03:42] DEBUG - Creating lock-file (Rexfile.lock)
[2023-06-03 19:03:42] DEBUG - Loading Rexfile
[2023-06-03 19:03:42] DEBUG - Enabling task_chaining_cmdline_args feature
[2023-06-03 19:03:42] DEBUG - Activating new template engine.
[2023-06-03 19:03:42] DEBUG - Disabling usage of a tty
[2023-06-03 19:03:42] DEBUG - Activating autodie.
[2023-06-03 19:03:42] DEBUG - Using Net::OpenSSH if present.
[2023-06-03 19:03:42] DEBUG - Add service check.
[2023-06-03 19:03:42] DEBUG - Setting set() to not append data.
[2023-06-03 19:03:42] DEBUG - Registering CMDB as template variables.
[2023-06-03 19:03:42] DEBUG - activating featureset >= 0.51
[2023-06-03 19:03:42] DEBUG - activating featureset >= 0.40
[2023-06-03 19:03:42] DEBUG - activating featureset >= 0.35
[2023-06-03 19:03:42] DEBUG - activating featureset >= 0.31
[2023-06-03 19:03:42] DEBUG - Enabling exec_autodie
[2023-06-03 19:03:42] DEBUG - Creating new distribution class of type: Base
[2023-06-03 19:03:42] DEBUG - new distribution class of type Rex::TaskList::Base created.
[2023-06-03 19:03:42] DEBUG - Creating task: uptime
[2023-06-03 19:03:42] DEBUG - Found Net::OpenSSH and Net::SFTP::Foreign - using it as default
[2023-06-03 19:03:42] DEBUG - Registering task: uptime
[2023-06-03 19:03:42] DEBUG - Initializing Logger from parameters found in Rexfile
[2023-06-03 19:03:42] DEBUG - Returning existing distribution class of type: Rex::TaskList::Base
[2023-06-03 19:03:42] DEBUG - Returning existing distribution class of type: Rex::TaskList::Base
[2023-06-03 19:03:42] INFO - Running task uptime on 2001:db8::1
[2023-06-03 19:03:42] DEBUG - Rex::Group::Entry::Server (private_key): returning
[2023-06-03 19:03:42] DEBUG - Rex::Group::Entry::Server (public_key): returning
[2023-06-03 19:03:42] DEBUG - $VAR1 = '';

[2023-06-03 19:03:42] DEBUG - Auth-Information inside Task:
[2023-06-03 19:03:42] DEBUG - public_key => [[]]
[2023-06-03 19:03:42] DEBUG - password => [[%s]]
[2023-06-03 19:03:42] DEBUG - sudo_password => [[**********]]
[2023-06-03 19:03:42] DEBUG - auth_type => [[try]]
[2023-06-03 19:03:42] DEBUG - private_key => [[]]
[2023-06-03 19:03:42] DEBUG - sudo => [[]]
[2023-06-03 19:03:42] DEBUG - user => [[me]]
[2023-06-03 19:03:42] DEBUG - port => [[]]
[2023-06-03 19:03:42] DEBUG - Using Net::OpenSSH for connection
[2023-06-03 19:03:42] DEBUG - Using user: me
[2023-06-03 19:03:42] DEBUG - Connecting to 2001:db8 (me)
[2023-06-03 19:03:42] DEBUG - get_openssh_opt()
[2023-06-03 19:03:42] DEBUG - $VAR1 = {};

[2023-06-03 19:03:42] DEBUG - OpenSSH: key_auth or not defined: 2001:db8 - me
[2023-06-03 19:03:42] DEBUG - OpenSSH options:
[2023-06-03 19:03:42] DEBUG - $VAR1 = [
          '2001',
          'user',
          'me',
          'port',
          'db8',
          'master_opts',
          [
            '-o',
            'ConnectTimeout=2',
            '-o',
            'LogLevel=QUIET'
          ],
          'default_ssh_opts',
          $VAR1->[6]
        ];

[2023-06-03 19:03:42] DEBUG - OpenSSH constructor options:
[2023-06-03 19:03:42] DEBUG - $VAR1 = {};

[2023-06-03 19:03:42] DEBUG - Trying following auth types:
[2023-06-03 19:03:42] DEBUG - $VAR1 = [
          'key',
          'pass'
        ];

Bad port 'db8'
Bad port 'db8'
Bad port 'db8'
Bad port 'db8'
Bad port 'db8'
Bad port 'db8'
[2023-06-03 19:03:42] WARN - Can't authenticate against 2001 (unable to establish master SSH connection: master process exited unexpectedly)
[2023-06-03 19:03:42] DEBUG - Connections in queue: 1
[2023-06-03 19:03:42] DEBUG - Destroying all cached os information
[2023-06-03 19:03:43] DEBUG - Waiting for children to finish
[2023-06-03 19:03:43] DEBUG - Need to reinitialize connections.
[2023-06-03 19:03:43] DEBUG - Returning existing distribution class of type: Rex::TaskList::Base
[2023-06-03 19:03:43] ERROR - 1 out of 1 task(s) failed:
[2023-06-03 19:03:43] ERROR - 	uptime failed on 2001:db8::1
[2023-06-03 19:03:43] ERROR - 		Couldn't authenticate against 2001:db8::1. It may be caused by one or more of:
[2023-06-03 19:03:43] ERROR - 		 - wrong username, password, key or passphrase
[2023-06-03 19:03:43] ERROR - 		 - changed remote host key
[2023-06-03 19:03:43] ERROR - 		 at /Users/me/perl5/lib/perl5/Rex/TaskList/Base.pm line 340.
[2023-06-03 19:03:43] DEBUG - Removing lockfile
[2023-06-03 19:03:43] DEBUG - Returning existing distribution class of type: Rex::TaskList::Base

My suspicion is that the error lies here https://github.com/RexOps/Rex/blob/6fe5da1832927b22ad940248fa15cc393c224169/lib/Rex/Interface/Connection/OpenSSH.pm#L67

Rex::Helper::IP::get_server_and_port expects a string for the $server but gets a server entry object instead which makes all the is_ip functions inside get_server_and_port fail. Maybe passing "$server" would solve the issue?

Rex version

(R)?ex 1.14.2

Perl version

This is perl 5, version 30, subversion 3 (v5.30.3) built for darwin-thread-multi-2level

Operating system running rex

macOS 12.0.1 (21A559)

Operating system managed by rex

none

How rex was installed?

cpan client

[ferki]

Thanks for the report, @verbit!

At first glance it indeed looks like $server is not stringifed correctly when passed as an argument to is_ip* calls in Rex::Helper::IP.

I expect adding double quotes, forcing an early stringification, or changing what is being passed to get_server_and_port() in the first place may solve this. Perhaps get_server_and_port() may throw a warning or even error if it receives an unexpected argument format.

Either way, it certainly would need additional tests. t/issue/1008.t already has quite a few related tests, but not for the case when a server entry object is being passed to it.