Socket.gethostname ignores encoding settings #2558

jsvd · 2015-02-04T09:23:28Z

tested on windows but also happens on macosx. I can't make Socket.gethostname return a UTF-8 string even when fiddling with Encoding.default_internal and external settings.

Example:

C:\Users\Jsvd>irb
io/console not supported; tty will not be manipulated
irb(main):001:0>  puts ''.encoding; require 'socket'; puts Socket.gethostname.encoding
Windows-1252
Windows-1252
=> nil
irb(main):002:0>

C:\Users\Jsvd>irb -EUTF-8
io/console not supported; tty will not be manipulated
irb(main):001:0> puts ''.encoding; require 'socket'; puts Socket.gethostname.encoding
UTF-8
Windows-1252
=> nil
irb(main):002:0>

C:\Users\Jsvd>jruby -v
jruby 1.7.17 (1.9.3p392) 2014-12-09 fafd1a7 on Java HotSpot(TM) 64-Bit Server VM 1.8.0_31-b13 +jit [Windows 7-amd64]

jordansissel · 2015-02-05T18:30:41Z

MRI 2.1.4 also has this bug.

headius · 2015-02-24T20:12:43Z

I believe this would fix it. I noticed gethostbyname and others are not using this path either.

If someone can confirm this fix (I'm on Windows and struggling with dev env at the moment), we can get it merged in.

diff --git a/core/src/main/java/org/jruby/ext/socket/SocketUtils.java b/core/src/main/java/org/jruby/ext/socket/SocketUtils.java        
index b1a457e..65110a7 100644                                                                                                           
--- a/core/src/main/java/org/jruby/ext/socket/SocketUtils.java                                                                          
+++ b/core/src/main/java/org/jruby/ext/socket/SocketUtils.java                                                                          
@@ -74,12 +74,12 @@ public class SocketUtils {                                                                                          
         Ruby runtime = context.runtime;                                                                                                

         try {                                                                                                                          
-            return runtime.newString(InetAddress.getLocalHost().getHostName());                                                        
+            return RubyString.newInternalFromJavaExternal(runtime, InetAddress.getLocalHost().getHostName());                          

         } catch(UnknownHostException e) {                                                                                              

             try {                                                                                                                      
-                return runtime.newString(InetAddress.getByAddress(new byte[]{0,0,0,0}).getHostName());                                 
+                return RubyString.newInternalFromJavaExternal(runtime, InetAddress.getByAddress(new byte[]{0, 0, 0, 0}).getHostName());

             } catch(UnknownHostException e2) {                                                                                         
                 throw sockerr(runtime, "gethostname: name or service not known");

I'm confused why it would be Windows-1252, though. At least on 9k, the logic for runtime.newString above will always use UTF-8 as the string encoding by default.

Fixes #2558.

headius · 2015-07-02T20:43:20Z

Fixed in 9k.rc2 and 1.7.21.

jsvd · 2015-10-26T11:10:00Z

This is still an issue on 1.7.22

C:\Users\Jsvd>jruby -v
jruby 1.7.22 (1.9.3p551) 2015-08-20 c28f492 on Java HotSpot(TM) 64-Bit Server VM 1.8.0_31-b13 +jit [Windows 7-amd64]

C:\Users\Jsvd>jruby -S irb -EUTF-8
io/console not supported; tty will not be manipulated
irb(main):001:0> puts ''.encoding; require 'socket'; puts Socket.gethostname.encoding
UTF-8
Windows-1252
=> nil
irb(main):003:0> s = TCPServer.new(3335)
=> #<TCPServer:fd 15>
irb(main):005:0> c = TCPSocket.new("localhost", 3335)
=> #<TCPSocket:fd 16>
irb(main):006:0> c.peeraddr
=> ["AF_INET", 3335, "127.0.0.1", "127.0.0.1"]
irb(main):007:0> c.peeraddr.first.encoding
=> #<Encoding:Windows-1252>
irb(main):008:0> c.peeraddr.last.encoding
=> #<Encoding:Windows-1252>
irb(main):009:0> "".encoding
=> #<Encoding:UTF-8>
irb(main):010:0>

jsvd · 2016-04-12T15:51:48Z

a shout out that this is still an issue in 1.7.25-snapshot as of 12th of april 2016:

C:\projects\logstash>irb -EUTF-8
irb(main):001:0> RUBY_DESCRIPTION
=> "jruby 1.7.25-SNAPSHOT (1.9.3p551) 2016-04-12 5693183 on Java HotSpot(TM) 64-Bit Server VM 1.8.0_65-b17 +jit [Windows 7-amd64]"
irb(main):002:0> puts ''.encoding; require 'socket'; puts Socket.gethostname.encoding
UTF-8
Windows-1252
=> nil
irb(main):003:0> s = TCPServer.new(3335); c = TCPSocket.new("localhost", 3335)
=> #<TCPSocket:fd 14>
irb(main):004:0> c.peeraddr
=> ["AF_INET", 3335, "127.0.0.1", "127.0.0.1"]
irb(main):005:0> c.peeraddr.first.encoding
=> #<Encoding:Windows-1252>
irb(main):006:0> c.peeraddr.last.encoding
=> #<Encoding:Windows-1252>
irb(main):007:0> "".encoding
=> #<Encoding:UTF-8>

nirvdrum · 2016-04-13T16:04:00Z

While I agree that it's a bit confusing, neither Encoding.default_internal nor Encoding.default_external are documented to do anything with sockets. MRI also appears to ignore them for the same socket calls. Changing that behavior in JRuby would then break compatibility with MRI, as far as I can tell.

kares · 2020-03-31T11:22:28Z

an update here, most of these respect default_external, Socket.gethostname is still problematic :

jruby 9.2.11.1 (2.5.7) 2020-03-25 b1f55b1a40 OpenJDK 64-Bit Server VM 11+28 on 11+28 +jit [mswin32-x86_64]
2020-03-31T02:40:37.913-07:00 [main] WARN FilenoUtil : Native subprocess control requires open access to sun.nio.ch
Pass '--add-opens java.base/sun.nio.ch=org.jruby.dist' or '=org.jruby.core' to enable.
irb(main):001:0> puts ''.encoding; require 'socket'; puts Socket.gethostname.encoding
Windows-1252
Windows-1252
=> nil
irb(main):002:0>
irb(main):003:0> s = TCPServer.new(3335); c = TCPSocket.new("localhost", 3335)
=> #<TCPSocket:fd 100027>
irb(main):004:0> s
=> #<TCPServer:fd 100026>
irb(main):006:0> c.peeraddr
=> ["AF_INET", 3335, "127.0.0.1", "127.0.0.1"]
irb(main):008:0> c.peeraddr.first.encoding
=> #<Encoding:UTF-8>
irb(main):009:0> c.peeraddr.map { |addr| addr.is_a?(String) ? addr.encoding : addr }
=> [#<Encoding:UTF-8>, 3335, #<Encoding:UTF-8>, #<Encoding:UTF-8>]
irb(main):010:0> ''.encoding
=> #<Encoding:Windows-1252>

C:\jruby-9.2.11.1>bin\jruby -v -S irb -EUTF-8
jruby 9.2.11.1 (2.5.7) 2020-03-25 b1f55b1a40 OpenJDK 64-Bit Server VM 11+28 on 11+28 +jit [mswin32-x86_64]
2020-03-31T02:45:02.680-07:00 [main] WARN FilenoUtil : Native subprocess control requires open access to sun.nio.ch
Pass '--add-opens java.base/sun.nio.ch=org.jruby.dist' or '=org.jruby.core' to enable.
irb(main):001:0>  puts ''.encoding; require 'socket'; puts Socket.gethostname.encoding
UTF-8
Windows-1252
=> nil
irb(main):002:0>
irb(main):003:0> s = TCPServer.new(3335); c = TCPSocket.new("localhost", 3335)
=> #<TCPSocket:fd 100027>
irb(main):004:0> c.peeraddr
=> ["AF_INET", 3335, "127.0.0.1", "127.0.0.1"]
irb(main):005:0> c.peeraddr.first
=> "AF_INET"
irb(main):006:0> c.peeraddr.map { |addr| addr.is_a?(String) ? addr.encoding : addr }
=> [#<Encoding:UTF-8>, 3335, #<Encoding:UTF-8>, #<Encoding:UTF-8>]

headius · 2020-03-31T21:16:23Z

@kares And CRuby does this differently I assume? There's no examples on MRI.

kares · 2020-04-01T06:34:09Z

sorry, did not want to get into the hustle of installing MRI on Windows 😒
assumed JRuby could follow ~ same logic as on non-Windows for consistency...

in either case, ENV keys/values are platform specific regardless of setting external.
hostname is ASCII-8BIT (sounds like a bug) would have assumed US-ASCII if not respecting external :

C:\Ruby26-x64>bin\irb -EUTF-8
irb(main):001:0> ENV['USERNAME'].encoding
=> #<Encoding:IBM437>
irb(main):002:0> Encoding.default_external
=> #<Encoding:UTF-8>
irb(main):003:0> ENV.keys.first.encoding
=> #<Encoding:IBM437>

irb(main):005:0> require 'socket'
=> true
irb(main):006:0> Socket.gethostname
=> "MSEDGEWIN10"
irb(main):007:0> Socket.gethostname.encoding
=> #<Encoding:ASCII-8BIT>

headius · 2020-04-02T22:53:02Z

Still unclear whether there's a difference from MRI here. If so, someone should open a separate bug for that since the original issue here was patched and closed (though maybe not fixed the way everyone wants).

If both JRuby and MRI have the same bug here, then we'd need to work with ruby-core to determine if this should be fixed. JRuby will follow MRI's lead.

jsvd mentioned this issue Feb 4, 2015

fix encoding issue on windows+jruby elastic/logstash#2507

Closed

headius added stdlib encoding JRuby 1.7.x labels Feb 24, 2015

headius added this to the 9.0.0.0.pre2 milestone Feb 24, 2015

headius modified the milestones: 9.0.0.0.rc1, 9.0.0.0.pre2 Apr 17, 2015

enebo modified the milestones: JRuby 9.0.0.0.rc1, JRuby 9.0.0.0.rc2 Jun 10, 2015

headius added a commit that referenced this issue Jul 2, 2015

Fix Socket.gethostname and IPSocket.getaddress encoding.

af48e11

Fixes #2558.

headius closed this as completed in 0429799 Jul 2, 2015

This was referenced Oct 26, 2015

force encoding for source host logstash-plugins/logstash-input-gelf#23

Closed

Encoding issue in JRuby on windows produces non utf8 from IPSocket.recvfrom elastic/logstash#4090

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

GitHub Sponsors

Socket.gethostname ignores encoding settings #2558

Socket.gethostname ignores encoding settings #2558

jsvd commented Feb 4, 2015

jordansissel commented Feb 5, 2015

headius commented Feb 24, 2015

headius commented Jul 2, 2015

jsvd commented Oct 26, 2015

jsvd commented Apr 12, 2016

nirvdrum commented Apr 13, 2016

kares commented Mar 31, 2020

headius commented Mar 31, 2020

kares commented Apr 1, 2020

headius commented Apr 2, 2020

Socket.gethostname ignores encoding settings #2558

Socket.gethostname ignores encoding settings #2558

Comments

jsvd commented Feb 4, 2015

jordansissel commented Feb 5, 2015

headius commented Feb 24, 2015

headius commented Jul 2, 2015

jsvd commented Oct 26, 2015

jsvd commented Apr 12, 2016

nirvdrum commented Apr 13, 2016

kares commented Mar 31, 2020

headius commented Mar 31, 2020

kares commented Apr 1, 2020

headius commented Apr 2, 2020