Getting "io/console not supported; tty will not be manipulated" on Windows

[davidhooey]

W:\rubyzip>java -version
java version "1.7.0_51"
Java(TM) SE Runtime Environment (build 1.7.0_51-b13)
Java HotSpot(TM) Client VM (build 24.51-b03, mixed mode, sharing)

W:\rubyzip>jruby -v
jruby 1.7.10 (1.9.3p392) 2014-01-09 c4ecd6b on Java HotSpot(TM) Client VM 1.7.0_51-b13 [Windows 7-x86]

W:\rubyzip>jruby -S rake -T
io/console not supported; tty will not be manipulated
rake build    # Build rubyzip-1.1.2.gem into the pkg directory
rake install  # Build and install rubyzip-1.1.2.gem into system gems
rake release  # Create tag v1.1.2 and build and push rubyzip-1.1.2.gem to R...
rake test     # Run tests

[denserv]

This issue prevents a command line app from hiding a user's password when entered via STDIN.

require 'io/console'
password = STDIN.noecho(&:gets)

Text is still echoed back to the console in Windows.

[djberg96]

Please fix.

[mikebridge]

Same error here using rake on windows, e.g. rake -T

    C:\> jruby -v
    jruby 1.7.19 (1.9.3p551) 2015-01-29 20786bd on Java HotSpot(TM) 64-Bit Server VM 1.8.0_31-b13 +jit [Windows 8.1-amd64]

    C:\>java -version
    java version "1.8.0_31" Java(TM) SE Runtime Environment (build 1.8.0_31-b13) Java HotSpot(TM) 64-Bit Server VM (build 25.31-b07, mixed mode)

[PragTob]

I'm getting the same thing on the appveyor (CI for Windows). Quite curiosly I have a build where the error pops up after gem install bundler and it continues to run and one where the error pops up and console output halts. Unfortunately there are more of the latter ones... :-/

That the output stops could also be on appveyors side...

edit: for me that's using jruby-1.7.19
edit2: And of course I'm an idiot and deleted that appveyor project together with all history... I have a new one though that will soon exhibit similar issues: https://ci.appveyor.com/project/PragTob/shoes4/build/1.0.2

[hathawad]

Getting as well on:
jruby 1.7.19 (1.9.3p551) 2015-01-29 20786bd on Java HotSpot(TM) 64-Bit Server VM 1.8.0_31-b13 +jit [Windows Server 2008 R2-amd64]

[mjstrasser]

I am getting the same message on Windows and Solaris with JRuby 1.7.19. It does not occur on JRuby 1.7.18.

Windows

D:\>jruby -v
jruby 1.7.19 (1.9.3p551) 2015-01-29 20786bd on Java HotSpot(TM) Client VM 1.7.0_67-b01 +jit [Windows 7-x86]

D:\>gem list
io/console not supported; tty will not be manipulated

Solaris

-bash-3.2$ jruby -v
jruby 1.7.19 (1.9.3p551) 2015-01-29 20786bd on Java HotSpot(TM) Server VM 1.7.0_67-b01 +jit [SunOS-sparc]
-bash-3.2$ gem list
io/console not supported; tty will not be manipulated

[headius]

This is usually caused when some native support can't load. Windows should be falling back on a dummy implementation.

Of course I just got back up and running in my OS X env, so I'm not going to be looking at this in the short term, but if someone wants to poke around the code is in lib/ruby/1.9/io/console.rb. At the very least it should be possible to sort out what supressed errors are causing it to fall into the noisy path.

[djberg96]

For solaris, it looks to me like you're just missing a sunos/solaris string in the regex check.

[enebo]

For Solaris I would not expect to see the error since it will try and stty and I would have thought that would work on Solaris but @djberg96 is right in that we should probably be following the first if in io/console.rb:

if RbConfig::CONFIG['host_os'].downcase =~ /darwin|openbsd|freebsd|netbsd|linux/

@mjstrasser can you print out RbConfig::CONFIG['host_os'] on Solaris. I believe it will be 'sunos' but I
would like to make sure. You could also edit lib/ruby/1.9/io/console.rb to see if adding solaris host_os to that regexp fixes it. What is weird about the Solaris observation is that we have not change io/console.rb on 1.7 in almost exactly 2 years. So we either made some seemingly unrelated changes in IO which is causing the stty path to break or perhaps something maybe changed in your env recently? In any case, we should go for real native support and not rely on stty on Solaris so let's fix the regexp...

For Windows, we clearly never do anything but try to stty which we know will fail in the next if statement where we unconditionally warn and setup dummy IO. So I guess we can at least eliminate the bogus stty check (unless perhaps we try it in cygwin or mingw? ), but that does not solve the warning and warning is meaningful. Short of trying to port windows side of ext/io/console.c I am not sure the best way to solve this.

[mjstrasser]

It returns solaris. I also noticed that JRuby 1.7.19 outputs a different message on JDK 6:

-bash-3.2$ export JAVA_HOME=/usr/jdk/instances/jdk1.6.0
-bash-3.2$ jruby -v
jruby 1.7.19 (1.9.3p551) 2015-01-29 20786bd on Java HotSpot(TM) Server VM 1.6.0_37-b06 +jit [SunOS-sparc]
-bash-3.2$ jruby -e 'puts RbConfig::CONFIG["host_os"]'
io/console on JRuby shells out to stty for most operations
solaris

On JDK 7:

-bash-3.2$ export JAVA_HOME=/usr/jdk/instances/jdk1.7.0_67
-bash-3.2$ jruby -v
jruby 1.7.19 (1.9.3p551) 2015-01-29 20786bd on Java HotSpot(TM) Server VM 1.7.0_67-b01 +jit [SunOS-sparc]
-bash-3.2$ jruby -e 'puts RbConfig::CONFIG["host_os"]'
io/console not supported; tty will not be manipulated
solaris

[enebo]

@mjstrasser If you could modify the regexp in io/console.rb and try it out it would help in resolving this since I have no immediate access to Solaris instance. Your observation re Java 6/7 is interesting...perhaps we are not properly invoking processes correctly? on Java 6 it appears the stty path must have worked but not on Java 7 :|

[mjstrasser]

I tried the following edit:

-bash-3.2$ diff console.rb.orig console.rb
24c24
< if RbConfig::CONFIG['host_os'].downcase =~ /darwin|openbsd|freebsd|netbsd|linux/
---
> if RbConfig::CONFIG['host_os'].downcase =~ /darwin|openbsd|freebsd|netbsd|linux|solaris/
28c28
<     if RbConfig::CONFIG['host_os'].downcase =~ /darwin|openbsd|freebsd|netbsd/
---
>     if RbConfig::CONFIG['host_os'].downcase =~ /darwin|openbsd|freebsd|netbsd|solaris/

Then, with $VERBOSE=true:

-bash-3.2$ export JAVA_HOME=/usr/jdk/instances/jdk1.7.0_67
-bash-3.2$ jruby -ve 'puts RbConfig::CONFIG["host_os"]'
jruby 1.7.19 (1.9.3p551) 2015-01-29 20786bd on Java HotSpot(TM) Server VM 1.7.0_67-b01 +jit [SunOS-sparc]
failed to load native console support: Could not load FFI Provider: (NotImplementedError) FFI not available: null
 See http://jira.codehaus.org/browse/JRUBY-4583
io/console not supported; tty will not be manipulated
solaris

On JDK 6:

-bash-3.2$ export JAVA_HOME=/usr/jdk/instances/jdk1.6.0
-bash-3.2$ jruby -ve 'puts RbConfig::CONFIG["host_os"]'
jruby 1.7.19 (1.9.3p551) 2015-01-29 20786bd on Java HotSpot(TM) Server VM 1.6.0_37-b06 +jit [SunOS-sparc]
failed to load native console support: Could not load FFI Provider: (NotImplementedError) FFI not available: null
 See http://jira.codehaus.org/browse/JRUBY-4583
io/console on JRuby shells out to stty for most operations
solaris

Does this shed any light?