Encoding error while installing rails on ruby 1.9.2

[Bek Bek]

Hello everybody,

I am making my first steps on Ruby on Rails. I have a very simple
question. When I am trying to install rails (as well as every gem) I am
getting this error:

ERROR: While executing gem ... (Encoding::ConverterNotFoundError)
code converter not found (UTF-16LE to IBM737)

My Operating System is Microsoft Windows XP and I have installed Ruby
1.9.2. With previous releases of Ruby (for example 1.8.7) everything is
fine and gems installed successfully. But in version 1.9.2, when i try
"gem install GEMNAME" in command prompt I get this error.

Thank you in advance

 

[Brian Candler]

Please be more specific. Which *exact* version of ruby 1.9.2 did you
install, and where did you get it from? For example was it a Cygwin
package, the rubyinstaller package, some other pre-compiled package, or
did you build it yourself from source? In the latter case, what compiler
did you use, and what options did you configure it with?

Also, what output do you get from these commands?

ruby -v
ruby -e 'puts Encoding.find("locale")'
ruby -e 'puts $:'

 

[Bek Bek]

I've installed "RubyInstaller 1.9.2-p0" from http://rubyinstaller.org/.
The output I get from the commands you asked me is:

ruvy -v:
ruby 1.9.2p0 (2010-08-18) [i386-mingw32]

ruby -e 'puts Encoding.find("locale")':
IBM737

ruby -e 'puts $:':
C:/Ruby192/lib/ruby/site_ruby/1.9.1
C:/Ruby192/lib/ruby/site_ruby/1.9.1/i386-msvcrt
C:/Ruby192/lib/ruby/site_ruby
C:/Ruby192/lib/ruby/vendor_ruby/1.9.1
C:/Ruby192/lib/ruby/vendor_ruby/1.9.1/i386-msvcrt
C:/Ruby192/lib/ruby/vendor_ruby
C:/Ruby192/lib/ruby/1.9.1
C:/Ruby192/lib/ruby/1.9.1/i386-mingw32

 

[Luis Lavena]

Hello everybody,

I am making my first steps on Ruby on Rails. I have a very simple
question. When I am trying to install rails (as well as every gem) I am
getting this error:

ERROR:  While executing gem ... (Encoding::ConverterNotFoundError)
    code converter not found (UTF-16LE to IBM737)

Why is your terminal IBM737?

Can you change it with chcp to either 850 or 1251/1252?

It seems that IBM737 do not have a Unicode converter match from/to
UTF.

If that works, please let us know.

 

[Bek Bek]

I changed it to 850 and it worked!!
So that was the problem..
Thank you very much!

 

[Brian Candler]

I changed it to 850 and it worked!!
So that was the problem..

Or to give a different point of view: the problem is the fundamentally
flawed design of ruby 1.9, which means that:

(1) The same program, running on the same version of ruby and the same
operating system, will work fine on one system but crash on another.

(2) It is so complex that it is virtually impossible to build complete
test coverage to prevent this sort of problem happening.

Fortunately, there is an easy solution: use ruby 1.8.

 

[Benoit Daloze]

Or to give a different point of view: the problem is the fundamentally
flawed design of ruby 1.9, which means that:

(1) The same program, running on the same version of ruby and the same
operating system, will work fine on one system but crash on another.

(2) It is so complex that it is virtually impossible to build complete
test coverage to prevent this sort of problem happening.

Fortunately, there is an easy solution: use ruby 1.8.
--

Well, that is rather pessimist.

AFAIK, with 1.8, he would just get unreadable characters in his
terminal, and would have to figure himself what to do to resolve it.
With 1.9, at least the error is shown and says the term's encoding is
not compatible.

But I agree, this kind of error just running `gem` is unexpected and
non user-friendly.
But he would anyway have to change one day to use other characters
than pure ASCII in his term.

P.S.: Why `gem` is trying to convert from UTF-16LE ?

 

[Brian Candler]

AFAIK, with 1.8, he would just get unreadable characters in his
terminal

Given that the error messages from ruby are themselves readable, it's a
fair bet that IBM737 is at least ASCII-compatible.

I'm pretty sure that rubygems under 1.8 would just have worked. After
all, rubygems is a system utility. Its job is to fetch files and install
them in the right places. "Encodings" don't help it to do that job, but
they do make programs crash.

P.S.: Why `gem` is trying to convert from UTF-16LE ?

I wondered that too.

But more importantly, why is the locale's default_encoding significant?
Is it reading files from the filesystem without specifying the locale
(or binary mode)? If so that's a "bug" in rubygems, but it's the
language which makes these bugs occur.

 

[Luis Lavena]

Given that the error messages from ruby are themselves readable, it's a
fair bet that IBM737 is at least ASCII-compatible.

This is what Roger Pack asked on Ruby core about IBM737

I wondered that too.

But more importantly, why is the locale's default_encoding significant?
Is it reading files from the filesystem without specifying the locale
(or binary mode)? If so that's a "bug" in rubygems, but it's the
language which makes these bugs occur.

Excellent question, seems to me that is trying to parse some gemspec
contents (my guess)

Complete output of "gem install xxx --debug -V" will be great to have
when chcp is set to 737

 

[Bernard Kenik]

This is what Roger Pack asked on Ruby core about IBM737


Excellent question, seems to me that is trying to parse some gemspec
contents (my guess)

Complete output of "gem install xxx --debug -V" will be great to have
when chcp is set to 737

Pardon if this considered to be high-jacking --
I installed ruby-1.9.2p0 using ruby installer. This is the responses I
get that was asked to the op.


ruby -v:
ruby 1.9.2p0 (2010-08-18) [i386-mingw32]


ruby -e 'puts Encoding.find "local"'
IBM437

ruby -e 'puts $:'
C:/Ruby192/lib/ruby/site_ruby/1.9.1
C:/Ruby192/lib/ruby/site_ruby/1.9.1/i386-msvcrt
C:/Ruby192/lib/ruby/vendor_ruby/1.9.1
C:/Ruby192/lib/ruby/vendor_ruby/1.9.1/i386-msvcrt
C:/Ruby192/lib/ruby/vendor_ruby
C:/Ruby192/lib/ruby/1.9.1
C:/Ruby192/lib/ruby/1.9.1/i386-msvcrt


Is IBM437 an appropriate answer?
If not, how do I correct it?

Thank You

(e-mail address removed)

 

[Luis Lavena]

Pardon if this considered to be high-jacking --
I installed ruby-1.9.2p0 using ruby installer.  This is the responses I
get that was asked to the op.

It fails? if not, then it is thread hijacking

Is IBM437 an appropriate answer?

That depends. I prefer switch to latin (1252) due accented characters/
unicode of Cucumber.

If not, how do I correct it?

A hidden trick, either you change in the current terminal by using
"chcp XYZ" or editing the registry for permanent modification.

http://github.com/aslakhellesoy/cucumber/wiki/Troubleshooting