Why does a file written out encoded as UTF-8 end up being ISO-8859-1 instead? - ruby

Question

I am reading an ISO-8859-1 encoded text file, transcoding it to UTF-8, and writing out a different file as UTF-8. However, when I inspect the output file, it is still encoded as ISO-8859-1! What am I doing wrong?
Here is my ruby class:
module EF
class Transcoder
# app_path ......... Path to the java console application (InferEncoding.jar) that infers the character encoding.
# target_encoding .. Transcodes the text loaded from the file into this encoding.
attr_accessor :app_path, :target_encoding
def initialize(consoleAppPath)
#app_path = consoleAppPath
#target_encoding = "UTF-8"
end
def detect_encoding(filename)
encoding = `java -jar #{#app_path} \"#{filename}\"`
encoding = encoding.strip
end
def transcode(filename)
original_encoding = detect_encoding(filename)
content = File.open(filename, "r:#{original_encoding}", &:read)
content = content.force_encoding(original_encoding)
content.encode!(#target_encoding, :invalid => :replace)
end
def transcode_file(input_filename, output_filename)
content = transcode(input_filename)
File.open(output_filename, "w:#{#target_encoding}") do |f|
f.write content
end
end
end
end
By way of explanation, #app_path is the path to a Java jar file. This console application will read a text file and tell me what its current encoding is (printing it to stdout). It uses the ubiquitous ICU library. (I tried using the ruby gem charlock-holmes, but I cannot get it to compile on Windows for MINGW. The Java bindings to ICU are good, so I wrote a Java application instead.)
To call the above class, I do this in irb:
require_relative 'transcoder'
tc = EF::Transcoder.new("C:/Users/paul.chernoch/Documents/java/InferEncoding.jar")
tc.detect_encoding "C:/temp/infer-encoding-test/ISO-8859-1.txt"
tc.transcode_file("C:/temp/infer-encoding-test/ISO-8859-1.txt", "C:/temp/infer-encoding-test/output-utf8.txt")
tc.detect_encoding "C:/temp/infer-encoding-test/output-utf8.txt"
The file ISO-8859-1.txt is encoded like it sounds. I used Notepad++ to write the file using that encoding.
I used my Java application to test the file. It concurs that it is in ISO-8859-1 format.
I also created a file in Notepad++ and saved it as UTF-8. I then verified using my java app that it was in UTF-8.
After I perform the above in irb, I used my java app to test the output file and it says the format is still ISO-8859-1.
What am I doing wrong? If you hard-code the method detect_encoding to return "ISO-8859-1", you do not need my java application to replicate the part that reads the file.
Any solution must NOT use charlock-holmes.