Paperclip content type validation uses wrong encoding in file command on jruby

[mmustala]

I have recently moved my project from MRI ruby to JRuby. At that time I needed to change my backticks calls to encode the ruby strings into ascii-8bit when passing it into the call to ensure that the correct file is found for example.

`file -b --mime-type #{path.encode('ASCII-8BIT').shellescape}`.try(:strip)

Now I noticed that when Paperclip 4.2.0 (and earlier versions) tries to validate the content type of an attachment whose file name contains utf-8 characters like Öäå the file name goes in with a wrong encoding. The error message is

Command :: file -b --mime '/tmp/ea_Neliö20141119-12290-a5imen'
[paperclip] Error while determining content type: Command 'file -b --mime '/tmp/ea_Neliö20141119-12290-a5imen'' returned 1. Expected 0
Here is the command output:
ERROR: cannot open `/tmp/ea_Neliö20141119-12290-a5imen' (No such file or directory)

I'm running the application on Ubuntu in development and Debian in production.

I would love to offer a patch for this but finding the correct line to fix is not so easy. The way validations are working in Rails seems to be magic.

[mmustala]

lib/paperclip/file_command_content_type_detector.rb calls

Paperclip.run("file", "-b --mime :file", :file => @filename)

which instead calls

Cocaine::CommandLine.new(cmd, arguments, local_options).run(interpolation_values)

This probably is best to fix in Cocaine #run method where the interpolation_values hash values should be encoded if using jruby and unix os.

[tute]

Thanks @mmustala! Thanks for your input and for reporting this issue to the other projects.