Adds a test case for decoding UTF-8 strings and fixes this bug

[keichan34]

Multibyte strings were not being encoded back into UTF-8, so I wrote a test case (and fixed the bug)

[jqr]

I'll need to look into this a bit more but this looks like a great change.

[Freaky]

PHP serialization format doesn't do encodings - similar to Ruby 1.8-era, they're just vague blobs of bytes. UTF-8 is surely a good bet these days, but maybe it should be configurable?

Module instance variable I guess.

[keichan34]

@jqr @Freaky Used an module variable to set encoding. You think this should be documented somewhere?

[Freaky]

Well, the point would be to make it easy to override if necessary, a constant would throw a warning if you redefined it. I was thinking more along the lines of:

module PHP
  class << self
    attr_accessor :encoding
  end
end

PHP.encoding ='UTF-8'

Then users can set it to whatever weirdness they're using in PHP. Might be mostly theoretical?

[kaorukobo]

Today, I tried to merge this PR and #11 onto commit e23fcfd. I noticed that these two patches are incompatible with each other.

#11 resolves the issue using a different approach, as keichan34 mentioned. Thus, I think this PR can be closed in favor of #11.

[jqr]

Yeah that force encoding seems problematic, closing in favor of #11.