Unicode file packager

[kripken]

Adds a test for #4100.

Passes on linux, since it looks like the error reported in that issue happens in windows-specific code. The code checks for the 'hidden' attribute. @juj, can we just ignore unicode paths in that method perhaps?

[kripken]

Also added a commit that would ignore errors in that method, which seems ok since it just checks for a path being hidden - if we can't check, we might as well continue normally? Or is there a better fix, @juj?

[TannerRogalsky]

Thanks for the quick response, @kripken! My inclination to fix that line was to call decode('unicode-escape') on filepath before passing it to unicode. Seemed to resolve that issue.

The reason I didn't submit a PR with that, though, was that another error occurred after that when generating the json metadata (on this line). Could we also pass a --js-output=test.data.js arg to the file packager in the test and then make that pass?

[juj]

Looks good to me. Tested to pass on Windows as well.

[kripken]

@TannerRogalsky why is that extra flag needed? It just makes it write to a file, not to stdout, and both cases should hit the line you mention with metadata?

[TannerRogalsky]

@kripken You're right. The test passes for me on Windows as well but I still get errors when trying to run the script from various CLIs. This is good for me, though. If someone wants to figure out the precise nature of the issue surrounding unicode encoding of file names in Python 2 in Windows, I will direct them to this starting point.

Thank you!

[kripken]

A possible cause for the failure you still see, @TannerRogalsky, is if not just the dir name but also a filename contains unicode? Worth testing that on windows to see what happens. (On linux it already works here.)