Fix watcher failure due to incorrect file name encoding

[stkao05]

When the watched file name isn't US-ASCII encoded (i.e. UTF-8), the watcher will raise error when file modification is made:

processing events: U+6211 from UTF-8 to US-ASCII Backtrace:

[2018-10-13T15:17:43.736524 #83757] ERROR -- : exception while processing events: U+6211 from UTF-8 to US-ASCII Backtrace:
 -- .rbenv/versions/2.5.1/lib/ruby/gems/2.5.0/gems/jekyll-watch-2.1.1/lib/jekyll/watcher.rb:71:in `encode!'
 -- .rbenv/versions/2.5.1/lib/ruby/gems/2.5.0/gems/jekyll-watch-2.1.1/lib/jekyll/watcher.rb:71:in `block in normalize_encoding'
 -- .rbenv/versions/2.5.1/lib/ruby/gems/2.5.0/gems/jekyll-watch-2.1.1/lib/jekyll/watcher.rb:71:in `map'
 -- .rbenv/versions/2.5.1/lib/ruby/gems/2.5.0/gems/jekyll-watch-2.1.1/lib/jekyll/watcher.rb:71:in `normalize_encoding'
 -- .rbenv/versions/2.5.1/lib/ruby/gems/2.5.0/gems/jekyll-watch-2.1.1/lib/jekyll/watcher.rb:59:in `block in listen_handler'

The issue is fixed by avoiding the need for encoding conversion when watcher outputting log to the console.

Steps to reproduce

• Create a post with file name "2018-01-01-夏天.markdown"
• Run jekyll (bundle exec jekyll serve)
• Make any modification to the post file to trigger file change event, and error will be raised.

[ashmaroli]

This still breaks watching on Windows and source at D:/testé-çasé

Configuration file: D:/testé-çasé/_config.yml
            Source: D:/testé-çasé
       Destination: public
 Incremental build: disabled. Enable with --incremental
      Generating...
                    done in 0.749 seconds.
 Auto-regeneration: enabled for 'D:/testé-çasé'
    Server address: http://127.0.0.1:4000
  Server running... press ctrl-c to stop.
      Regenerating: 1 file(s) changed at 2018-10-13 05:27:29
E, [2018-10-13T05:27:29.792599 #11068] ERROR -- : exception while processing events: incompatible character encodings: UTF-8 and Windows-1252

[DirtyF]

@ashmaroli this PR focuses on files inside the source dir, fine by me, but we need to add a test.

[ashmaroli]

@DirtyF I agree this fix is legit for watching 2018-01-01-夏天.markdown inside a "latin-named" source_dir, and therefore should definitely be merged..
Just pointing out that this "breaks" a "non-latin-named" source_dir and a "latin-named regular" dest_dir on Windows..

[DirtyF]

Would be nice to have a test checkin' this, especially if we want Jekyll 4 to properly handle i18n.

[stkao05]

Having thought about it a bit, I think the fix I made previously was wrong.

The issue codes was outputting file path info to the console, and therefore, it is the encoding of the file name that we are working with, not the encoding of file content. The previous fix incorrectly encodes the file name encoding with the content encoding specified in the Jekyll config.

I change my commit to fix this issue with a different approach which avoids the need for encoding conversion altogether. I have also added a new post file with Chinese file name for test purpose.

@ashmaroli Could you check if the fix works on your machine? Thanks!

[ashmaroli]

@stkao05 You have effectively replaced the fix introduced in #69 with a better alternative. The issue I noted in #78 (comment) is still present but that shouldn't be a concern..

Great Job!! 👍 💯

[ashmaroli]

👍 For a better alternative

[ashmaroli]

@DirtyF I'm dismissing your "Approval" and requesting a re-review of the "current implementation"..

Branch content changed significantly since approval.

[DirtyF]

Works on my end.

[DirtyF]

@jekyllbot: merge +fix