Fix processing of utf-8 files

Fixes #1517

Author: lsegal

lib/yard/parser/source_parser.rb

@@ -107,7 +107,8 @@ def parse(paths = DEFAULT_PATH_GLOB, excluded = [], level = log.level)
           files = [paths].flatten.
             map {|p| File.directory?(p) ? "#{p}/**/*.{rb,c,cc,cxx,cpp}" : p }.
             map {|p| p.include?("*") ? Dir[p].sort_by {|d| [d.length, d] } : p }.flatten.
-            reject {|p| !File.file?(p) || excluded.any? {|re| p =~ re } }
+            reject {|p| !File.file?(p) || excluded.any? {|re| p =~ re } }.
+            map {|p| p.encoding == Encoding.default_external ? p : p.dup.force_encoding(Encoding.default_external) }
 
           log.enter_level(level) do
             parse_in_order(*files.uniq)

spec/parser/source_parser_spec.rb

@@ -499,6 +499,15 @@ class Foo < Bar
       YARD.parse ['foo']
     end
 
+    it "converts globs into UTF-8" do
+      expect(Dir).to receive(:[]).with('lib/**/*.rb').and_return(['lib/é.rb'])
+      expect(File).to receive(:file?).with('lib/é.rb').and_return(true)
+      expect(File).to receive(:read_binary).with('lib/é.rb').and_return("class A; end")
+
+      YARD.parse ['lib/**/*.rb']
+      expect(Registry.at('A')).not_to be nil
+    end
+
     it "uses Registry.checksums cache if file is cached" do
       data = 'DATA'
       hash = Registry.checksum_for(data)