Fix for unicode filenames

Fix for the issue discussed here:
https://github.com/rg3/youtube-dl/issues/4787
https://github.com/rg3/youtube-dl/issues/4787 by hashing the input and
output filenames before processing them and then restoring them.

youtube_dl/YoutubeDL.py

@@ -570,8 +570,8 @@ class YoutubeDL(object):
             # Temporary fix for #4787
             # 'Treat' all problem characters by passing filename through preferredencoding
             # to workaround encoding issues with subprocess on python2 @ Windows
-            if sys.version_info < (3, 0) and sys.platform == 'win32':
-                filename = encodeFilename(filename, True).decode(preferredencoding())
+            #if sys.version_info < (3, 0) and sys.platform == 'win32':
+            #    filename = encodeFilename(filename, True).decode(preferredencoding())
             return filename
         except ValueError as err:
             self.report_error('Error in output template: ' + str(err) + ' (encoding: ' + repr(preferredencoding()) + ')')

youtube_dl/postprocessor/ffmpeg.py

@@ -4,7 +4,7 @@ import io
 import os
 import subprocess
 import time
-
+import hashNames
 
 from .common import AudioConversionError, PostProcessor
 
@@ -131,12 +131,17 @@ class FFmpegPostProcessor(PostProcessor):
             os.stat(encodeFilename(path)).st_mtime for path in input_paths)
 
         files_cmd = []
+        original_names = []
         for path in input_paths:
-            files_cmd.extend([encodeArgument('-i'), encodeFilename(path, True)])
+            hashed_path = hashNames.hashRename(path)
+            os.rename(path, hashed_path)
+            original_names.append([hashed_path, path])
+            files_cmd.extend([encodeArgument('-i'), encodeFilename(hashed_path, True)])
+        hashed_out_path = hashNames.hashRename(out_path)
         cmd = ([encodeFilename(self.executable, True), encodeArgument('-y')] +
                files_cmd +
                [encodeArgument(o) for o in opts] +
-               [encodeFilename(self._ffmpeg_filename_argument(out_path), True)])
+               [encodeFilename(self._ffmpeg_filename_argument(hashed_out_path), True)])
 
         if self._downloader.params.get('verbose', False):
             self._downloader.to_screen('[debug] ffmpeg command line: %s' % shell_quote(cmd))
@@ -146,6 +151,9 @@ class FFmpegPostProcessor(PostProcessor):
             stderr = stderr.decode('utf-8', 'replace')
             msg = stderr.strip().split('\n')[-1]
             raise FFmpegPostProcessorError(msg)
+        for o_name in original_names:
+            os.rename(o_name[0], o_name[1])
+        os.rename(hashed_out_path, out_path)
         self.try_utime(out_path, oldest_mtime, oldest_mtime)
 
     def run_ffmpeg(self, path, out_path, opts):

youtube_dl/postprocessor/hashNames.py

@@ -0,0 +1,9 @@
+import os
+import hashlib
+
+#Hash the file names to drop unicode characters for FFMPEG
+def hashRename(fileN):
+    ext = os.path.splitext(fileN)
+    fileHash = hashlib.sha224(fileN.encode('utf-8')).hexdigest()
+    hash_name = fileHash+ext[1]
+    return hash_name