[ie/youtube] Adjust OAuth refresh token handling (#11414)

Removes support for using '' as an empty password in netrc, e.g.:
machine youtube login oauth password ''

Double-quotes ("") are valid and must be used instead, e.g.:
machine youtube login oauth password ""

Authored by: bashonly

test/test_InfoExtractor.py

@@ -53,6 +53,18 @@ class TestInfoExtractor(unittest.TestCase):
     def test_ie_key(self):
         self.assertEqual(get_info_extractor(YoutubeIE.ie_key()), YoutubeIE)
 
+    def test_get_netrc_login_info(self):
+        for params in [
+            {'usenetrc': True, 'netrc_location': './test/testdata/netrc/netrc'},
+            {'netrc_cmd': f'{sys.executable} ./test/testdata/netrc/print_netrc.py'},
+        ]:
+            ie = DummyIE(FakeYDL(params))
+            self.assertEqual(ie._get_netrc_login_info(netrc_machine='normal_use'), ('user', 'pass'))
+            self.assertEqual(ie._get_netrc_login_info(netrc_machine='empty_user'), ('', 'pass'))
+            self.assertEqual(ie._get_netrc_login_info(netrc_machine='empty_pass'), ('user', ''))
+            self.assertEqual(ie._get_netrc_login_info(netrc_machine='both_empty'), ('', ''))
+            self.assertEqual(ie._get_netrc_login_info(netrc_machine='nonexistent'), (None, None))
+
     def test_html_search_regex(self):
         html = '<p id="foo">Watch this <a href="http://www.youtube.com/watch?v=BaW_jenozKc">video</a></p>'
         search = lambda re, *args: self.ie._html_search_regex(re, html, *args)

test/testdata/netrc/netrc

@@ -0,0 +1,4 @@
+machine normal_use login user password pass
+machine empty_user login "" password pass
+machine empty_pass login user password ""
+machine both_empty login "" password ""

test/testdata/netrc/print_netrc.py

@@ -0,0 +1,2 @@
+with open('./test/testdata/netrc/netrc', encoding='utf-8') as fp:
+    print(fp.read())

yt_dlp/extractor/common.py

@@ -1409,6 +1409,13 @@ class InfoExtractor:
             return None, None
 
         self.write_debug(f'Using netrc for {netrc_machine} authentication')
+
+        # compat: <=py3.10: netrc cannot parse tokens as empty strings, will return `""` instead
+        # Ref: https://github.com/yt-dlp/yt-dlp/issues/11413
+        #      https://github.com/python/cpython/commit/15409c720be0503131713e3d3abc1acd0da07378
+        if sys.version_info < (3, 11):
+            return tuple(x if x != '""' else '' for x in info[::2])
+
         return info[0], info[2]
 
     def _get_login_info(self, username_option='username', password_option='password', netrc_machine=None):

yt_dlp/extractor/youtube.py

@@ -644,13 +644,14 @@ class YoutubeBaseInfoExtractor(InfoExtractor):
         YoutubeBaseInfoExtractor._OAUTH_ACCESS_TOKEN_CACHE[self._OAUTH_PROFILE] = {}
 
         if refresh_token:
-            refresh_token = refresh_token.strip('\'') or None
-
-        # Allow refresh token passed to initialize cache
-        if refresh_token:
+            msg = f'{self._OAUTH_DISPLAY_ID}: Using password input as refresh token'
+            if self.get_param('cachedir') is not False:
+                msg += ' and caching token to disk; you should supply an empty password next time'
+            self.to_screen(msg)
             self.cache.store(self._NETRC_MACHINE, self._oauth_cache_key, refresh_token)
+        else:
+            refresh_token = self.cache.load(self._NETRC_MACHINE, self._oauth_cache_key)
 
-        refresh_token = refresh_token or self.cache.load(self._NETRC_MACHINE, self._oauth_cache_key)
         if refresh_token:
             YoutubeBaseInfoExtractor._OAUTH_ACCESS_TOKEN_CACHE[self._OAUTH_PROFILE]['refresh_token'] = refresh_token
             try: