Merge 9035c78016 into be3579aaf0

The byte string-format should be applied after the sanitization is done,
as sanitize might replace a single byte character with a multi-byte one,
e.g. '/' with '⧸', making the resulting string go over the desired byte
limit.

Fixes #10060

test/test_YoutubeDL.py

@@ -712,13 +712,13 @@ class TestYoutubeDL(unittest.TestCase):
     }
 
     def test_prepare_outtmpl_and_filename(self):
-        def test(tmpl, expected, *, info=None, **params):
+        def test(tmpl, expected, *, info=None, sanitize=False, **params):
             params['outtmpl'] = tmpl
             ydl = FakeYDL(params)
             ydl._num_downloads = 1
             self.assertEqual(ydl.validate_outtmpl(tmpl), None)
 
-            out = ydl.evaluate_outtmpl(tmpl, info or self.outtmpl_info)
+            out = ydl.evaluate_outtmpl(tmpl, info or self.outtmpl_info, sanitize=sanitize)
             fname = ydl.prepare_filename(info or self.outtmpl_info)
 
             if not isinstance(expected, (list, tuple)):
@@ -917,6 +917,7 @@ class TestYoutubeDL(unittest.TestCase):
         test('Hello %(title2)s', 'Hello %PATH%')
         test('%(title3)s', ('foo/bar\\test', 'foo⧸bar⧹test'))
         test('folder/%(title3)s', ('folder/foo/bar\\test', f'folder{os.path.sep}foo⧸bar⧹test'))
+        test('%(title3).7B', 'foo⧸b', sanitize=True)
 
     def test_format_note(self):
         ydl = YoutubeDL()

yt_dlp/YoutubeDL.py

@@ -1374,9 +1374,6 @@ class YoutubeDL:
             elif fmt[-1] == 'q':  # quoted
                 value = map(str, variadic(value) if '#' in flags else [value])
                 value, fmt = shell_quote(value, shell=True), str_fmt
-            elif fmt[-1] == 'B':  # bytes
-                value = f'%{str_fmt}'.encode() % str(value).encode()
-                value, fmt = value.decode('utf-8', 'ignore'), 's'
             elif fmt[-1] == 'U':  # unicode normalized
                 value, fmt = unicodedata.normalize(
                     # "+" = compatibility equivalence, "#" = NFD
@@ -1393,7 +1390,7 @@ class YoutubeDL:
                     value = str(value)[0]
                 else:
                     fmt = str_fmt
-            elif fmt[-1] not in 'rsa':  # numeric
+            elif fmt[-1] not in 'rsaB':  # numeric
                 value = float_or_none(value)
                 if value is None:
                     value, fmt = default, 's'
@@ -1405,9 +1402,13 @@ class YoutubeDL:
                     value, fmt = repr(value), str_fmt
                 elif fmt[-1] == 'a':
                     value, fmt = ascii(value), str_fmt
-                if fmt[-1] in 'csra':
+                if fmt[-1] in 'csraB':
                     value = sanitizer(last_field, value)
 
+            if fmt[-1] == 'B':  # bytes
+                value = f'%{str_fmt}'.encode() % str(value).encode()
+                value, fmt = value.decode('utf-8', 'ignore'), 's'
+
             key = '{}\0{}'.format(key.replace('%', '%\0'), outer_mobj.group('format'))
             TMPL_DICT[key] = value
             return '{prefix}%({key}){fmt}'.format(key=key, fmt=fmt, prefix=outer_mobj.group('prefix'))