[f4m] Prefer baseURL for relative URLs (closes #14660)

youtube_dl/downloader/f4m.py

@@ -243,8 +243,17 @@ def remove_encrypted_media(media):
                        media))
 
 
-def _add_ns(prop):
+def _add_ns(prop, ver=1):
-    return '{http://ns.adobe.com/f4m/1.0}%s' % prop
+    return '{http://ns.adobe.com/f4m/%d.0}%s' % (ver, prop)
+
+
+def get_base_url(manifest):
+    base_url = xpath_text(
+        manifest, [_add_ns('baseURL'), _add_ns('baseURL', 2)],
+        'base URL', default=None)
+    if base_url:
+        base_url = base_url.strip()
+    return base_url
 
 
 class F4mFD(FragmentFD):
@@ -330,13 +339,13 @@ class F4mFD(FragmentFD):
             rate, media = list(filter(
                 lambda f: int(f[0]) == requested_bitrate, formats))[0]
 
-        base_url = compat_urlparse.urljoin(man_url, media.attrib['url'])
+        # Prefer baseURL for relative URLs as per 11.2 of F4M 3.0 spec.
+        man_base_url = get_base_url(doc) or man_url
+
+        base_url = compat_urlparse.urljoin(man_base_url, media.attrib['url'])
         bootstrap_node = doc.find(_add_ns('bootstrapInfo'))
-        # From Adobe F4M 3.0 spec:
+        boot_info, bootstrap_url = self._parse_bootstrap_node(
-        # The <baseURL> element SHALL be the base URL for all relative
+            bootstrap_node, man_base_url)
-        # (HTTP-based) URLs in the manifest. If <baseURL> is not present, said
-        # URLs should be relative to the location of the containing document.
-        boot_info, bootstrap_url = self._parse_bootstrap_node(bootstrap_node, man_url)
         live = boot_info['live']
         metadata_node = media.find(_add_ns('metadata'))
         if metadata_node is not None:

youtube_dl/extractor/common.py

@@ -29,7 +29,10 @@ from ..compat import (
     compat_urlparse,
     compat_xml_parse_error,
 )
-from ..downloader.f4m import remove_encrypted_media
+from ..downloader.f4m import (
+    get_base_url,
+    remove_encrypted_media,
+)
 from ..utils import (
     NO_DEFAULT,
     age_restricted,
@@ -1239,11 +1242,8 @@ class InfoExtractor(object):
         media_nodes = remove_encrypted_media(media_nodes)
         if not media_nodes:
             return formats
-        base_url = xpath_text(
+
-            manifest, ['{http://ns.adobe.com/f4m/1.0}baseURL', '{http://ns.adobe.com/f4m/2.0}baseURL'],
+        manifest_base_url = get_base_url(manifest)
-            'base URL', default=None)
-        if base_url:
-            base_url = base_url.strip()
 
         bootstrap_info = xpath_element(
             manifest, ['{http://ns.adobe.com/f4m/1.0}bootstrapInfo', '{http://ns.adobe.com/f4m/2.0}bootstrapInfo'],
@@ -1275,7 +1275,7 @@ class InfoExtractor(object):
                     continue
                 manifest_url = (
                     media_url if media_url.startswith('http://') or media_url.startswith('https://')
-                    else ((base_url or '/'.join(manifest_url.split('/')[:-1])) + '/' + media_url))
+                    else ((manifest_base_url or '/'.join(manifest_url.split('/')[:-1])) + '/' + media_url))
                 # If media_url is itself a f4m manifest do the recursive extraction
                 # since bitrates in parent manifest (this one) and media_url manifest
                 # may differ leading to inability to resolve the format by requested