Start new release.

[youtubedl] / youtube_dl / downloader / f4m.py

diff --git a/youtube_dl/downloader/f4m.py b/youtube_dl/downloader/f4m.py
index 6170cc1552194104ea1f029ec9b54c2b051e7a9c..8dd3c2eeb3154b98e48cf6a58fe50ca50943dada 100644 (file)
--- a/youtube_dl/downloader/f4m.py
+++ b/youtube_dl/downloader/f4m.py
@@ -1,47 +1,57 @@
 from __future__ import division, unicode_literals
 
 from __future__ import division, unicode_literals
 

-import base64

 import io
 import itertools
 import io
 import itertools

-import os

 import time
 
 from .fragment import FragmentFD
 from ..compat import (
 import time
 
 from .fragment import FragmentFD
 from ..compat import (

+    compat_b64decode,

     compat_etree_fromstring,
     compat_urlparse,
     compat_urllib_error,
     compat_urllib_parse_urlparse,
     compat_etree_fromstring,
     compat_urlparse,
     compat_urllib_error,
     compat_urllib_parse_urlparse,

+    compat_struct_pack,
+    compat_struct_unpack,

 )
 from ..utils import (
 )
 from ..utils import (

-    encodeFilename,
-    sanitize_open,
-    struct_pack,
-    struct_unpack,
+    fix_xml_ampersands,

     xpath_text,
 )
 
 
     xpath_text,
 )
 
 

+class DataTruncatedError(Exception):
+    pass
+
+

 class FlvReader(io.BytesIO):
     """
     Reader for Flv files
     The file format is documented in https://www.adobe.com/devnet/f4v.html
     """
 
 class FlvReader(io.BytesIO):
     """
     Reader for Flv files
     The file format is documented in https://www.adobe.com/devnet/f4v.html
     """
 

+    def read_bytes(self, n):
+        data = self.read(n)
+        if len(data) < n:
+            raise DataTruncatedError(
+                'FlvReader error: need %d bytes while only %d bytes got' % (
+                    n, len(data)))
+        return data
+

     # Utility functions for reading numbers and strings
     def read_unsigned_long_long(self):
     # Utility functions for reading numbers and strings
     def read_unsigned_long_long(self):

-        return struct_unpack('!Q', self.read(8))[0]
+        return compat_struct_unpack('!Q', self.read_bytes(8))[0]

 
     def read_unsigned_int(self):
 
     def read_unsigned_int(self):

-        return struct_unpack('!I', self.read(4))[0]
+        return compat_struct_unpack('!I', self.read_bytes(4))[0]

 
     def read_unsigned_char(self):
 
     def read_unsigned_char(self):

-        return struct_unpack('!B', self.read(1))[0]
+        return compat_struct_unpack('!B', self.read_bytes(1))[0]

 
     def read_string(self):
         res = b''
         while True:
 
     def read_string(self):
         res = b''
         while True:

-            char = self.read(1)
+            char = self.read_bytes(1)

             if char == b'\x00':
                 break
             res += char
             if char == b'\x00':
                 break
             res += char


@@ -52,18 +62,18 @@ class FlvReader(io.BytesIO):
         Read a box and return the info as a tuple: (box_size, box_type, box_data)
         """
         real_size = size = self.read_unsigned_int()
         Read a box and return the info as a tuple: (box_size, box_type, box_data)
         """
         real_size = size = self.read_unsigned_int()

-        box_type = self.read(4)
+        box_type = self.read_bytes(4)

         header_end = 8
         if size == 1:
             real_size = self.read_unsigned_long_long()
             header_end = 16
         header_end = 8
         if size == 1:
             real_size = self.read_unsigned_long_long()
             header_end = 16

-        return real_size, box_type, self.read(real_size - header_end)
+        return real_size, box_type, self.read_bytes(real_size - header_end)

 
     def read_asrt(self):
         # version
         self.read_unsigned_char()
         # flags
 
     def read_asrt(self):
         # version
         self.read_unsigned_char()
         # flags

-        self.read(3)
+        self.read_bytes(3)

         quality_entry_count = self.read_unsigned_char()
         # QualityEntryCount
         for i in range(quality_entry_count):
         quality_entry_count = self.read_unsigned_char()
         # QualityEntryCount
         for i in range(quality_entry_count):


@@ -84,7 +94,7 @@ class FlvReader(io.BytesIO):
         # version
         self.read_unsigned_char()
         # flags
         # version
         self.read_unsigned_char()
         # flags

-        self.read(3)
+        self.read_bytes(3)

         # time scale
         self.read_unsigned_int()
 
         # time scale
         self.read_unsigned_int()
 


@@ -118,7 +128,7 @@ class FlvReader(io.BytesIO):
         # version
         self.read_unsigned_char()
         # flags
         # version
         self.read_unsigned_char()
         # flags

-        self.read(3)
+        self.read_bytes(3)

 
         self.read_unsigned_int()  # BootstrapinfoVersion
         # Profile,Live,Update,Reserved
 
         self.read_unsigned_int()  # BootstrapinfoVersion
         # Profile,Live,Update,Reserved


@@ -183,6 +193,11 @@ def build_fragments_list(boot_info):
     first_frag_number = fragment_run_entry_table[0]['first']
     fragments_counter = itertools.count(first_frag_number)
     for segment, fragments_count in segment_run_table['segment_run']:
     first_frag_number = fragment_run_entry_table[0]['first']
     fragments_counter = itertools.count(first_frag_number)
     for segment, fragments_count in segment_run_table['segment_run']:

+        # In some live HDS streams (for example Rai), `fragments_count` is
+        # abnormal and causing out-of-memory errors. It's OK to change the
+        # number of fragments for live streams as they are updated periodically
+        if fragments_count == 4294967295 and boot_info['live']:
+            fragments_count = 2

         for _ in range(fragments_count):
             res.append((segment, next(fragments_counter)))
 
         for _ in range(fragments_count):
             res.append((segment, next(fragments_counter)))
 


@@ -193,11 +208,11 @@ def build_fragments_list(boot_info):
 
 
 def write_unsigned_int(stream, val):
 
 
 def write_unsigned_int(stream, val):

-    stream.write(struct_pack('!I', val))
+    stream.write(compat_struct_pack('!I', val))

 
 
 def write_unsigned_int_24(stream, val):
 
 
 def write_unsigned_int_24(stream, val):

-    stream.write(struct_pack('!I', val)[1:])
+    stream.write(compat_struct_pack('!I', val)[1:])

 
 
 def write_flv_header(stream):
 
 
 def write_flv_header(stream):


@@ -222,8 +237,23 @@ def write_metadata_tag(stream, metadata):
         write_unsigned_int(stream, FLV_TAG_HEADER_LEN + len(metadata))
 
 
         write_unsigned_int(stream, FLV_TAG_HEADER_LEN + len(metadata))
 
 

-def _add_ns(prop):
-    return '{http://ns.adobe.com/f4m/1.0}%s' % prop
+def remove_encrypted_media(media):
+    return list(filter(lambda e: 'drmAdditionalHeaderId' not in e.attrib
+                                 and 'drmAdditionalHeaderSetId' not in e.attrib,
+                       media))
+
+
+def _add_ns(prop, ver=1):
+    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):
 
 
 class F4mFD(FragmentFD):


@@ -237,15 +267,13 @@ class F4mFD(FragmentFD):
         media = doc.findall(_add_ns('media'))
         if not media:
             self.report_error('No media found')
         media = doc.findall(_add_ns('media'))
         if not media:
             self.report_error('No media found')

-        for e in (doc.findall(_add_ns('drmAdditionalHeader')) +
-                  doc.findall(_add_ns('drmAdditionalHeaderSet'))):
+        for e in (doc.findall(_add_ns('drmAdditionalHeader'))
+                  + doc.findall(_add_ns('drmAdditionalHeaderSet'))):

             # If id attribute is missing it's valid for all media nodes
             # without drmAdditionalHeaderId or drmAdditionalHeaderSetId attribute
             if 'id' not in e.attrib:
                 self.report_error('Missing ID in f4m DRM')
             # If id attribute is missing it's valid for all media nodes
             # without drmAdditionalHeaderId or drmAdditionalHeaderSetId attribute
             if 'id' not in e.attrib:
                 self.report_error('Missing ID in f4m DRM')

-        media = list(filter(lambda e: 'drmAdditionalHeaderId' not in e.attrib and
-                                      'drmAdditionalHeaderSetId' not in e.attrib,
-                            media))
+        media = remove_encrypted_media(media)

         if not media:
             self.report_error('Unsupported DRM')
         return media
         if not media:
             self.report_error('Unsupported DRM')
         return media


@@ -272,28 +300,38 @@ class F4mFD(FragmentFD):
         return fragments_list
 
     def _parse_bootstrap_node(self, node, base_url):
         return fragments_list
 
     def _parse_bootstrap_node(self, node, base_url):

-        if node.text is None:
+        # Sometimes non empty inline bootstrap info can be specified along
+        # with bootstrap url attribute (e.g. dummy inline bootstrap info
+        # contains whitespace characters in [1]). We will prefer bootstrap
+        # url over inline bootstrap info when present.
+        # 1. http://live-1-1.rutube.ru/stream/1024/HDS/SD/C2NKsS85HQNckgn5HdEmOQ/1454167650/S-s604419906/move/four/dirs/upper/1024-576p.f4m
+        bootstrap_url = node.get('url')
+        if bootstrap_url:

             bootstrap_url = compat_urlparse.urljoin(
             bootstrap_url = compat_urlparse.urljoin(

-                base_url, node.attrib['url'])
+                base_url, bootstrap_url)

             boot_info = self._get_bootstrap_from_url(bootstrap_url)
         else:
             bootstrap_url = None
             boot_info = self._get_bootstrap_from_url(bootstrap_url)
         else:
             bootstrap_url = None

-            bootstrap = base64.b64decode(node.text.encode('ascii'))
+            bootstrap = compat_b64decode(node.text)

             boot_info = read_bootstrap_info(bootstrap)
             boot_info = read_bootstrap_info(bootstrap)

-        return (boot_info, bootstrap_url)
+        return boot_info, bootstrap_url

 
     def real_download(self, filename, info_dict):
         man_url = info_dict['url']
         requested_bitrate = info_dict.get('tbr')
         self.to_screen('[%s] Downloading f4m manifest' % self.FD_NAME)
 
     def real_download(self, filename, info_dict):
         man_url = info_dict['url']
         requested_bitrate = info_dict.get('tbr')
         self.to_screen('[%s] Downloading f4m manifest' % self.FD_NAME)

-        urlh = self.ydl.urlopen(man_url)
+
+        urlh = self.ydl.urlopen(self._prepare_url(info_dict, man_url))

         man_url = urlh.geturl()
         man_url = urlh.geturl()

-        manifest = urlh.read()
+        # Some manifests may be malformed, e.g. prosiebensat1 generated manifests
+        # (see https://github.com/ytdl-org/youtube-dl/issues/6215#issuecomment-121704244
+        # and https://github.com/ytdl-org/youtube-dl/issues/7823)
+        manifest = fix_xml_ampersands(urlh.read().decode('utf-8', 'ignore')).strip()

 
         doc = compat_etree_fromstring(manifest)
         formats = [(int(f.attrib.get('bitrate', -1)), f)
                    for f in self._get_unencrypted_media(doc)]
 
         doc = compat_etree_fromstring(manifest)
         formats = [(int(f.attrib.get('bitrate', -1)), f)
                    for f in self._get_unencrypted_media(doc)]

-        if requested_bitrate is None:
+        if requested_bitrate is None or len(formats) == 1:

             # get the best format
             formats = sorted(formats, key=lambda f: f[0])
             rate, media = formats[-1]
             # get the best format
             formats = sorted(formats, key=lambda f: f[0])
             rate, media = formats[-1]


@@ -301,18 +339,23 @@ class F4mFD(FragmentFD):
             rate, media = list(filter(
                 lambda f: int(f[0]) == requested_bitrate, formats))[0]
 
             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'))
         bootstrap_node = doc.find(_add_ns('bootstrapInfo'))

-        boot_info, bootstrap_url = self._parse_bootstrap_node(bootstrap_node, base_url)
+        boot_info, bootstrap_url = self._parse_bootstrap_node(
+            bootstrap_node, man_base_url)

         live = boot_info['live']
         metadata_node = media.find(_add_ns('metadata'))
         if metadata_node is not None:
         live = boot_info['live']
         metadata_node = media.find(_add_ns('metadata'))
         if metadata_node is not None:

-            metadata = base64.b64decode(metadata_node.text.encode('ascii'))
+            metadata = compat_b64decode(metadata_node.text)

         else:
             metadata = None
 
         fragments_list = build_fragments_list(boot_info)
         else:
             metadata = None
 
         fragments_list = build_fragments_list(boot_info)

-        if self.params.get('test', False):
+        test = self.params.get('test', False)
+        if test:

             # We only download the first fragment
             fragments_list = fragments_list[:1]
         total_frags = len(fragments_list)
             # We only download the first fragment
             fragments_list = fragments_list[:1]
         total_frags = len(fragments_list)


@@ -322,23 +365,28 @@ class F4mFD(FragmentFD):
         ctx = {
             'filename': filename,
             'total_frags': total_frags,
         ctx = {
             'filename': filename,
             'total_frags': total_frags,

+            'live': live,

         }
 
         self._prepare_frag_download(ctx)
 
         dest_stream = ctx['dest_stream']
 
         }
 
         self._prepare_frag_download(ctx)
 
         dest_stream = ctx['dest_stream']
 

-        write_flv_header(dest_stream)
-        if not live:
-            write_metadata_tag(dest_stream, metadata)
+        if ctx['complete_frags_downloaded_bytes'] == 0:
+            write_flv_header(dest_stream)
+            if not live:
+                write_metadata_tag(dest_stream, metadata)

 
         base_url_parsed = compat_urllib_parse_urlparse(base_url)
 
         self._start_frag_download(ctx)
 
 
         base_url_parsed = compat_urllib_parse_urlparse(base_url)
 
         self._start_frag_download(ctx)
 

-        frags_filenames = []
+        frag_index = 0

         while fragments_list:
             seg_i, frag_i = fragments_list.pop(0)
         while fragments_list:
             seg_i, frag_i = fragments_list.pop(0)

+            frag_index += 1
+            if frag_index <= ctx['fragment_index']:
+                continue

             name = 'Seg%d-Frag%d' % (seg_i, frag_i)
             query = []
             if base_url_parsed.query:
             name = 'Seg%d-Frag%d' % (seg_i, frag_i)
             query = []
             if base_url_parsed.query:


@@ -348,24 +396,26 @@ class F4mFD(FragmentFD):
             if info_dict.get('extra_param_to_segment_url'):
                 query.append(info_dict['extra_param_to_segment_url'])
             url_parsed = base_url_parsed._replace(path=base_url_parsed.path + name, query='&'.join(query))
             if info_dict.get('extra_param_to_segment_url'):
                 query.append(info_dict['extra_param_to_segment_url'])
             url_parsed = base_url_parsed._replace(path=base_url_parsed.path + name, query='&'.join(query))

-            frag_filename = '%s-%s' % (ctx['tmpfilename'], name)

             try:
             try:

-                success = ctx['dl'].download(frag_filename, {'url': url_parsed.geturl()})
+                success, down_data = self._download_fragment(ctx, url_parsed.geturl(), info_dict)

                 if not success:
                     return False
                 if not success:
                     return False

-                (down, frag_sanitized) = sanitize_open(frag_filename, 'rb')
-                down_data = down.read()
-                down.close()

                 reader = FlvReader(down_data)
                 while True:
                 reader = FlvReader(down_data)
                 while True:

-                    _, box_type, box_data = reader.read_box_info()
+                    try:
+                        _, box_type, box_data = reader.read_box_info()
+                    except DataTruncatedError:
+                        if test:
+                            # In tests, segments may be truncated, and thus
+                            # FlvReader may not be able to parse the whole
+                            # chunk. If so, write the segment as is
+                            # See https://github.com/ytdl-org/youtube-dl/issues/9214
+                            dest_stream.write(down_data)
+                            break
+                        raise

                     if box_type == b'mdat':
                     if box_type == b'mdat':

-                        dest_stream.write(box_data)
+                        self._append_fragment(ctx, box_data)

                         break
                         break

-                if live:
-                    os.remove(encodeFilename(frag_sanitized))
-                else:
-                    frags_filenames.append(frag_sanitized)

             except (compat_urllib_error.HTTPError, ) as err:
                 if live and (err.code == 404 or err.code == 410):
                     # We didn't keep up with the live window. Continue
             except (compat_urllib_error.HTTPError, ) as err:
                 if live and (err.code == 404 or err.code == 410):
                     # We didn't keep up with the live window. Continue


@@ -376,7 +426,7 @@ class F4mFD(FragmentFD):
                 else:
                     raise
 
                 else:
                     raise
 

-            if not fragments_list and live and bootstrap_url:
+            if not fragments_list and not test and live and bootstrap_url:

                 fragments_list = self._update_live_fragments(bootstrap_url, frag_i)
                 total_frags += len(fragments_list)
                 if fragments_list and (fragments_list[0][1] > frag_i + 1):
                 fragments_list = self._update_live_fragments(bootstrap_url, frag_i)
                 total_frags += len(fragments_list)
                 if fragments_list and (fragments_list[0][1] > frag_i + 1):


@@ -385,7 +435,4 @@ class F4mFD(FragmentFD):
 
         self._finish_frag_download(ctx)
 
 
         self._finish_frag_download(ctx)
 

-        for frag_file in frags_filenames:
-            os.remove(encodeFilename(frag_file))
-

         return True
         return True