X-Git-Url: http://git.bitcoin.ninja/index.cgi?a=blobdiff_plain;f=youtube_dl%2Fextractor%2Ftwitch.py;h=8639293e35a79875c40b3933b5531381c2b3d077;hb=12557339453e25dbb18dfc51dc1e88ca5325d8e9;hp=69882da6337bc6f20d0578f6d612e799429db3e8;hpb=c3d3110f6a4b769e1ddb5532ac61f3da419ebd07;p=youtube-dl

diff --git a/youtube_dl/extractor/twitch.py b/youtube_dl/extractor/twitch.py
index 69882da63..8639293e3 100644
--- a/youtube_dl/extractor/twitch.py
+++ b/youtube_dl/extractor/twitch.py
@@ -17,6 +17,7 @@ from ..utils import (
     encode_dict,
     ExtractorError,
     int_or_none,
+    orderedSet,
     parse_duration,
     parse_iso8601,
     sanitized_Request,
@@ -281,17 +282,36 @@ class TwitchPlaylistBaseIE(TwitchBaseIE):
         entries = []
         offset = 0
         limit = self._PAGE_LIMIT
+        broken_paging_detected = False
+        counter_override = None
         for counter in itertools.count(1):
             response = self._download_json(
                 self._PLAYLIST_URL % (channel_id, offset, limit),
-                channel_id, 'Downloading %s videos JSON page %d' % (self._PLAYLIST_TYPE, counter))
+                channel_id,
+                'Downloading %s videos JSON page %s'
+                % (self._PLAYLIST_TYPE, counter_override or counter))
             page_entries = self._extract_playlist_page(response)
             if not page_entries:
                 break
+            total = int_or_none(response.get('_total'))
+            # Since the beginning of March 2016 twitch's paging mechanism
+            # is completely broken on the twitch side. It simply ignores
+            # a limit and returns the whole offset number of videos.
+            # Working around by just requesting all videos at once.
+            if not broken_paging_detected and total and len(page_entries) > limit:
+                self.report_warning(
+                    'Twitch paging is broken on twitch side, requesting all videos at once',
+                    channel_id)
+                broken_paging_detected = True
+                offset = total
+                counter_override = '(all at once)'
+                continue
             entries.extend(page_entries)
+            if broken_paging_detected or total and len(page_entries) >= total:
+                break
             offset += limit
         return self.playlist_result(
-            [self.url_result(entry) for entry in set(entries)],
+            [self.url_result(entry) for entry in orderedSet(entries)],
             channel_id, channel_name)
 
     def _extract_playlist_page(self, response):