youtube-dl sometimes extracts the wrong upload playlist for a channel

[ivan]

I spotted this in my logs after getting hundreds of videos from the wrong playlist:

$ youtube-dl https://www.youtube.com/user/talktomeinkorean/videos
[youtube:user] talktomeinkorean: Downloading channel page
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading webpage
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #1
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #2
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #3
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #4
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #5
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #6
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #7
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #8
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #9
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #10
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #11
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #12
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #13
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #14
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #15
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #16
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #17
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #18
[youtube:playlist] UUj2HtBTppiQLVrZfEjcFxig: Downloading page #19
[download] Downloading playlist: Uploads from SevenSuperGirls
[youtube:playlist] playlist Uploads from SevenSuperGirls: Collected 1923 video ids (downloading 1923 of them)
[download] Downloading video 1 of 1923
[youtube] qNmTlDGfX6w: Downloading webpage
[youtube] qNmTlDGfX6w: Extracting video information
[youtube] qNmTlDGfX6w: Searching for annotations.
[youtube] qNmTlDGfX6w: Downloading DASH manifest
[info] Writing video description to: Blast from the Past-qNmTlDGfX6w.description

It extracted the incorrect playlist https://www.youtube.com/playlist?list=UUj2HtBTppiQLVrZfEjcFxig

instead of the correct playlist https://www.youtube.com/playlist?list=UU5r3WHrX4Z7peSYpDlgktGw

Unfortunately, I haven't yet been able to reproduce the issue. Maybe YouTube was serving a temporary promotion or advertisement that matched either regexp?

        channel_playlist_id = self._search_regex(
            [r'<meta itemprop="channelId" content="([^"]+)">',
             r'data-channel-external-id="([^"]+)"'],
            channel_page, 'channel id', default=None)

[dstftw]

Serving promotion as channelId meta looks odd. However, if they render meta differently that may be possible.

[ivan]

I have a patch ready (testing now) that extracts this instead

        channel_playlist_id = self._search_regex(
            r'<link rel="alternate" type="application/rss\+xml" title="RSS"' +
            r' href="https://www\.youtube\.com/feeds/videos\.xml\?channel_id=([^"]+)">',
            channel_page, 'channel id', default=None)

[ivan]

I've stopped patching my fork and given the lack of complaints it's very likely that 3d8e957 fixed it. Closing; will reopen if it ever happens again.