Fixes#1179.
PR #1166 moved Rai 1, Rai 2, Rai 3, Rai Premium, Rai Storia and Rai
Scuola from the official mediapolis.rai.it relinker to a third-party
CloudFront mirror, because the relinker had become Italy-IP-gated and
was failing for everyone else (including the CI checker). That mirror
works for casual viewing, but as #1179 reports, it goes black during
broadcasts where RAI only holds domestic rights (e.g. EuroVolley) -
the official stream keeps working correctly for Italian-IP viewers
through exactly those broadcasts, because it's RAI's own domestic
feed, not a re-mirror of it.
Rather than trade one gap for the other, add the official mediapolis
URL back as a second, Ⓖ-marked entry for each of the six channels -
the same "CDN mirror as the general-access entry, official relinker
as the geo-gated alternative" split this file already uses for every
other Rai channel (Rai 4, Rai 5, Rai Movie, Rai Sport, etc., all still
on mediapolis.rai.it + Ⓖ). Viewers with an Italian IP get the more
correct source; everyone else keeps the mirror that already works for
them.
Also fixes a real gap in check_channels.py found while verifying this:
mediapolis.rai.it's relinker doesn't refuse an out-of-territory request
with an HTTP error - it answers 200 with a real MP4, but the file is
always the same "video_no_available.mp4" decoy regardless of which
channel was asked for. The checker had no way to tell that apart from
a working stream failing its playlist-header check, so every
mediapolis.rai.it entry (the pre-existing Rai 4/5/Movie/Sport ones
included) was scored `dead` instead of `blocked`, right down to the
newly-fixed pr_check.yml workflow that now hard-fails a PR over a
`dead`-scored touched channel. Added a check for this specific decoy
redirect target, classified as `blocked` like any other geo-refusal.
Regenerated playlist.m3u8 and playlists/playlist_italy.m3u8 via
make_playlist.py. pylint check_channels.py: 10.00/10.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Enus8N247jHM5r5Som5JuU
Some radio stations serve the live audio directly at the listed URL
(Content-Type: audio/mpeg, audio/aacp, etc.) rather than an HLS/DASH
manifest that points to one. probe() only checked for #EXTM3U/<MPD in
the response body, so every one of these was misclassified as dead
regardless of whether the station actually worked.
Found while verifying replacement URLs for Irish radio stations: curl
confirmed real audio bytes with correct audio/* Content-Type on all of
them, but check_channels.py reported every single one as dead. Now
checks the response's Content-Type first; a direct audio/video stream
counts as alive without needing playlist syntax. Verified this doesn't
regress playlist detection (existing HLS/DASH channels still classify
correctly).
The first scheduled run of #1157's workflows failed outright: a server
started a chunked HTTP response and then hung up mid-chunk, which
urllib surfaces as http.client.IncompleteRead - a subclass of
http.client.HTTPException, not of OSError/URLError/ValueError, so
probe()'s except clause let it propagate and took the whole run down
before it wrote anything to --json, including check_channels_fast.yml's
run.jsonl.
Catch http.client.HTTPException alongside the existing exceptions and
treat it as unreachable, same as any other broken connection. Verified
against the exact exception (mocked) and against the live lists.
Builds on the idea in #1145 (a checker that maps failures back to a channel
and list, and tells a geo-blocked channel from a dead one) with a few
additions:
check_channels.py
- Standard library only, same as #1145's version.
- Adds a `disputed` state: an optional --confirm-dead pass gives ffprobe a
second opinion on anything that looks dead over HTTP, before it gets
reported as dead. This is one-directional (can only pull a verdict out
of `dead`, never push one into it) because ffprobe itself is not
reliable enough to trust in the other direction - a known-good DASH
channel needed longer than any sane per-channel budget to open while
testing this, and a header check alone had already been fooled by an
isolated media fragment sitting at a URL that looked like a live channel
(a mistake made and caught while triaging #1149/#1151 - see the
docstring for the details).
- Restructured to share one worker pool across every list in a run
instead of a fresh pool per list, which stopped small lists from paying
the same wall-clock floor as large ones; a full run across all lists
dropped from not finishing in 15 minutes to about 10.
- --json writes one record per channel per run, for building a history.
generate_dashboard.py
- Turns that history into a single self-contained docs/index.html: current
state breakdown, an alive-share trend across every run kept, a per-list
breakdown, and a searchable/filterable table of everything that is not
currently alive.
Two new scheduled workflows
- check_channels_fast.yml: every 6 hours, HTTP checks only.
- check_channels_deep.yml: every 2 days, with --confirm-dead (needs ffmpeg).
Both append to .github/checker-history/history.jsonl (pruned to 90 days),
then build and deploy the dashboard to GitHub Pages.
Kálmán „KAMI” Szalai