check_channels.py's exit code reflects every dead channel across the
whole list file it was pointed at, not just the ones a PR added or
changed. Since pr_check.yml ran the checker directly and let its exit
code decide the job's pass/fail, any PR touching a list that already
had pre-existing dead entries (which is most of them - dozens of
already-known-broken channels exist across these lists) got a red X
for something it never caused, even though the workflow's own summary
step already computed the right "touched vs already broken" split.
Fixes this by:
- letting the check step run to completion regardless of its exit
code (the checker's own verdict on the whole file isn't this job's
call to make)
- moving pass/fail into the summarize step, which already has the
touched/others split: the job now fails only when a channel this PR
itself added or changed comes back `dead`
- deliberately not failing on `blocked`/`unreachable`/`flaky` for
touched channels either - these have real false-positive rates from
the runner's own network (geo-blocks, a datacenter IP some providers
reject, one slow probe), which is exactly why the *Channel check
(deep)* workflow exists to give `dead`-looking channels a second
opinion before anyone acts on them
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01Enus8N247jHM5r5Som5JuU
Replaces iptv-checker (npm + ffmpeg) with check_channels.py:
- Checks every channel in a list a PR touches, existing and new, not just
the added/changed rows - a PR editing italy.md now surfaces already-dead
channels in italy.md too, not only the one line it changed.
- Uses check_channels.py's dead/blocked/unreachable/flaky states instead
of a binary online/failed, so the summary no longer needs the "some of
these failures are geo-blocks, use your judgment" disclaimer without
telling reviewers which failures those are - blocked channels are now
named as such.
- Summary separates what the PR actually added/changed from what was
already broken in the same file, so a reviewer isn't left guessing
whether a failure is theirs to fix.
- No ffprobe pass here (kept fast for a PR gate) - anything this flags as
dead gets a second opinion from the scheduled Channel check (deep)
workflow before anyone acts on it.
Verified end to end against a simulated PR diff.
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.
Stores the failed-channel snapshot as a workflow artifact (90-day
retention) and downloads the previous run's artifact for comparison,
instead of committing a state file back to the repo.
ChannelCheckerBot
Kálmán „KAMI” SzalaiandClaude Sonnet 5
KAMI