Scripts now receive the spectrum, channel levels, sample rate and track,
queue and transport details; may return #{ commands, state } to keep state
between frames (with dt and frame, reset on reload and on opening); can draw
thin braille dots and lines (clipped to the area); and can set bold, dim,
italic, underline or reversed on any command. v2 scripts keep working.
The bundled "VFD panel" preset uses all of it for a vacuum-fluorescent hi-fi
front panel in a thin style: segment counter, peak level meter, 25-band
analyzer with peak hold, tone fans, routing diagram and transport lamps,
with a power-on segment test, a counter search on track changes and an
OVER lamp.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A device restarted with a "share" decision that another device of its
group revoked meanwhile sent that card again: the first reconciler pass
called sync_shares from the stale registers before device sync ran.
Until DeviceSync says the registers are current (a sync exchange with a
v4 device of the group after the store was attached, or no such device
to wait for), passes now only unpublish cards the local registers no
longer share and never call sync_shares.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
README, ARCHITECTURE, CHANGELOG and AGENTS describe the immediate
revocation and quit-time wait, decision-bound confirmations, the card
URL filter, devices that left the group, the separate bounded podcast
search, and which direction of a moved feed the search mark covers.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A slow server could hold one of the four refresh slots for 60 s per
redirect hop. A feed or cover fetch now has an overall deadline (120 s).
Tests cover feeds over the size cap with and without Content-Length
(error recorded with back-off, catalog kept), the redirect cap, a
redirect loop, a non-http redirect, a hanging server, and covers (not an
image, over the cap, written atomically). The test server can stall.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The device-sync stream handlers exported and merged podcast snapshot
sections and operations synchronously on a Tokio worker. While a large
feed refresh held the store (up to about half a second), the worker's
other tasks (feed checks, streams, art) waited. Run that work in
block_in_place on the multi-thread runtime, and say so in ARCHITECTURE.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A 301 to a new host followed by a 302 (CDN, tracker) never moved the
feed, because only all-permanent chains counted; every check then went
through the old address without validators and re-applied the whole
feed. The target of the leading permanent hops is now recorded as the
new fetch URL (the old one becomes an alias); the document is still
parsed against the URL that served it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A device below sync v4 got no playback snapshot while the queue held an
episode, so its engine never heard the owner's heartbeat and took
playback over by automatic failover after two minutes; this device then
adopted its queue and the episodes were gone. Send such peers the
coordination envelope with the queue redacted like for Jam: every
episode becomes an opaque "Private item", nothing is stripped, and their
commands stay refused while the queue has episodes.
This departs from FRID PLAYBACK_PROTOCOL.md rule 1 ("send None") and is
reported as a contract gap.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
README explains that podcast state is private by default, syncs only
between trusted devices and streams audio from the publisher, plus
handoff, Jam and older devices. ARCHITECTURE, AGENTS and the changelog
describe protocol v4, version gating, the podcast store bridge and the
invariants (no numeric ids for episodes, versions only from the device
itself).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>