Commit Graph
56 Commits
Author SHA1 Message Date
ekklesiaandClaude Opus 5 a3dba6e8fb feat: provision Chrome and DVD codecs without a manual step
A fresh classroom needed two things installed by hand on every box:
Google Chrome, which is not in Debian, and the DVD codecs, which Debian
cannot ship. Both now happen on their own.

ekklesia-classroom-provision.service runs after every boot and does the
work that needs the network and the apt lock, neither of which a package
postinst can have. Each step returns immediately if it is already done,
so a provisioned machine costs nothing and a machine that was offline
the first time gets another go at the next boot. Nothing there is
allowed to fail the boot: a classroom with no DVD codecs is still a
working classroom, and it says so in the log rather than silently.

The postinst enables the service but deliberately does not start it.
Every OTA update reboots the moment the install finishes, and an apt run
killed by that reboot would leave dpkg half-configured.

Chrome comes from Google's own repository rather than a one-off .deb, so
it keeps getting security updates. Google's package migrates whatever we
write into a deb822 file of its own, so the repository is looked for
across every source file instead of only the one we would write;
otherwise a machine that had Chrome removed would end up with two copies
of the same repository.

The DVD codec helper previously assumed deb822 sources and did nothing
on a machine using a classic /etc/apt/sources.list, which is what the
hardware actually runs. It now handles both formats, and only ever adds
a missing component to an existing Debian entry — so it cannot produce a
duplicate repository or a duplicate component however many times it
runs, and third-party repositories like Chrome's are never touched. It
also answers libdvd-pkg's debconf questions, so nothing blocks on a
prompt in a service with no terminal.

Tested against fixture APT trees for classic sources.list, deb822
.sources, contrib already enabled, repeated runs, third-party
repositories and the awkward real lines: bracketed options, indentation,
trailing comments and flat repositories.

Co-Authored-By: Claude Opus 5 <[email protected]>
2026-09-04 21:33:44 -05:00
ekklesiaandClaude Opus 5 c81fc1b0b9 feat: dress VLC down to an appliance screen
Stock VLC looks like a desktop media player, which is wrong for a room
where the TV is the whole interface. VLC stays the backend because its
dvdnav module is what provides the disc menus, and menus are the part
that matters on a commercial disc.

It is dressed down with documented VLC 3.x options, all verified against
the 3.0.23 build in Debian Trixie: minimal view, dark palette, no cone,
no tray, no now-playing popups, no resume prompt, no recent-disc
history, no title banner. sway draws no border on the window either, so
there is nothing above the video.

The auto-hiding fullscreen controller is deliberately kept: it is not
persistent chrome, and it is the only way a volunteer with a mouse can
pause or change chapter.

No skin is involved. Skins replace the interface wholesale and take the
keyboard and mouse handling that DVD menu navigation depends on with
them, as would --intf dummy.

Also drops --no-qt-updates-notif, which was not an option in Debian's
VLC at all. It made VLC exit before it started, so the DVD button could
never have launched a disc.

Co-Authored-By: Claude Opus 5 <[email protected]>
2026-09-04 21:33:22 -05:00
ekklesiaandClaude Opus 5 428eb27a74 feat: scale the shell to the room's television
A 32" panel a few feet from the children and a 65" set across a hall
want the same design at different sizes. UI_SCALE in device.conf now
sets that, and nobody has to edit CSS or source to tune a room.

The stylesheet and the notification config become templates carrying
measurement tokens. classroom-dock resolves the stylesheet in memory at
startup; classroom-notifications resolves the notification config into
the session's runtime directory and execs mako on it, so the process is
still just mako and everything that looks for it by name still finds it.
Neither template is rewritten on disk, so an update cannot lose a room's
setting and a room cannot drift from the packaged design.

Dock height and the reserved strip, corner radii, padding, button size,
spacing, every icon, the clock and its inset, and the notification
geometry all follow the scale. Colours, translucency and the hairline
borders do not: those are the design, not the size.

At UI_SCALE=1.0 the rendered stylesheet is byte-identical to the one the
shell used before any of this existed, which is the property worth
keeping. Values outside 0.8 to 1.5 are clamped and anything that is not
a number falls back to 1.0, so a typo in device.conf cannot produce an
unusable classroom.

Device configuration moves into a shared module, since the dock and the
notification daemon now have to agree about what a setting means.

Co-Authored-By: Claude Opus 5 <[email protected]>
2026-09-04 21:33:09 -05:00
ekklesiaandClaude Opus 5 ffd18edf7e fix: deliver test notifications into the classroom session
The classroom desktop belongs to the classroom user and mako sits on
that user's D-Bus session. An administrator over SSH is in a different
session where nothing owns org.freedesktop.Notifications, so notify-send
asked D-Bus to start a second mako, which had no display to open and
timed out with NoReply.

Find the running sway process for the classroom user and deliver into
its session instead. XDG_RUNTIME_DIR and the bus address are read from
that process, falling back to /run/user/<uid> and its bus socket when
the environment cannot be read — root has ptrace access on the endpoint,
but that is not somewhere to be brittle. The Wayland display is read the
same way, and failing that from the wayland-N.lock sway holds open, so
no socket name is ever assumed. notify-send only needs the bus, so a
display that cannot be resolved is not worth refusing to send over.

Refuses to send when mako is not already running, rather than letting
D-Bus start one, and says plainly what is wrong when there is no
session, no privileges, or no bus socket. --dry-run reports what it
resolved without sending.

Behaviour inside the session is unchanged, and the sample notifications
and styling are untouched.

The tests stand a real background process in for sway so /proc/<pid>/environ
is genuine rather than mocked, and cover both invocation paths plus the
fallbacks and every refusal. The build runs them.

Co-Authored-By: Claude Opus 5 <[email protected]>
2026-09-04 17:38:44 -05:00
ekklesiaandClaude Opus 5 36100ad40c fix: let restore rescue windows left in sway's scratchpad
Nothing puts windows there any more, but a window parked by an older
classroom build — the exact case this work exists to fix — would
otherwise be invisible to restore and stranded until the next reboot.
Counting the scratchpad as hidden means the Browser button and
Super+Shift+M both pull it back, tiled.

Co-Authored-By: Claude Opus 5 <[email protected]>
2026-09-04 17:13:51 -05:00
ekklesiaandClaude Opus 5 a8cf9ef337 docs: record the classroom test checklist and DVD notes
The checklist covers the bottom-shell regression explicitly, since that
is the part most likely to be broken by a later change, and records the
known limitation that Chrome's own minimize button does nothing under
sway.

Co-Authored-By: Claude Opus 5 <[email protected]>
2026-09-04 17:07:30 -05:00