`engines`, `@types/node` and the three esbuild targets said node20 while both
stages of the image have been node:24-alpine for a while — and since
24.10.2026 the development machines run 24 too. So the bundles are built for
exactly the runtime that carries them instead of for a version nobody runs
any more.
`npm run check`: 110 files, 1825 tests, green on node 24.22.1.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
test-filius refused every ticket as from the future: its clock was 886 ms
behind the dashboard's machine, with NTP still 8 ms (one ticket in 125, as a
ticket carries whole seconds). F now takes a ticket signed up to 5 s ahead
(TICKET_AHEAD_SECONDS); itsdangerous refuses one second, and the oracle
keeps that on record.
tools/live measures and accepts on a live F, with fresh projects each run:
m3 (200 rooms of the largest example with pings: probe p95 37 ms), m4 (63 KB
document; dragging makes an update per pointer move, 59/s), m5 (an hour
through the proxy), acceptance (two browsers, 14 checks, passed three times)
and clock (F's clock against ours).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A new package, packages/server: Hocuspocus 3.4.4 on /sync, a project a
document, refused with a reason when the store has not got it; every update
stored in Postgres as it comes and compacted after 500 updates, a minute of
quiet and when the last person leaves (FOR UPDATE on the row written when
the project is made); the janitor runs only here, at loading and 250 ms
after a change that is not its own; readers change nothing; /healthz; JSON
log. On closing it takes no one new and closes the sockets too (Hocuspocus'
close is a message after which a client signs in again on the same socket).
The tests use real clients against a throwaway Postgres in Docker.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The touch test runs in Chromium on the desktop and in WebKit as an iPad
(810x1080 and 1080x810): the acceptance of F4 as a script ("zwei
Rechnernetze" built anew by finger, set up and pinged across), two fingers
zoom the surface and never the page (M7, without a frame), every action
without a right click, every target at least 44 px. Real gestures: CDP in
Chromium, WebKit's own Input.dispatchTouchEvent through Playwright's
in-process connection (Playwright pinned to 1.62.1). A dragged finger holds
still before it is lifted, or Chromium takes it for a fling.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
packages/editor (React 18, Vite) draws a document as Java FILIUS does: Java's
default hardware icons, the device button of JSidebarButton, cables as
JCablePanel's quadratic curve between the buttons' middles, wireless cables
dashed, documentation items below. The surface moves with mouse and finger
(the pointer captured only once it moved) and zooms with Ctrl and the
wheel, a trackpad pinch, two fingers and the zoom bar; the first view shows
all. German and English catalogs with Java's device names, the dashboard's
colour tokens, the bridge with its origin check. packages/doc gets the
drawing's light view of a document.
Java's Switch.cloud (the Internet in the examples) was lost on reading; it
is now read, kept in the model and written, with FlsOracle and the corpus,
written and exported checks renewed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
packages/protocol holds the one list of /sim messages and checks every
field, naming what is wrong. packages/sim has the simulation room for F and
the web worker: the document's snapshot at start, the engine's clock on the
wall clock with Java's speed as cable delay, subscriptions to cable
activity, a card's new frames and a program's new window lines, capture and
clock kept over a stop, the command line as first action, a core failure
told to every client. LocalSimHost mirrors the document into the worker and
back without echo. M9 in Chromium: 17 examples give the same capture as in
Node, character for character; the room runs in a real worker.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
packages/doc holds the editor's model (ImportedProject) losslessly in a Yjs
document (schema 1): ports with their settings, each node's settings as
maps per field, programs, files as a flat path map in creation order (text
as Y.Text), own Java applications, links, documentation items, archive
entries. The core's snapshot and the .fls export go through the project.
The janitor repairs dangling links, two links at one port and files
without their folder, deterministically and idempotently. All 23 corpus
files come back through a Yjs update with the same snapshot and export.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Reader and writer ported from filius-ipad with the same path policy,
limits and error codes; DEFLATE via fflate, cut at the declared size
like Apple Compression. 54 tests: the Swift archive rejection tests,
DEFLATE edge cases, and all 23 corpus files cross-checked against
fflate.unzipSync and round-tripped.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
tools/java-oracle builds FILIUS at 6640354 and loads every .fls through
its own private loader under Xvfb. All 23 files load with every node
positioned and every cable resolved (Swift: 20 of 25 nodes in the
Internet example). MAC addresses FILIUS generates are left out, so two
runs are byte-identical.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
tools/swift-oracle builds the unchanged Swift runtime (pinned 3b5ded0)
in Docker and replays JSON scenarios; the TypeScript engine must produce
the same trace events, results and end state. Five scenarios, 884
events, identical.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
First slice of the TypeScript port of the filius-ipad runtime engine,
with 29 tests ported from the Swift suite. Parts not ported yet throw
NotPortedError instead of guessing.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
julian.kiedaischandClaude Opus 5