Paul HittandClaude Opus 5
d8b8d63687
Verifying a layout meant booting an 8 GB AVD, installing an APK, tapping to a screen and pulling a screencap — and the emulator was killed by memory pressure four times in one session, each time mid-check. Robolectric in native-graphics mode draws the same Compose pixels on the JVM. `Scripts/screenshots.sh` now renders the screens in about four seconds with no device. Two things had to be got right and both fail in a way that reads as something else: `isIncludeAndroidResources` is not optional — without it the theme cannot resolve and every render throws on a missing resource, which looks like a Compose problem rather than a configuration one. 🔴 And the capture. `captureToImage()` goes through PixelCopy and waits for a frame-commit callback Robolectric never delivers, so every render died as "Condition still not satisfied after 2000 ms" — which reads as a hung composition and is nothing of the sort. Drawing the decor view into a software canvas produces the same pixels with no window involved. `createComposeRule()` has the same problem for the same reason; it needs the Activity rule. ⚠️ These are output, not assertions: they write files and pass unless something throws. A screenshot test that diffs against a stored baseline is a different thing, worth adding once the design stops moving — right now it is moving, and a baseline would be re-blessed on every change, which proves nothing. This immediately paid for itself. The address card had only ever been seen in its empty state, because the emulator's stored session predates the fnid and a fresh sign-in erases that phone's keys. It renders populated here, and the QR in it — encoded by our own Kotlin encoder, drawn by the app's own Compose code — decodes back to the exact address through CoreImage's detector, which shares no line with any of it. Rendered at the kit's own 375x812 canvas, so a PNG can be put beside the Sketch artboard rather than mentally rescaled from a 1080x2400 phone. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Flamenet Messenger — Android
Nothing here is an app yet. This is the engine, and it is deliberately first.
An Android client needs the protocol implemented a third time. That third
implementation is the one that finds the places where the specification says
less than the two existing engines quietly agreed on — but only if it is
written against the spec and the published vectors rather than against a
reading of our own Swift and JavaScript. So the engine comes first, it is a
plain Kotlin/JVM library with no Android in it at all, and it is held to
engine/test/vectors/protocol-v1.json — the same file the other two engines
are checked against, read from that one location rather than copied here.
./gradlew :engine:test
No Android SDK, no emulator, no device. It runs anywhere a JVM does, which is also what lets it run in CI on Linux.
What the engine covers today
| Area | Spec | State |
|---|---|---|
| Safety numbers | §7 | Done — fingerprints and conversation ordering match the published vectors |
| Key transparency verification | §19 | Done — RFC 6962 leaf/node hashing, unbalanced split, bottom-up audit paths |
| Byte helpers | — | Unsigned comparison, constant-time equality |
What it does not cover
Listed rather than left to be discovered, because a table of green ticks with the hard parts quietly missing is worse than no table:
| Area | Spec | State |
|---|---|---|
| X25519 / Ed25519 | §3 | Not started — needs a provider that works on both JVM and Android |
| X3DH key agreement | §4 | Not started |
| Double Ratchet | §6 | Not started |
| Sealed sender | §12 | Not started |
| ML-KEM-768 hybrid | §11 | Not started |
| ML-DSA-65 identities | §20 | Not started |
| Relay client | §8 | Not started |
| The app itself | — | Not started |
Until those exist, Android is honestly marked "not started" on the website, and a conversation gets the protocol's guarantees only when both ends are iOS.
Why the vectors matter more than the code
Two implementations that agree may both be wrong in the same way, especially when one was written while reading the other. The audit-path ordering in §19 is the worked example: a top-down reading of a bottom-up path produces the correct root for every left-spine index, so it passes any test that only checks index 0. The vector file pins a right-descending index for that reason, and the Kotlin tests assert both that the published path verifies and that the same path reversed does not.
If this engine ever disagrees with the file, the interesting question is which of the three is wrong — not this one by default.