[SunkenInTime]

Test only, nothing on screen. What it catches:

Code under test	Seeds failing
main before #550 and #551	10 of the first 200
main now	0 of 10,000

Why

The sync tests are one scenario per bug: each was written after someone found the bug. #550 and #551 fixed four bugs nobody had a scenario for, and this test found three of them (a lost item edit, a theme change overwriting a teammate's map, and a regression in the first version of that fix). It is here so the next one is found before a user does.

What it does

One group, random schedules, in strategy_page_session_provider_test.dart. It uses the file's existing harness: the real page session, the real op queue and outbox, and _FakeServer.

From a seed it draws 30 steps out of: the user changes the map or theme, moves or retypes one of two texts; a teammate changes the map or retypes a third text; the next answer is lost; the connection drops or returns; the live read arrives; the pointer goes down or up; save; send. Then it goes online, lets everything send, and reads the server back. If the user is being asked about refused work, it takes the cloud's version.

It then requires:

• the canvas shows what the server holds: map, theme, and each text's words and place;
• the user's last words and place for their two texts are there, unless they were asked and took the cloud's;
• the last change to the map is the one kept, with one stated exception (the user set it back to what their device last read while a teammate's newer change had not reached them);
• the teammate's words are on the server and on the canvas;
• nothing is left waiting and nobody is still being asked.

A failure prints the seed, every step, the canvas, queue and server after each settle round, and every batch sent.

200 seeds run by default, about 11 s. SYNC_SEEDS=10000 runs more; SYNC_SEEDS=1907,2673 replays named seeds.

Limits

• The server is the file's Dart model, not Convex. The sync gauntlet covers the real server with a few scripted schedules; this covers many schedules against the model.
• One user and one teammate, one page, texts only. No page add, delete or reorder, no lineups, no restart mid-schedule.
• It does not model a read older than an acknowledged change, so it cannot see the one weakness Publish the outbox's summary and its waiting work together #550 left open.

ci: analyze test

[coderabbitai]

Pre-merge checks: 1 · 1

Failed checks: 1 inconclusive

Check name	Status	Explanation	Resolution
Title check		This assessment has not completed. The check will be retried on the next pre-merge check run.

Passed checks: 1 passed

Check name	Status	Explanation
Description Check		Check skipped - CodeRabbit’s high-level summary is enabled.

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

Share

• X
• Mastodon
• Reddit
• LinkedIn