Revert the motion capability: the clock, the probe and the filmstrip, until a method is proven #433
Loading…
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
What happens. The motion capability shipped on 11-07 (#407) and 11-08 (#415) has not paid off and is broken in part. Run 182's judge: the filmstrip helper fails on production (it resolves the Playwright CLI symlink before walking up to the package, so Node finds no
playwrightand no strip was ever composed); the pasted motion probe went unused (36 freezes of the judge's own, because the probe cannot wait for a server reply); and the paused clock caused two 1,200 s stalls, about 50 minutes (findings 14, 17, 18 on #425). The judge read single frames instead, as before #407.Fix. Take the capability out whole, back to the judge's wording before #407, until #434 finds a method that works: the
motion-filmstripspec (the page clock, the frozen frame, the filmstrip, frames-only conclusions, the paused clock holding what a server pushes) and the judge's two requirements "A check freezes motion shorter than a turn" and "A check captures motion with a shipped probe". In code:src/osf/agent/braid_clock.jsand its install by the browser server (OSF_BRAID_CLOCK),src/osf/engine/probes/motion-frame.js,src/osf/engine/filmstrip.pyand.mjs, the motion paragraphs of the judge's andbrowser's briefs (_SEE_MOTION, the motion lines of_PLAN_THE_STATES, the probe paragraph of_PROBES), the guide's section, and their tests (test_see_motion.py,test_filmstrip.py, the clock parts oftest_shape_browser.pyandtest_path_guard.py). The 180 s page-script bound (PR #428) stays, on the machine's plain timer. The text and canvas contrast probes stay. An animation scenario is judged from its start and end states and what a screenshot shows; the brief says motion itself is not asserted. #429 closes as moot.Done when. The spec and brief tests for motion are gone and the suite is green; a replay of run 182's judge step with the capsule writes no clock calls and names no filmstrip; a simulated run passes; the guide no longer describes the clock.
From the review of run 182's findings (braid#425).
Shipped in PR #435 (merged
5983d1f8), archived in PR #436. Gone: Braid's page clock and its install by the browser server (OSF_BRAID_CLOCK), the motion probe, the filmstrip helper and script, the brief's paragraphs on them and the rule "never write a freeze of your own", theinit_scriptoption ofmcp_config, their tests and the clock journeys, and themotion-filmstripcapability. Kept: the braid#346 one-call freeze in thebrowserprompt and the judge's own-browser rules, the 180 s page-script bound on the plain timer, and the text and canvas probes.Evidence from the laptop:
./tools/test.shgreen (531 unit tests, 44 journeys),./tools/sim.sh judge-requirement-by-namepasses with the Visual check's brief no longer naming the filmstrip, and a new brief test holds that the probes are the text and canvas probes only. The research into a method is #434.