[trissim]

The pilot settles on owed work instead of sleeping on a CPU-time guess.

Pilot.press slept wait_for_idle(0) twice per key (at least 20 ms each, looping in 20 ms steps while the loop thread burned more than 1 ms per step), and every pause slept it again. The exact signal was already there: each queue says Owing after a put and Settled when its pump finds it empty.

	main	this branch
full suite (-n auto, snapshots excluded)	39.8 s, 21 failed	26.5 s, 22 failed (only test_slug 4/5 and a gc-timed worker test differ, both seen failing on main runs)
test_pilot.py, test_data_table.py, test_app.py, serial	27.9 s	12.8 s
Toad's 5 keypress-heavy files, serial, one worker	135.9 s	~88 s

One owner for "the app has handled what it was given"

App._settle repeats until nothing is owed:

1. one loop turn, so input a driver handed over through call_soon_threadsafe is on its queue;
2. waits behind every owing pump on the screen; work that posts more makes its target owe again;
3. delivers a resize still waiting on its 1/120 s debounce;
4. runs the screen update itself (Screen._update_now: the timer's update, the rest of a layout split into steps, then a prompt so the idle check resumes the timer for work requested meanwhile), at most SETTLE_UPDATES (2) times: a layout's own consequences land in the second, and a widget that asks for a layout on every layout never stops asking.

It stops when the app raises or exits (the app's loop task ending is part of every wait). A running animation is not owed work: it may run for a minute, and callers who need it finished use Pilot.wait_for_animation.

press (before its first key and after each), pause, exit, the animation waits and run_test all call it. Pilot._wait_for_screen is deleted; WaitForScreenTimeout moves to app.py beside its raiser; textual._wait has no caller left and goes with its test.

Bugs the sleep was hiding

• The key raced its own wait. _press_keys sent keys through driver.send_message, a thread-safe hop that lands a loop turn later. Keys are now posted on the loop _press_keys already runs on.
• A key that stops the app hung on the dead app's animator; only assert driver is not None got it out.
• call_after_refresh strands its callback while a refresh stays pending without a published frame (my first cut waited on it: 49 suite tests sat on the 30 s round timeout). The settle runs the update instead of waiting for a frame.
• text_wrap and text_overflow were declared repaint-only but decide how many lines an auto-sized widget measures. Their test passed on main only because the startup resize, still behind its debounce, re-laid the screen out after the change.

Tests that leaned on the sleep

• Tabs moves its underline UNDERLINE_DELAY (0.02 s) after activation: the literal is now on the class, and the animation tests pause for it.
• test_log_from_worker's app exited itself after one refresh, racing the thread it tested.
• test_data_table pauses through the pilot instead of importing wait_for_idle.
• Toad: Neovim's mode after escape and a click aimed during a sidebar scroll animation now wait for their receipts (Two tests wait for what they assert on, not for a pilot sleep to cover it toad#1610, merged).

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.

An organization owner can choose the repository organization as the billing source for reviews requested by members with a paid Copilot license. Learn about Copilot code review billing.