Repository navigation
fix(ios): restore the back swipe's pre-swipe still span - #4515
Conversation
#4508 cut the post-swipe delay to 0.1 s, and the required still span was derived from that delay, so it shrank to about 0.4 s on a simulator. A screen that repaints itself every 0.4-0.55 s could then look still, and its own repaint after a no-op swipe read as a back. The still span now uses its own 250 ms margin instead of the delay: about 0.55 s on a simulator again. The 0.1 s after-sample and the 0.05 s sampling pause stay. The settle budget derives from the span, so it follows. The slow-toggle fixture repaints every 0.5 s again and runs without the edge pan, so the swipe is a no-op; back is tried at five timer phases and must answer back_swipe_unverified each time. Co-Authored-By: Claude Opus 5.5 <[email protected]>
Size Report
Startup median (7 runs, lower is better):
|
![cubic-dev-ai[bot]](https://avatars.githubusercontent.com/in/1082092?s=60&v=4){kind=small}
There was a problem hiding this comment.
All reported issues were addressed across 5 files
Reply with feedback, questions, or to request a fix.
View guided diff | Turn on auto-fix | Re-trigger cubic
Apex Sentinel · ✅ ApproveSince the last review: 0 new · 0 still open · 0 no longer raised No open findings. Reviewed with this repository's own Sentinel configuration ( 💬 Talking to SentinelIn a finding's thread, reply with:
On this conversation, comment Earlier reviewsReview of
|
| Where | Finding |
|---|---|
website/docs/docs/commands.md:127 |
🆕 User docs claim the runner tests use a 0.5 s repaint; the fixture repaints every 0.45 s |
Reviewed with this repository's own Sentinel configuration (ddbfc855661f).
Review of 8946026c · 13.11.2026 21:06 UTC
Apex Sentinel · ✅ Approve
1 finding(s): 1 minor.
🟡 Minor (1)
| Where | Finding |
|---|---|
apple/runner/AgentDeviceRunner/AgentDeviceRunner/AgentDeviceRunnerAppFixtures.m:788 |
New 0.5 s repaint fixture leaves the 0.55 s still span only 50 ms of slack, so the test can flake |
1 finding below minor hidden (user_settings.min_severity).
Reviewed with this repository's own Sentinel configuration (ddbfc855661f).
![apex-sentinel-staging[bot]](https://avatars.githubusercontent.com/in/5196457?s=60&v=4){kind=small}
There was a problem hiding this comment.
Apex Sentinel: 1 new finding on 8946026. The verdict and the full list are in the summary comment.
The `--slow-toggle` fixture now repaints on whole multiples of 0.45 s of the wall clock, which the test process shares, so the test waits for each target offset (0, 0.09, 0.18, 0.27, 0.36 s) instead of cumulative sleeps. The 0.45 s period leaves 100 ms below the ~0.55 s still span and stays above the old ~0.4 s span. The shade steps through 8 grays by wall-clock period, so samples on either side of a capture stall cannot alias. Co-Authored-By: Claude Opus 5.5 <[email protected]>
There was a problem hiding this comment.
All reported issues were addressed across 2 files (changes from recent commits).
Reply with feedback, questions, or to request a fix.
View guided diff | Turn on auto-fix | Re-trigger cubic
… its own slot Co-Authored-By: Claude Opus 5.5 <[email protected]>
There was a problem hiding this comment.
Apex Sentinel: 1 new finding on c5765c4. The verdict and the full list are in the summary comment.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
This PR is ready from the code side at 8dfb672: I found no problems in the restored still span. CI was still running when I checked, so no job has a failure to attribute yet. Smoke Tests (ios) is the job that overlaps this change, because the span governs the swipe-fallback back and the settle loop. Smoke Tests (macos) overlaps only through the shared runner build. Numeric Ranges and Coverage run no TypeScript from this diff. There are no conflicts. Before merge, Smoke Tests (ios), Smoke Tests (macos), Numeric Ranges, Coverage and the external review check need to finish green. Not blocking, and you can take or leave these: (1) the PR body's cost of about 0.15 s per swipe-path back is a simulator figure with 0.1 s captures, and on a device with 0.4 s captures the formula gives about +0.45 s, so scoping that sentence to the simulator would be more accurate; (2) the new doc line on I could not run Xcode or XCTest. The five-phase test result comes from your thread replies, and no physical-device run exists. The 1.05 s span and 1.4 s settle time are derived from the formula and the earlier capture measurement. Of the earlier review threads, these no longer apply and can be resolved: the cumulative-phase concern (#4515 (comment)) because every attempt now waits for the wall-clock offset; the 50 ms slack concern (#4515 (comment)) because the fixture period is 0.45 s and a unit test pins the 100 ms margin; the floor shade concern (#4515 (comment)) because a late callback now keeps its own slot; and the commands.md wording concern (#4515 (comment)) because the docs now match the 0.45 s fixture. |
thymikee
added
the
ready-for-human
* origin/main: fix(android): carry an empty text on a node labelled by its content description (#4531) fix(ios): restore the back swipe's pre-swipe still span (#4515) fix(daemon-client): type the remote daemon refusal with its dispatch fact (#4532) fix: prepare reinstall artifacts before removing apps (#4533) fix(ios): classify the keyboard dismiss refusal by what it dispatched (#4536) fix(ios): press the navigation bar back button and speed up the back swipe (#4508) refactor(gestures): one owner for the densified gesture path; drop the unproven sendevent transport (#4239) fix(host-kit): keep Linux process identity stable across clock steps (#4438) fix(ios): bound the idle wait of keyboard dismiss and keyboard enter (#4454) fix(device-selection): fail an ambiguous device request with only the tied devices (#4506)
Summary
Follow-up to #4508. iOS
back's swipe fallback required the screen to be still for a span derived from the post-swipe delay. When #4508 cut that delay to 0.1 s, the span shrank from about 0.55 s to 0.4 s on a simulator. A screen that repaints itself every 0.4–0.55 s (carousel, pulsing badge) could then look still, and its own repaint after a no-op swipe was read as a successful back. The agent then works on the wrong screen, and a retry can go back twice.The still span now has its own 250 ms margin, independent of the delay: about 0.55 s on a simulator again. The 0.1 s after-sample and the 0.05 s sampling pause stay. The settle budget derives from the span. The cost is about 0.15 s on swipe-path backs only; the navigation-bar button path is unaffected. 5 files.
Validation
Head 8946026, base ddbfc85.
back_swipe_unverified. With the old 100 ms span, that test fails (misread as success).check:xctest-selection, focused vitest (109), typecheck, and lint pass.backin Safari history, 8 runs each, from runner trace: settle samples 4 → 5, still time median 422 → 574 ms (+152 ms). After-sample gap unchanged (~320 ms).check:affected.🤖 Generated with Claude Code
Open in [code]smith · Need help on this PR? Tag
@codesmith-botwith what you need.