[teamleaderleo]

Fixes two internal reports about NEXT iOS ignoring Reduce Motion:

1. The team drawer always slid the card and parallaxed the sidebar on open and close, including the scrim dismissal and a released drag.
2. Jump to latest from far up a transcript hid the whole transcript, waited 50 ms, faded it back over 150 ms and sprang the last screen over 400 ms.

They share one fix pattern, so they are in one PR: each controller reads UIAccessibility.isReduceMotionEnabled through a reducesMotion seam that the tests set, and settles at once when it is on.

Change

• SidebarContainerViewController.settle: under Reduce Motion the drawer writes its end geometry at once and runs its settle steps in place, with no animator. Those steps were moved out of the animator's completion into didSettle: model progress; a closed card's square corners and hidden sidebar; VoiceOver's screen-changed post to the opened sidebar. Unchanged: the publish hit-testing and accessibility state, finger tracking during a drag (track writes the geometry per frame) and the animated path with Reduce Motion off.
• ChatTranscriptController.scroll(to:animated:): under Reduce Motion an animated jump becomes an immediate offset change. Nothing hides the transcript and there is no fade or spring. The long-jump cut, the hidden frame (the stale-tile-over-the-composer workaround) and the spring are unchanged with Reduce Motion off, and the withoutReduceMotion… test still covers them.
• AgentChatSurface.scrollToBottom's SwiftUI fallback (used when no UIKit scroller is attached) also skips its 0.35 s animation under Reduce Motion.

Tests

Each test reads the layers' running animations as well as the end point, because an animator writes the final model values at once and the end point alone would pass on the old code.

• CmuxMobileShellTests/DrawerReduceMotionTests:
  • Open and close under Reduce Motion: the end geometry is set, no slide or parallax animation runs, and isOpen, progress, the hidden sidebar, corners and accessibility hiding are all set at once.
  • A released drag settles the same way, and the card follows the finger during the drag.
  • Control: with Reduce Motion off, the drawer still animates.
• CmuxMobileAgentChatTests/JumpReduceMotionTests:
  • From the top of a 9000 pt transcript, Jump to latest lands on the bottom at once, with the transcript at alpha 1 and no animation.
  • Control: with Reduce Motion off, the long jump still animates.

Red/green (cmux-next-ios kit tests, iOS simulator):

• red: the tests at this head with the fix commit (2e7f346) reverted (4908364), fleet step 708d4ad9 (cmux-next-ios-sim.sh test only=...JumpReduceMotionTests,...DrawerReduceMotionTests): 2 passed, 3 failed. aLongJumpLandsAtOnceWithTheTranscriptShown (the jump cut short to 7398 before 8272), openingAndClosingSettleInPlace and aReleasedDragSettlesInPlaceAndTheDragFollowsTheFinger (animation running, sidebar not hidden, corners 44) fail; both withoutReduceMotion… controls pass, so the oracle sees motion.
• green at d320e90: fleet step a804f2ce, 5 passed, 0 failed. Kit CI: https://github.com/manaflow-ai/cmux/actions/runs/41017568819

The controls failed on the first push because of how the tests were set up, not because of the fix. UIKit animates from the last committed frame, and a layout pass still pending when the jump starts snaps a following transcript to the bottom. So the tests now commit the start position first. The jump test records every offset and opacity the jump writes through KVO, and it jumps through the same animated scroll that scrollTranscript(toContentY:) uses.

No capture: Reduce Motion end frames match the animated end frames; tests assert no animation runs

UI: with Reduce Motion off, nothing changes. With it on, the drawer and the jump land in their final place, so a before/after still shows the same end frame; the tests above record the difference in motion.

Agent session: agent session

🤖 Generated with Claude Code

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}

---

Summary by cubic

Fixes NEXT iOS ignoring Reduce Motion: the team drawer always slid and parallaxed the sidebar on open and close, and Jump to latest from far up a transcript hid the transcript, faded it back, and sprang to the bottom. Both now settle at once when Reduce Motion is on, via a reducesMotion seam each controller reads.

Changes

• Under Reduce Motion, SidebarContainerViewController writes the drawer's end geometry and runs its settle steps (corners, hidden sidebar, VoiceOver focus) in place with no animator; drags still follow the finger.
• ChatTranscriptController turns an animated jump into an immediate offset change, keeping the transcript visible throughout.
• The SwiftUI fallback in AgentChatSurface.scrollToBottom also skips its animation under Reduce Motion.
• New tests verify no animations run and the end state lands at once; with Reduce Motion off, both behaviors are unchanged.

^{Written for commit 198d09d. Summary will update on new commits.}

[blacksmith-sh]

Found 5 test failures on Blacksmith runners:

Failures

Test	View Logs
main.DeveloperIdAppTests.test_download_prints_the_active_profile_only/ test_download_prints_the_active_profile_only	View Logs
stubs: /System/Library/Frameworks/Security.framework/Versions/A/Security has no stub/ Security has no stub	View Logs
stubs: CoreFoundation.tbd: install-name /System/Library/Frameworks/ CoreFoundation.framework/Versions/A/CoreFoundation, missing 1: ['_CFRetain']/ CoreFoundation, missing 1: ['_CFRetain']	View Logs
stubs: CoreFoundation.tbd: install-name /System/Library/Frameworks/ CoreFoundation.framework/Versions/A/CoreFoundation, missing 1: ['_log2']/ CoreFoundation, missing 1: ['_log2']	View Logs
test_the_ten_first_party_gpl_crates_are_named_exactly (main.AppCrateLicenseGate.tes t_the_ten_first_party_gpl_crates_are_named_exactly)/ test_the_ten_first_party_gpl_crates_are_named_exactly (main.AppCrateLicenseGate.tes t_the_ten_first_party_gpl_crates_are_named_exactly)	View Logs

^{Need help on this PR? Tag @codesmith-bot with what you need.}

[teamleaderleo]

Owned by an agent session · took over from an earlier session · session

[teamleaderleo]

Exact-head review at 2e76d48b50bcb7b638bd356c53af03254e4bfd50. This head is d320e909 merged with base, and the PR diff is unchanged.

Astra (gpt-6-astra, high): BLOCK.

• [P2] Reduce Motion is read when a settle or jump starts. Nothing watches reduceMotionStatusDidChangeNotification, so an animation already running when the setting changes keeps going.
• [P3] The jump control test that runs with Reduce Motion off checks only synchronous writes.

Sol (gpt-6.1-sol, high): VERDICT: SHIP 2e76d48b50bcb7b638bd356c53af03254e4bfd50.
Sol rejects the P2. Reading the setting when an animation starts matches UIKit. Flipping it means leaving cmux for Settings or Control Center, and the next settle or jump reads the new value. Sol found no P1, P2 or P3.

The P3, a stronger motion assertion in withoutReduceMotionTheLongJumpStillAnimates, is worth a follow-up but doesn't block.

[teamleaderleo]

Review (gpt-6.1-sol, high): SHIP at 57b130d

This head is 2e76d48b merged with base abe139dc, and the PR diff is unchanged. Sol found no P1 or P2. It again rejects the Reduce Motion notification P2 and repeats only the test-strength P3 at JumpReduceMotionTests.swift:130. I will land this once the kit job finishes and the classifier excuses every red.

[teamleaderleo]

Kit run 41101909944 at 57b130d2a20efa21f43d5c98a4b2926bcaf406ed: swift test (macOS) passed. The iOS simulator step failed 11 of 1651 tests:

• MacHomeFeedTests.* (feed.rows?.count == 4)
• HomeRetryTests (two tests: the same MacHomeFeedTests().until feed wait)
• TerminalBenchProbeAccountingTests (latency 299 ms against a 150 ms limit)
• TerminalSurfaceFramePacingTests ("too few batches (6)")

There was no crash and the run wasn't cut short. None of these tests touch the drawer or the transcript-jump code this PR changes.

base_red_excuse.py exits 1 on them, because the newest base kit run, 41095745168, passed. These are known timing flakes, but under the landing rule a red that passes on base is never excused. So I'm rerunning the failed job and will classify again; this does not land until the classifier exits 0.

[teamleaderleo]

The kit rerun at 57b130d2a20efa21f43d5c98a4b2926bcaf406ed (run 41101909944, attempt 2) still has 3 iOS simulator failures, and macOS swift test passed again:

• HomeRetryTests.failedHomeLoadRetriesTheFeedOnTheExistingLink
• MacHomeFeedTests.aFailedOpenIsTriedAgain
• TerminalSurfaceFramePacingTests.busyOutputDrawsAtMostOneFramePerDisplayRefresh

All three also failed in attempt 1. base_red_excuse.py exits 1, because the newest base kit run, 41095745168, passed. This PR doesn't touch the Home feed or terminal frame pacing, but under the landing rule a red that passes on base blocks.

This stays unlanded until those suites pass at this head, or until a fix for those flakes lands on base. Review status: Sol SHIP at this head.

[github-actions]

cmux status _{(one comment, edited in place; each line is updated by the job that owns it)}

Review 198d09d: HQ feed card · gallery before/after runs · checks

App of this head (app.zip, kept 1 day)

gh workflow run reload-build.yml -R manaflow-ai/cmux --ref feat-cmux-next -f tag=pr27776 -f ref=lane/hq5040-ios-reduce-motion -f platform=macos -f backend_mode=local

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.4115755828101 (7c0adaf184)
• cmux NEXT nightly 1.0.0-nightly.4115843934401 (8b3fa55ef5)
• cmux NEXT nightly 1.0.0-nightly.4115962701801 (631d310eb9)