You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
fix(mobile): iOS 18 iPad sidebar and thread feed start below the header - #32936
On iPadOS / iOS before 26 (no Liquid Glass, so the bars are solid), two lists start underneath their column's header. The native column lays content out under the bar, but these lists add no top inset there:
In the iPad sidebar, the newest threads sit behind the title bar and the Search field.
In the right-hand inspector pane showing a subagent, the top of the loaded page is behind the pane's header. The "Load earlier activity" button can't be tapped, and the first rows are hidden under the title.
The same happens to the top of an ordinary thread feed, to the empty-thread placeholder, and to the first message sent into an empty thread, on iPad and iPhone.
Change
Sidebar:ThreadNavigationSidebar now uses UIKit's automatic content inset (contentInsetAdjustmentBehavior="automatic", with indicator insets) on every iOS version. Before, it was only on when glass was supported. UIKit then clears the bar and its stacked Search field. On iOS 26 the props are the same as before. The WORKSPACE_LIST_SCROLL_INSETS constant only existed to turn this off on iOS < 26, so it is removed.
Feed:ThreadFeed no longer takes a contentTopInset prop. Its only caller always passed 0.
When the feed does not get the native contentInset (iOS < 26), its top inset is useContentTopInset(), the column's measured top safe area. That is the bar height where the column lays content under the bar, and 0 on Android.
On iOS 26, where the feed already gets a native contentInset of the bar height, it stays 0.
This value drives the existing list-header spacer, the empty and unavailable placeholders, the scroll-indicator inset and the first-send anchor offset.
The header-material scroll check adds the bar height only on the native-inset path, so the spacer does not count as a scroll.
The doc comment on useContentTopInset now says it applies to solid bars too.
vp run check:changed passes on the head 08e1fb258a (format, lint, knip, typecheck, react compiler, tests beside the changed files).
This is a visual bug, so the proof is device captures and no new test. Before is main at ca796d5211 (N1: 80e1e4d14d). After is eb39186ba6, f76e21be6c (comment-only on top), or 4980e33628, as each row says; the head 08e1fb258a only merges main with no apps/mobile changes. Both sides share the native fingerprint and Legend List 3.7.2. Main merged since eb39186ba6 changes only file preview, asset loading and the Android rail width (0 off Android) under apps/mobile. The changed inset paths are identical apart from comments.
Environment: pool iOS simulators driven by agent-device, 1x capture pixels per point, default text size, fresh dev servers seeded with synthetic Replay (dev) threads. iPad (7th generation) iOS 18 1080×810 landscape, iPhone Xs iOS 18 / 18.6 375×812 portrait, iPad Pro 11-inch (M4) iPadOS 26.0 1210×834 landscape. Each comparison is cropped to the pane it covers, or has the status strip cropped and generated titles and timestamps masked identically on both sides. Diff0 marks every changed pixel; DiffT uses 5% fuzz.
Row
Before
After
Diff0
DiffT (5%)
S1 iPad 7 iOS 18, light. Newest rows behind bar and Search → row 20 first below Search. ca796d5211 → eb39186ba6
S2 iPad 7 iOS 18, light, status strip cropped. Empty sidebar: "No threads yet" moves below Search. ca796d5211 → eb39186ba6
F1 iPad 7 iOS 18, light, detail pane crop. Parent thread top: first rows under the header → below it. ca796d5211 → eb39186ba6
F8 iPhone Xs iOS 18, light, timestamps masked. Short settled feed: zero changed pixels. ca796d5211 → eb39186ba6
I2 iPad 7 iOS 18, light, inspector crop. Earlier-page error: control unreachable → error wraps below it. ca796d5211 → f76e21be6c
I3 iPhone Xs iOS 18, light, keyboard up, status strip cropped. First send into an empty thread: bubble under the bar → 16 pt below it. ca796d5211 → f76e21be6c
N1a iPhone Xs iOS 18.6, light, keyboard up, title/timestamps masked. New task → first send: bubble hidden under the bar → below it. 80e1e4d14d → 4980e33628
N1b iPhone Xs iOS 18.6, light, keyboard down, same masks. Reply 88 pt lower, clear of the bar. 80e1e4d14d → 4980e33628
I3:before and after. Fill at 2 s, send at 6 s, then the reply, keyboard dismissed, and pans.
N1:before and after. New task at 1 s, send at 23 s, refocus at 25–26 s, reply at 26–29 s. Send closes the keyboard on both sides.
S5/S6b sidebar Search, dark: focus, results, no matches, clear, cancel. I4 sidebar hide/restore: row 07 stays in place. I5 Search jump: lands at about 122 pt before and 120 pt after, unchanged. All three are in run 3.
What else changed in the captures: only the moved content above. Inside each crop or mask, the following still differ:
the status-bar clock;
generated thread titles, Replay timestamps and durations;
Expo dev-client overlays.
The recordings also carry touch markers and grey boxes over model labels. In the I1 before recording the inspector is 266 pt wide; after, it is 260 pt. Run 3's Search recordings show a Working counter only on the after side.
Not checked:
Android. useContentTopInset returns 0 there, and the sidebar change is in the iOS-only branch.
The compact iOS 26 send lifecycle. The feed keeps its native-inset path there with unchanged values, and G1 shows the glass inspector pixel-identical.
Verification log
Main merged into the branch again (4980e33628, then 08e1fb258a), with no conflicts.
Reviews: three earlier rounds and a round after each later merge, each with two independent reviewers. No code defects found. Findings were missing or stale captures and body text, now addressed. On 4980e33628 both reviewers returned MERGE with no findings.
Proof audit: the matrix was re-derived on 4980e33628. Its two passes found clock, timestamp and parent-pane noise in S2, F2, F3, F5, F8, I2, I3 and N1, and a before/before G1 pair. Those comparisons were rebuilt as above. The auditor rebuilt all ten independently with zero pixel delta and returned sufficient.
A message-search jump with the keyboard up can land the target partly under the bar on iOS 18. It lands in the same place before and after this change (scrollToIndex with viewPosition: 0.2), so it is out of scope here.
AI-assisted; verified with the tests and tools listed above.
Before iOS 26 the native bars are opaque, but the column still lays content
out under them. The iPad thread sidebar turned UIKit's automatic inset off
there, and ThreadFeed always got a top inset of 0. So the first rows and the
"Load earlier activity" button sat behind the bar, where they could not be
tapped.
- The sidebar list uses UIKit's automatic content inset on every iOS version,
as Home does, so it clears the bar and its stacked search field.
- When UIKit doesn't inset the feed, ThreadFeed sizes its top spacer, anchors
and indicator insets from the column's measured top safe area
(useContentTopInset). That value is 0 on Android and on iOS 26 it stays 0
because UIKit owns the inset there.
Macroscope's review found this PR approvable — This corrects sidebar and thread-feed spacing beneath iOS headers and aligns header appearance with scrolling. The changes are limited to layout behavior, with no persistent-data or external-system effects.
No code changes detected at 08e1fb2. Prior analysis still applies.
You can add or adjust custom eligibility rules. Learn more.
Handoff from the issue-watch investigation of #32884. I did not reproduce this on a device: the iOS/Android pools were empty and shiv was unreachable during my run. No app build, native interaction, screenshot or recording was produced, and no format/lint/typecheck or repository test suite ran.
Source investigation at 22dc6f9e98cc826ac3581aeeb0fb1852f49a43f1 found the sidebar's glass-gated automatic inset and the feed's pre-26 combination of automatic insets disabled plus contentTopInset={0}. The sidebar overlaps the existing work in #22550. I also identified the separate end-follow/snapback path covered by #31907; I did not change that patch.
A local alternative fix, f253083cc3629d72cf6ca12b4dd97e9a48d51df2, enabled automatic insets for the sidebar and all iOS thread routes. Independent source review found no definite new defect, but blocked readiness on missing native evidence and repository checks. The only executed validation was git diff --check, which passed. That local commit was never pushed and is being left behind now that this PR owns the issue. No runtime evidence from my work is available to add to yours.
No actionable comments were generated in the recent review.
Walkthrough
ThreadFeed now gets its top inset from useContentTopInset and adjusts header-material visibility using barInset when native automatic insets are enabled. The thread navigation sidebar now uses UIKit automatic inset adjustment on iOS.
The inset documentation now describes the iOS bar height for glass or solid bars and manual insets when UIKit automatic insets are disabled. ThreadFeed removes the contentTopInset prop, reads the column top inset from useContentTopInset, and uses barInset for the header-material visibility offset when native automatic insets are enabled.
The shared WORKSPACE_LIST_SCROLL_INSETS configuration is removed. On iOS, the sidebar list now enables automatic scroll-indicator and content-inset adjustment.
This change makes the iPad sidebar and thread feed start below their headers on iOS 18. No concrete merge-blocking risk was found in the supplied changes.
Pre-merge checks: 3 · 1
Failed checks: 1 warning
Check name
Status
Explanation
Resolution
Description check
The description clearly explains the problem, change, and detailed verification. The Scope and approval section links triaged issues but does not include a maintainer approval comment or explain why t…
Add a link to the triaged issue or discussion with the maintainer’s explicit approval of the direction and scope, including the approval comment. If the change qualifies as a small, focused fix of an obvious bug without prior approval, expl…
Passed checks: 3 passed
Check name
Status
Explanation
Title check
The title clearly identifies the iOS 18 sidebar and thread-feed inset fix.
Linked Issues check
Open, directly linked issue #32884 requires sidebar content to remain visible below its header. ThreadNavigationSidebar now enables UIKit automatic content and scroll-indicator insets on iOS. `Threa…
Out of Scope Changes check
All reviewed changes support issue #32884. The sidebar inset change fixes the left list. The feed inset change and removal of its contentTopInset override fix the subagent pane and apply the same in…
Full details: Linked Issues check
Explanation
Open, directly linked issue #32884 requires sidebar content to remain visible below its header. ThreadNavigationSidebar now enables UIKit automatic content and scroll-indicator insets on iOS. ThreadFeed uses useContentTopInset() when UIKit does not apply automatic insets, and keeps the manual inset at zero on the automatic path. This clears the left sidebar and subagent feed below their bars. The PR description reports iPadOS 18 captures for both cases, including the earlier-activity button below the inspector header. Issue #32884 does not require an automated test, so the absence of a new test is not an unmet requirement.
Full details: Out of Scope Changes check
Explanation
All reviewed changes support issue #32884. The sidebar inset change fixes the left list. The feed inset change and removal of its contentTopInset override fix the subagent pane and apply the same inset to the list spacer, placeholder, indicators, and anchors. Removing WORKSPACE_LIST_SCROLL_INSETS follows from enabling automatic insets on the sidebar. The useContentTopInset comment documents the new solid-bar use. No unrelated change is evident.
Full details: Description check
Explanation
The description clearly explains the problem, change, and detailed verification. The Scope and approval section links triaged issues but does not include a maintainer approval comment or explain why the change qualifies for the small, obvious bug exemption.
Resolution
Add a link to the triaged issue or discussion with the maintainer’s explicit approval of the direction and scope, including the approval comment. If the change qualifies as a small, focused fix of an obvious bug without prior approval, explain why.
Finishing Touches
Generate docstrings
Commit to this branch
Create a new PR
Generate unit tests (beta)
Commit to this branch
Create a new PR
Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Reviewing files that changed from the base of the PR and between eb39186 and f76e21b.
Files selected for processing: 2
apps/mobile/src/components/ui/useBarInset.ts
apps/mobile/src/features/threads/ThreadFeed.tsx
Files skipped from review as they are similar to previous changes: 2
apps/mobile/src/components/ui/useBarInset.ts
apps/mobile/src/features/threads/ThreadFeed.tsx
Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 0 remain after this review.
… (Checking unused code)
Failing check: Checking unused code
**What failed:** "Checking unused code". Knip reported one unused file: `apps/mobile/src/features/layout/workspace/workspace-list-insets.ts`.
**Why:** Before the merge, Loom used this file through its `WORKSPACE_LIST_SCROLL_INSETS` export in `ThreadNavigationSidebar.tsx`. Upstream's pingdotgg#32936 (80e6261) removed that export and its import. All that's left in the file is `useWorkspaceListEmptyTopInset`, and upstream calls it only from the early-return empty state in `HomeScreen.tsx`. Loom's `project-navigation` seam replaces that early return with empty states rendered inside the list. The list already sits below the glass bar through `contentTopInset`, so the fork has no caller for the hook and doesn't need one. Pre-merge, the hook was already an unused export, but knip's export pass doesn't cover `apps/mobile`. Nothing flagged it until it became the file's only export.
**What I changed:** Deleting upstream's file isn't allowed, and adding the inset to the in-list empty state would offset it twice. Instead I added a marked seam to `knip.jsonc`: an `ignore` entry for that file in the `apps/mobile` workspace, marked `// fork: ext-lint`. This copies the existing `apps/web` thread-inspector ignore. I bumped the `knip.jsonc ext-lint` row in `docs/fork/seams.tsv` from 7 to 8 and added a sentence to the knip note in FORK.md. `vp run knip:check:all` now passes (3/3), and `check_seams` exits 0.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
bugfixFix PR opened by an issue lead; enters the PR looppr:readyEvery gate passed; waits for Julius's land labelsize:S10-29 changed lines (additions + deletions).vouch:trustedPR author is trusted by repo permissions or the VOUCHED list.
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
github-actions
Bot
added
vouch:trusted
![macroscopeapp[bot]](https://avatars.githubusercontent.com/in/900172?s=60&v=4){kind=small}
juliusmarminge
added
bugfix
Problem
On iPadOS / iOS before 26 (no Liquid Glass, so the bars are solid), two lists start underneath their column's header. The native column lays content out under the bar, but these lists add no top inset there:
The same happens to the top of an ordinary thread feed, to the empty-thread placeholder, and to the first message sent into an empty thread, on iPad and iPhone.
Change
ThreadNavigationSidebarnow uses UIKit's automatic content inset (contentInsetAdjustmentBehavior="automatic", with indicator insets) on every iOS version. Before, it was only on when glass was supported. UIKit then clears the bar and its stacked Search field. On iOS 26 the props are the same as before. TheWORKSPACE_LIST_SCROLL_INSETSconstant only existed to turn this off on iOS < 26, so it is removed.ThreadFeedno longer takes acontentTopInsetprop. Its only caller always passed0.contentInset(iOS < 26), its top inset isuseContentTopInset(), the column's measured top safe area. That is the bar height where the column lays content under the bar, and 0 on Android.contentInsetof the bar height, it stays 0.useContentTopInsetnow says it applies to solid bars too.Scope and approval
Fixes #32884
Fixes #22543
Triaged via the automated issue flow. https://git.cafe/pingdotgg/t3code/issues/22543 reports the sidebar half on its own.
Verification
vp run check:changedpasses on the head08e1fb258a(format, lint, knip, typecheck, react compiler, tests beside the changed files).ca796d5211(N1:80e1e4d14d). After iseb39186ba6,f76e21be6c(comment-only on top), or4980e33628, as each row says; the head08e1fb258aonly merges main with noapps/mobilechanges. Both sides share the native fingerprint and Legend List 3.7.2. Main merged sinceeb39186ba6changes only file preview, asset loading and the Android rail width (0 off Android) underapps/mobile. The changed inset paths are identical apart from comments.ca796d5211→eb39186ba6ca796d5211→eb39186ba6ca796d5211→eb39186ba6ca796d5211→eb39186ba6ca796d5211→4980e33628ca796d5211→eb39186ba6ca796d5211→f76e21be6cca796d5211→eb39186ba6ca796d5211→f76e21be6cca796d5211→f76e21be6c80e1e4d14d→4980e3362880e1e4d14d→4980e33628ca796d5211→4980e33628Recordings:
Evidence write-ups:
Seed and driver scripts: seed-rpc.mjs, driver.md, capture.py, seed.py, recapture-seed.mjs, repair-comparisons.py, repair-four-comparisons.py.
What else changed in the captures: only the moved content above. Inside each crop or mask, the following still differ:
The recordings also carry touch markers and grey boxes over model labels. In the I1 before recording the inspector is 266 pt wide; after, it is 260 pt. Run 3's Search recordings show a Working counter only on the after side.
Not checked:
useContentTopInsetreturns 0 there, and the sidebar change is in the iOS-only branch.Verification log
4980e33628, then08e1fb258a), with no conflicts.4980e33628both reviewers returned MERGE with no findings.4980e33628. Its two passes found clock, timestamp and parent-pane noise in S2, F2, F3, F5, F8, I2, I3 and N1, and a before/before G1 pair. Those comparisons were rebuilt as above. The auditor rebuilt all ten independently with zero pixel delta and returned sufficient.scrollToIndexwithviewPosition: 0.2), so it is out of scope here.AI-assisted; verified with the tests and tools listed above.