fix(mobile): keep the iOS home header stable

[rynfar]

Rewrites the react-native-screens@4.25.2 patch so iOS header subviews are
collected into explicit left/right bar-button arrays instead of the navigation
item's items being cleared and rebuilt, which is what made the home header jump.

Adopted from T3 Code b982847ab201795711884d1dfadfddb3864adb23
(pingdotgg/t3code#8467), cherry-picked clean. Patch file plus lockfile only, no
Pylon source touched.

Scope note

This was originally scoped alongside upstream #8399 (Git action toast glass) as
one "mobile cosmetics" change. #8399 is not here because #138 already
contains it: #7327 rewrites GitActionProgressOverlay.tsx on top of #8399, and
that PR takes upstream's version of the file wholesale. Cherry-picking #8399
separately would have re-applied a superseded version.

Verification

• pnpm install succeeds, so the rewritten patch applies against
  react-native-screens@4.25.2
• Confirmed the patch actually landed rather than being silently skipped: the new
  subviewLeftItems symbol is present in the installed
  ios/RNSScreenStackHeaderConfig.mm
• pnpm-lock.yaml is unchanged after install, so the committed lockfile matches
  what pnpm resolves
• vp run -F @t3tools/mobile typecheck — clean; vp test run on mobile — 25 passed

Not verified

The patch is iOS native Objective-C. Nothing here proves the header actually
stops jumping — that needs a simulator or device run, which this change has not
had. Tests and typecheck can only show the patch is well-formed and the JS side
still resolves. Worth pairing with the test-pylon-mobile pass that #138 also
wants, since both land on the same surface.

Claude Opus 5 via Claude Code.

---

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

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.3 KiB	+47 B (+0.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	+1 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.4 KiB	+46 B (+0.7%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	55.6 KiB	+44 B (+0.1%)	66.4 KiB	✅
Codex	Live turn messages	9	10	+1 (+11.1%)	21	✅
Claude	Total thread wire	13.3 KiB	13.3 KiB	−3 B (−0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	+16 B (+0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.5 KiB	6.4 KiB	−19 B (−0.3%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	56.4 KiB	−44 B (−0.1%)	66.4 KiB	✅
Claude	Live turn messages	11	10	−1 (−9.1%)	21	✅

Baseline: 3d6035f · PR result: bb99f51 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[rynfar]

Independent review pass. Two findings, both latent, neither fixed.

The review verified more than my own pass could: it reverse-applied the previous
patch against pristine react-native-screens@4.25.2, applied the new one, and
read the resulting ios/RNSScreenStackHeaderConfig.mm in final form. It cleared
the things worth worrying about in a native patch — no retain cycle from
retaining config on the navigation item (screenView is weak and
RNSScreen.shouldBeRecycled is NO), the @[...] literal cannot take a nil
element, dropping the leftBarButtonItems = nil resets is safe because iOS 16+
never writes that property now, and the nil→empty-array change only affects the
pre-iOS-16 branch, which is dead at the app's 18.0 deployment target.

Stale grouping when hidesSharedBackground toggles. The reuse cache is keyed
on [subview getUIBarButtonItem] object identity, but -setHidesSharedBackground:
mutates that cached item in place, so toggling the prop on a React header subview
no longer recomputes leading/trailingItemGroups. Previously any subview item
forced a rebuild every pass. Mechanism confirmed; not reachable today, because
Pylon only sets that prop on dict-based items, which are in the key.

Reuse key omits navitem.searchController. The searchBarPlacement branch
builds items from it, so a first pass taken before the search bar subview mounts
could be cached permanently. Newly reachable for screens with React header
subviews, but not exercised by current Pylon mobile code.

Both are latent upstream issues in vendored native code and neither is verifiable
without a device, so I have not touched them. They are worth re-checking during
the test-pylon-mobile pass this PR already needs, particularly if a header
search bar or a hidesSharedBackground toggle lands later.

[rynfar]

Rewrites the react-native-screens@4.25.2 patch so iOS header subviews are
collected into explicit left/right bar-button arrays instead of the navigation
item's items being cleared and rebuilt, which is what made the home header jump.

Adopted from T3 Code b982847ab201795711884d1dfadfddb3864adb23
(pingdotgg/t3code#8467), cherry-picked clean. Patch file plus lockfile only, no
Pylon source touched.

Scope note

This was originally scoped alongside upstream #8399 (Git action toast glass) as
one "mobile cosmetics" change. #8399 is not here because #138 already
contains it: #7327 rewrites GitActionProgressOverlay.tsx on top of #8399, and
that PR takes upstream's version of the file wholesale. Cherry-picking #8399
separately would have re-applied a superseded version.

Verification

• pnpm install succeeds, so the rewritten patch applies against
  react-native-screens@4.25.2
• Confirmed the patch actually landed rather than being silently skipped: the new
  subviewLeftItems symbol is present in the installed
  ios/RNSScreenStackHeaderConfig.mm
• pnpm-lock.yaml is unchanged after install, so the committed lockfile matches
  what pnpm resolves
• vp run -F @t3tools/mobile typecheck — clean; vp test run on mobile — 25 passed

Not verified

The patch is iOS native Objective-C. Nothing here proves the header actually
stops jumping — that needs a simulator or device run, which this change has not
had. Tests and typecheck can only show the patch is well-formed and the JS side
still resolves. Worth pairing with the test-pylon-mobile pass that #138 also
wants, since both land on the same surface.

Claude Opus 5 via Claude Code.

---

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

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.3 KiB	+47 B (+0.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	+1 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.4 KiB	+46 B (+0.7%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	55.6 KiB	+44 B (+0.1%)	66.4 KiB	✅
Codex	Live turn messages	9	10	+1 (+11.1%)	21	✅
Claude	Total thread wire	13.3 KiB	13.3 KiB	−3 B (−0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	+16 B (+0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.5 KiB	6.4 KiB	−19 B (−0.3%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	56.4 KiB	−44 B (−0.1%)	66.4 KiB	✅
Claude	Live turn messages	11	10	−1 (−9.1%)	21	✅

Baseline: 3d6035f · PR result: bb99f51 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[rynfar]

Independent review pass. Two findings, both latent, neither fixed.

The review verified more than my own pass could: it reverse-applied the previous
patch against pristine react-native-screens@4.25.2, applied the new one, and
read the resulting ios/RNSScreenStackHeaderConfig.mm in final form. It cleared
the things worth worrying about in a native patch — no retain cycle from
retaining config on the navigation item (screenView is weak and
RNSScreen.shouldBeRecycled is NO), the @[...] literal cannot take a nil
element, dropping the leftBarButtonItems = nil resets is safe because iOS 16+
never writes that property now, and the nil→empty-array change only affects the
pre-iOS-16 branch, which is dead at the app's 18.0 deployment target.

Stale grouping when hidesSharedBackground toggles. The reuse cache is keyed
on [subview getUIBarButtonItem] object identity, but -setHidesSharedBackground:
mutates that cached item in place, so toggling the prop on a React header subview
no longer recomputes leading/trailingItemGroups. Previously any subview item
forced a rebuild every pass. Mechanism confirmed; not reachable today, because
Pylon only sets that prop on dict-based items, which are in the key.

Reuse key omits navitem.searchController. The searchBarPlacement branch
builds items from it, so a first pass taken before the search bar subview mounts
could be cached permanently. Newly reachable for screens with React header
subviews, but not exercised by current Pylon mobile code.

Both are latent upstream issues in vendored native code and neither is verifiable
without a device, so I have not touched them. They are worth re-checking during
the test-pylon-mobile pass this PR already needs, particularly if a header
search bar or a hidesSharedBackground toggle lands later.

[rynfar]

Rewrites the react-native-screens@4.25.2 patch so iOS header subviews are
collected into explicit left/right bar-button arrays instead of the navigation
item's items being cleared and rebuilt, which is what made the home header jump.

Adopted from T3 Code b982847ab201795711884d1dfadfddb3864adb23
(pingdotgg/t3code#8467), cherry-picked clean. Patch file plus lockfile only, no
Pylon source touched.

Scope note

This was originally scoped alongside upstream #8399 (Git action toast glass) as
one "mobile cosmetics" change. #8399 is not here because #138 already
contains it: #7327 rewrites GitActionProgressOverlay.tsx on top of #8399, and
that PR takes upstream's version of the file wholesale. Cherry-picking #8399
separately would have re-applied a superseded version.

Verification

• pnpm install succeeds, so the rewritten patch applies against
  react-native-screens@4.25.2
• Confirmed the patch actually landed rather than being silently skipped: the new
  subviewLeftItems symbol is present in the installed
  ios/RNSScreenStackHeaderConfig.mm
• pnpm-lock.yaml is unchanged after install, so the committed lockfile matches
  what pnpm resolves
• vp run -F @t3tools/mobile typecheck — clean; vp test run on mobile — 25 passed

Not verified

The patch is iOS native Objective-C. Nothing here proves the header actually
stops jumping — that needs a simulator or device run, which this change has not
had. Tests and typecheck can only show the patch is well-formed and the JS side
still resolves. Worth pairing with the test-pylon-mobile pass that #138 also
wants, since both land on the same surface.

Claude Opus 5 via Claude Code.

---

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

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.3 KiB	+47 B (+0.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	+1 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.4 KiB	+46 B (+0.7%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	55.6 KiB	+44 B (+0.1%)	66.4 KiB	✅
Codex	Live turn messages	9	10	+1 (+11.1%)	21	✅
Claude	Total thread wire	13.3 KiB	13.3 KiB	−3 B (−0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	+16 B (+0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.5 KiB	6.4 KiB	−19 B (−0.3%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	56.4 KiB	−44 B (−0.1%)	66.4 KiB	✅
Claude	Live turn messages	11	10	−1 (−9.1%)	21	✅

Baseline: 3d6035f · PR result: bb99f51 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[rynfar]

Independent review pass. Two findings, both latent, neither fixed.

The review verified more than my own pass could: it reverse-applied the previous
patch against pristine react-native-screens@4.25.2, applied the new one, and
read the resulting ios/RNSScreenStackHeaderConfig.mm in final form. It cleared
the things worth worrying about in a native patch — no retain cycle from
retaining config on the navigation item (screenView is weak and
RNSScreen.shouldBeRecycled is NO), the @[...] literal cannot take a nil
element, dropping the leftBarButtonItems = nil resets is safe because iOS 16+
never writes that property now, and the nil→empty-array change only affects the
pre-iOS-16 branch, which is dead at the app's 18.0 deployment target.

Stale grouping when hidesSharedBackground toggles. The reuse cache is keyed
on [subview getUIBarButtonItem] object identity, but -setHidesSharedBackground:
mutates that cached item in place, so toggling the prop on a React header subview
no longer recomputes leading/trailingItemGroups. Previously any subview item
forced a rebuild every pass. Mechanism confirmed; not reachable today, because
Pylon only sets that prop on dict-based items, which are in the key.

Reuse key omits navitem.searchController. The searchBarPlacement branch
builds items from it, so a first pass taken before the search bar subview mounts
could be cached permanently. Newly reachable for screens with React header
subviews, but not exercised by current Pylon mobile code.

Both are latent upstream issues in vendored native code and neither is verifiable
without a device, so I have not touched them. They are worth re-checking during
the test-pylon-mobile pass this PR already needs, particularly if a header
search bar or a hidesSharedBackground toggle lands later.

[rynfar]

Rewrites the react-native-screens@4.25.2 patch so iOS header subviews are
collected into explicit left/right bar-button arrays instead of the navigation
item's items being cleared and rebuilt, which is what made the home header jump.

Adopted from T3 Code b982847ab201795711884d1dfadfddb3864adb23
(pingdotgg/t3code#8467), cherry-picked clean. Patch file plus lockfile only, no
Pylon source touched.

Scope note

This was originally scoped alongside upstream #8399 (Git action toast glass) as
one "mobile cosmetics" change. #8399 is not here because #138 already
contains it: #7327 rewrites GitActionProgressOverlay.tsx on top of #8399, and
that PR takes upstream's version of the file wholesale. Cherry-picking #8399
separately would have re-applied a superseded version.

Verification

• pnpm install succeeds, so the rewritten patch applies against
  react-native-screens@4.25.2
• Confirmed the patch actually landed rather than being silently skipped: the new
  subviewLeftItems symbol is present in the installed
  ios/RNSScreenStackHeaderConfig.mm
• pnpm-lock.yaml is unchanged after install, so the committed lockfile matches
  what pnpm resolves
• vp run -F @t3tools/mobile typecheck — clean; vp test run on mobile — 25 passed

Not verified

The patch is iOS native Objective-C. Nothing here proves the header actually
stops jumping — that needs a simulator or device run, which this change has not
had. Tests and typecheck can only show the patch is well-formed and the JS side
still resolves. Worth pairing with the test-pylon-mobile pass that #138 also
wants, since both land on the same surface.

Claude Opus 5 via Claude Code.

---

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

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.3 KiB	+47 B (+0.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	+1 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.4 KiB	+46 B (+0.7%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	55.6 KiB	+44 B (+0.1%)	66.4 KiB	✅
Codex	Live turn messages	9	10	+1 (+11.1%)	21	✅
Claude	Total thread wire	13.3 KiB	13.3 KiB	−3 B (−0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	+16 B (+0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.5 KiB	6.4 KiB	−19 B (−0.3%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	56.4 KiB	−44 B (−0.1%)	66.4 KiB	✅
Claude	Live turn messages	11	10	−1 (−9.1%)	21	✅

Baseline: 3d6035f · PR result: bb99f51 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[rynfar]

Independent review pass. Two findings, both latent, neither fixed.

The review verified more than my own pass could: it reverse-applied the previous
patch against pristine react-native-screens@4.25.2, applied the new one, and
read the resulting ios/RNSScreenStackHeaderConfig.mm in final form. It cleared
the things worth worrying about in a native patch — no retain cycle from
retaining config on the navigation item (screenView is weak and
RNSScreen.shouldBeRecycled is NO), the @[...] literal cannot take a nil
element, dropping the leftBarButtonItems = nil resets is safe because iOS 16+
never writes that property now, and the nil→empty-array change only affects the
pre-iOS-16 branch, which is dead at the app's 18.0 deployment target.

Stale grouping when hidesSharedBackground toggles. The reuse cache is keyed
on [subview getUIBarButtonItem] object identity, but -setHidesSharedBackground:
mutates that cached item in place, so toggling the prop on a React header subview
no longer recomputes leading/trailingItemGroups. Previously any subview item
forced a rebuild every pass. Mechanism confirmed; not reachable today, because
Pylon only sets that prop on dict-based items, which are in the key.

Reuse key omits navitem.searchController. The searchBarPlacement branch
builds items from it, so a first pass taken before the search bar subview mounts
could be cached permanently. Newly reachable for screens with React header
subviews, but not exercised by current Pylon mobile code.

Both are latent upstream issues in vendored native code and neither is verifiable
without a device, so I have not touched them. They are worth re-checking during
the test-pylon-mobile pass this PR already needs, particularly if a header
search bar or a hidesSharedBackground toggle lands later.

[rynfar]

Rewrites the react-native-screens@4.25.2 patch so iOS header subviews are
collected into explicit left/right bar-button arrays instead of the navigation
item's items being cleared and rebuilt, which is what made the home header jump.

Adopted from T3 Code b982847ab201795711884d1dfadfddb3864adb23
(pingdotgg/t3code#8467), cherry-picked clean. Patch file plus lockfile only, no
Pylon source touched.

Scope note

This was originally scoped alongside upstream #8399 (Git action toast glass) as
one "mobile cosmetics" change. #8399 is not here because #138 already
contains it: #7327 rewrites GitActionProgressOverlay.tsx on top of #8399, and
that PR takes upstream's version of the file wholesale. Cherry-picking #8399
separately would have re-applied a superseded version.

Verification

• pnpm install succeeds, so the rewritten patch applies against
  react-native-screens@4.25.2
• Confirmed the patch actually landed rather than being silently skipped: the new
  subviewLeftItems symbol is present in the installed
  ios/RNSScreenStackHeaderConfig.mm
• pnpm-lock.yaml is unchanged after install, so the committed lockfile matches
  what pnpm resolves
• vp run -F @t3tools/mobile typecheck — clean; vp test run on mobile — 25 passed

Not verified

The patch is iOS native Objective-C. Nothing here proves the header actually
stops jumping — that needs a simulator or device run, which this change has not
had. Tests and typecheck can only show the patch is well-formed and the JS side
still resolves. Worth pairing with the test-pylon-mobile pass that #138 also
wants, since both land on the same surface.

Claude Opus 5 via Claude Code.

---

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

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.3 KiB	+47 B (+0.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	+1 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.4 KiB	+46 B (+0.7%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	55.6 KiB	+44 B (+0.1%)	66.4 KiB	✅
Codex	Live turn messages	9	10	+1 (+11.1%)	21	✅
Claude	Total thread wire	13.3 KiB	13.3 KiB	−3 B (−0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	+16 B (+0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.5 KiB	6.4 KiB	−19 B (−0.3%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	56.4 KiB	−44 B (−0.1%)	66.4 KiB	✅
Claude	Live turn messages	11	10	−1 (−9.1%)	21	✅

Baseline: 3d6035f · PR result: bb99f51 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[rynfar]

Independent review pass. Two findings, both latent, neither fixed.

The review verified more than my own pass could: it reverse-applied the previous
patch against pristine react-native-screens@4.25.2, applied the new one, and
read the resulting ios/RNSScreenStackHeaderConfig.mm in final form. It cleared
the things worth worrying about in a native patch — no retain cycle from
retaining config on the navigation item (screenView is weak and
RNSScreen.shouldBeRecycled is NO), the @[...] literal cannot take a nil
element, dropping the leftBarButtonItems = nil resets is safe because iOS 16+
never writes that property now, and the nil→empty-array change only affects the
pre-iOS-16 branch, which is dead at the app's 18.0 deployment target.

Stale grouping when hidesSharedBackground toggles. The reuse cache is keyed
on [subview getUIBarButtonItem] object identity, but -setHidesSharedBackground:
mutates that cached item in place, so toggling the prop on a React header subview
no longer recomputes leading/trailingItemGroups. Previously any subview item
forced a rebuild every pass. Mechanism confirmed; not reachable today, because
Pylon only sets that prop on dict-based items, which are in the key.

Reuse key omits navitem.searchController. The searchBarPlacement branch
builds items from it, so a first pass taken before the search bar subview mounts
could be cached permanently. Newly reachable for screens with React header
subviews, but not exercised by current Pylon mobile code.

Both are latent upstream issues in vendored native code and neither is verifiable
without a device, so I have not touched them. They are worth re-checking during
the test-pylon-mobile pass this PR already needs, particularly if a header
search bar or a hidesSharedBackground toggle lands later.

[rynfar]

Rewrites the react-native-screens@4.25.2 patch so iOS header subviews are
collected into explicit left/right bar-button arrays instead of the navigation
item's items being cleared and rebuilt, which is what made the home header jump.

Adopted from T3 Code b982847ab201795711884d1dfadfddb3864adb23
(pingdotgg/t3code#8467), cherry-picked clean. Patch file plus lockfile only, no
Pylon source touched.

Scope note

This was originally scoped alongside upstream #8399 (Git action toast glass) as
one "mobile cosmetics" change. #8399 is not here because #138 already
contains it: #7327 rewrites GitActionProgressOverlay.tsx on top of #8399, and
that PR takes upstream's version of the file wholesale. Cherry-picking #8399
separately would have re-applied a superseded version.

Verification

• pnpm install succeeds, so the rewritten patch applies against
  react-native-screens@4.25.2
• Confirmed the patch actually landed rather than being silently skipped: the new
  subviewLeftItems symbol is present in the installed
  ios/RNSScreenStackHeaderConfig.mm
• pnpm-lock.yaml is unchanged after install, so the committed lockfile matches
  what pnpm resolves
• vp run -F @t3tools/mobile typecheck — clean; vp test run on mobile — 25 passed

Not verified

The patch is iOS native Objective-C. Nothing here proves the header actually
stops jumping — that needs a simulator or device run, which this change has not
had. Tests and typecheck can only show the patch is well-formed and the JS side
still resolves. Worth pairing with the test-pylon-mobile pass that #138 also
wants, since both land on the same surface.

Claude Opus 5 via Claude Code.

---

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

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.3 KiB	+47 B (+0.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	+1 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.4 KiB	+46 B (+0.7%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	55.6 KiB	+44 B (+0.1%)	66.4 KiB	✅
Codex	Live turn messages	9	10	+1 (+11.1%)	21	✅
Claude	Total thread wire	13.3 KiB	13.3 KiB	−3 B (−0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	+16 B (+0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.5 KiB	6.4 KiB	−19 B (−0.3%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	56.4 KiB	−44 B (−0.1%)	66.4 KiB	✅
Claude	Live turn messages	11	10	−1 (−9.1%)	21	✅

Baseline: 3d6035f · PR result: bb99f51 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[rynfar]

Independent review pass. Two findings, both latent, neither fixed.

The review verified more than my own pass could: it reverse-applied the previous
patch against pristine react-native-screens@4.25.2, applied the new one, and
read the resulting ios/RNSScreenStackHeaderConfig.mm in final form. It cleared
the things worth worrying about in a native patch — no retain cycle from
retaining config on the navigation item (screenView is weak and
RNSScreen.shouldBeRecycled is NO), the @[...] literal cannot take a nil
element, dropping the leftBarButtonItems = nil resets is safe because iOS 16+
never writes that property now, and the nil→empty-array change only affects the
pre-iOS-16 branch, which is dead at the app's 18.0 deployment target.

Stale grouping when hidesSharedBackground toggles. The reuse cache is keyed
on [subview getUIBarButtonItem] object identity, but -setHidesSharedBackground:
mutates that cached item in place, so toggling the prop on a React header subview
no longer recomputes leading/trailingItemGroups. Previously any subview item
forced a rebuild every pass. Mechanism confirmed; not reachable today, because
Pylon only sets that prop on dict-based items, which are in the key.

Reuse key omits navitem.searchController. The searchBarPlacement branch
builds items from it, so a first pass taken before the search bar subview mounts
could be cached permanently. Newly reachable for screens with React header
subviews, but not exercised by current Pylon mobile code.

Both are latent upstream issues in vendored native code and neither is verifiable
without a device, so I have not touched them. They are worth re-checking during
the test-pylon-mobile pass this PR already needs, particularly if a header
search bar or a hidesSharedBackground toggle lands later.

[rynfar]

Rewrites the react-native-screens@4.25.2 patch so iOS header subviews are
collected into explicit left/right bar-button arrays instead of the navigation
item's items being cleared and rebuilt, which is what made the home header jump.

Adopted from T3 Code b982847ab201795711884d1dfadfddb3864adb23
(pingdotgg/t3code#8467), cherry-picked clean. Patch file plus lockfile only, no
Pylon source touched.

Scope note

This was originally scoped alongside upstream #8399 (Git action toast glass) as
one "mobile cosmetics" change. #8399 is not here because #138 already
contains it: #7327 rewrites GitActionProgressOverlay.tsx on top of #8399, and
that PR takes upstream's version of the file wholesale. Cherry-picking #8399
separately would have re-applied a superseded version.

Verification

• pnpm install succeeds, so the rewritten patch applies against
  react-native-screens@4.25.2
• Confirmed the patch actually landed rather than being silently skipped: the new
  subviewLeftItems symbol is present in the installed
  ios/RNSScreenStackHeaderConfig.mm
• pnpm-lock.yaml is unchanged after install, so the committed lockfile matches
  what pnpm resolves
• vp run -F @t3tools/mobile typecheck — clean; vp test run on mobile — 25 passed

Not verified

The patch is iOS native Objective-C. Nothing here proves the header actually
stops jumping — that needs a simulator or device run, which this change has not
had. Tests and typecheck can only show the patch is well-formed and the JS side
still resolves. Worth pairing with the test-pylon-mobile pass that #138 also
wants, since both land on the same surface.

Claude Opus 5 via Claude Code.

---

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

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.3 KiB	+47 B (+0.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	+1 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.4 KiB	+46 B (+0.7%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	55.6 KiB	+44 B (+0.1%)	66.4 KiB	✅
Codex	Live turn messages	9	10	+1 (+11.1%)	21	✅
Claude	Total thread wire	13.3 KiB	13.3 KiB	−3 B (−0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	+16 B (+0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.5 KiB	6.4 KiB	−19 B (−0.3%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	56.4 KiB	−44 B (−0.1%)	66.4 KiB	✅
Claude	Live turn messages	11	10	−1 (−9.1%)	21	✅

Baseline: 3d6035f · PR result: bb99f51 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[rynfar]

Independent review pass. Two findings, both latent, neither fixed.

The review verified more than my own pass could: it reverse-applied the previous
patch against pristine react-native-screens@4.25.2, applied the new one, and
read the resulting ios/RNSScreenStackHeaderConfig.mm in final form. It cleared
the things worth worrying about in a native patch — no retain cycle from
retaining config on the navigation item (screenView is weak and
RNSScreen.shouldBeRecycled is NO), the @[...] literal cannot take a nil
element, dropping the leftBarButtonItems = nil resets is safe because iOS 16+
never writes that property now, and the nil→empty-array change only affects the
pre-iOS-16 branch, which is dead at the app's 18.0 deployment target.

Stale grouping when hidesSharedBackground toggles. The reuse cache is keyed
on [subview getUIBarButtonItem] object identity, but -setHidesSharedBackground:
mutates that cached item in place, so toggling the prop on a React header subview
no longer recomputes leading/trailingItemGroups. Previously any subview item
forced a rebuild every pass. Mechanism confirmed; not reachable today, because
Pylon only sets that prop on dict-based items, which are in the key.

Reuse key omits navitem.searchController. The searchBarPlacement branch
builds items from it, so a first pass taken before the search bar subview mounts
could be cached permanently. Newly reachable for screens with React header
subviews, but not exercised by current Pylon mobile code.

Both are latent upstream issues in vendored native code and neither is verifiable
without a device, so I have not touched them. They are worth re-checking during
the test-pylon-mobile pass this PR already needs, particularly if a header
search bar or a hidesSharedBackground toggle lands later.

[rynfar]

Rewrites the react-native-screens@4.25.2 patch so iOS header subviews are
collected into explicit left/right bar-button arrays instead of the navigation
item's items being cleared and rebuilt, which is what made the home header jump.

Adopted from T3 Code b982847ab201795711884d1dfadfddb3864adb23
(pingdotgg/t3code#8467), cherry-picked clean. Patch file plus lockfile only, no
Pylon source touched.

Scope note

This was originally scoped alongside upstream #8399 (Git action toast glass) as
one "mobile cosmetics" change. #8399 is not here because #138 already
contains it: #7327 rewrites GitActionProgressOverlay.tsx on top of #8399, and
that PR takes upstream's version of the file wholesale. Cherry-picking #8399
separately would have re-applied a superseded version.

Verification

• pnpm install succeeds, so the rewritten patch applies against
  react-native-screens@4.25.2
• Confirmed the patch actually landed rather than being silently skipped: the new
  subviewLeftItems symbol is present in the installed
  ios/RNSScreenStackHeaderConfig.mm
• pnpm-lock.yaml is unchanged after install, so the committed lockfile matches
  what pnpm resolves
• vp run -F @t3tools/mobile typecheck — clean; vp test run on mobile — 25 passed

Not verified

The patch is iOS native Objective-C. Nothing here proves the header actually
stops jumping — that needs a simulator or device run, which this change has not
had. Tests and typecheck can only show the patch is well-formed and the JS side
still resolves. Worth pairing with the test-pylon-mobile pass that #138 also
wants, since both land on the same surface.

Claude Opus 5 via Claude Code.

---

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

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.3 KiB	+47 B (+0.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	+1 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.4 KiB	+46 B (+0.7%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	55.6 KiB	+44 B (+0.1%)	66.4 KiB	✅
Codex	Live turn messages	9	10	+1 (+11.1%)	21	✅
Claude	Total thread wire	13.3 KiB	13.3 KiB	−3 B (−0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	+16 B (+0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.5 KiB	6.4 KiB	−19 B (−0.3%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	56.4 KiB	−44 B (−0.1%)	66.4 KiB	✅
Claude	Live turn messages	11	10	−1 (−9.1%)	21	✅

Baseline: 3d6035f · PR result: bb99f51 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[rynfar]

Independent review pass. Two findings, both latent, neither fixed.

The review verified more than my own pass could: it reverse-applied the previous
patch against pristine react-native-screens@4.25.2, applied the new one, and
read the resulting ios/RNSScreenStackHeaderConfig.mm in final form. It cleared
the things worth worrying about in a native patch — no retain cycle from
retaining config on the navigation item (screenView is weak and
RNSScreen.shouldBeRecycled is NO), the @[...] literal cannot take a nil
element, dropping the leftBarButtonItems = nil resets is safe because iOS 16+
never writes that property now, and the nil→empty-array change only affects the
pre-iOS-16 branch, which is dead at the app's 18.0 deployment target.

Stale grouping when hidesSharedBackground toggles. The reuse cache is keyed
on [subview getUIBarButtonItem] object identity, but -setHidesSharedBackground:
mutates that cached item in place, so toggling the prop on a React header subview
no longer recomputes leading/trailingItemGroups. Previously any subview item
forced a rebuild every pass. Mechanism confirmed; not reachable today, because
Pylon only sets that prop on dict-based items, which are in the key.

Reuse key omits navitem.searchController. The searchBarPlacement branch
builds items from it, so a first pass taken before the search bar subview mounts
could be cached permanently. Newly reachable for screens with React header
subviews, but not exercised by current Pylon mobile code.

Both are latent upstream issues in vendored native code and neither is verifiable
without a device, so I have not touched them. They are worth re-checking during
the test-pylon-mobile pass this PR already needs, particularly if a header
search bar or a hidesSharedBackground toggle lands later.