Repository navigation
[.NET 11 Failure] Fix NavigationPage and Shell: Respect back-button handling for interactive pop gestures - #39746
Open
Dhivya-SF4094 wants to merge 4 commits into
Open
Dhivya-SF4094 wants to merge 4 commits into
Dhivya-SF4094 wants to merge 4 commits into
Conversation
Dhivya-SF4094
temporarily deployed
to
copilot-pat-pool
October 29, 2026 05:58 — with 
GitHub Actions
Inactive
GitHub Actions
Inactive
Contributor
|
🚀 Dogfood this PR with:
curl -fsSL https://raw.githubusercontent.com/dotnet/maui/main/eng/scripts/get-maui-pr.sh | bash -s -- 39746Or
iex "& { $(irm https://raw.githubusercontent.com/dotnet/maui/main/eng/scripts/get-maui-pr.ps1) } 39746" |
Dhivya-SF4094
temporarily deployed
to
copilot-pat-pool
October 29, 2026 05:58 — with
GitHub Actions
Inactive
GitHub Actions
Inactive
This comment has been minimized.
This comment has been minimized.
dotnet-policy-service
Bot
added
the
partner/syncfusion
Dhivya-SF4094
temporarily deployed
to
copilot-pat-pool
October 29, 2026 05:59 — with
GitHub Actions
Inactive
GitHub Actions
Inactive
Dhivya-SF4094
temporarily deployed
to
copilot-pat-pool
October 29, 2026 06:01 — with
GitHub Actions
Inactive
GitHub Actions
Inactive
Tamilarasan-Paranthaman
added
area-navigation
Dhivya-SF4094
temporarily deployed
to
copilot-pat-pool
October 29, 2026 06:03 — with
GitHub Actions
Inactive
GitHub Actions
Inactive
Dhivya-SF4094
temporarily deployed
to
copilot-pat-pool
October 29, 2026 06:03 — with
GitHub Actions
Inactive
GitHub Actions
Inactive
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
github-actions
Bot
added
the
s/agent-review-in-progress
This comment has been minimized.
This comment has been minimized.
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
This comment has been minimized.
This comment has been minimized.
MauiBot
removed
the
s/agent-review-in-progress
This was referenced Oct 29, 2026
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
github-actions
Bot
added
the
s/agent-review-in-progress
This was referenced Oct 29, 2026
Open
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
github-actions
Bot
added
the
s/agent-review-in-progress
Collaborator
PR Test Gate
🧪 Gate analysis — click to expand⚠ Inconclusive. Setup, build, or execution did not produce a conclusive result. This does not prove the fix is incorrect. 📱 Device tests
Execution log excerptsWithout fix With fix Without fix With fix PR head: 09afac82. Without-fix baseline: 7de2a9f2. |
MauiBot
removed
the
s/agent-review-in-progress
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
vishnumenon2684
changed the title
sheiksyedm
marked this pull request as ready for review
October 30, 2026 15:05
Contributor
There was a problem hiding this comment.
🟡 Changes recommended
The navigation policy can admit a UIKit transition after OnBackButtonPressed disposes the manager, requiring a post-policy disposal check and a NavigationPage regression test.
1 open issue
What changed in this PR
This PR applies the back-button handling fix from #38071 to the .NET 11 NavigationPage and Shell handlers on iOS and Mac Catalyst, using their shared navigation controller manager.
Changes:
- Centralizes edge and iOS 26 content-pop gesture admission, consulting managed back-button handling before allowing navigation.
- Caches decisions per touch and tracks pending gestures to avoid duplicate back callbacks.
- Adds and updates device tests for handled back actions, cancellation, handler disconnection, and shared gesture decisions.
| File | Description |
|---|---|
| src/Core/src/Platform/iOS/NavigationControllerManager/NavigationControllerManager.cs | Centralizes gesture admission, touch caching, and pending-state cleanup. |
| src/Core/src/Platform/iOS/NavigationControllerManager/INavigationManagerDelegate.cs | Adds a dedicated interactive-pop policy callback. |
| src/Core/src/Handlers/NavigationPage/NavigationViewHandler.iOS.cs | Routes gesture admission through NavigationPage back handling. |
| src/Controls/tests/DeviceTests/Elements/Shell/ShellTests.iOS.cs | Adds Shell back-navigation and gesture-lifecycle regression tests. |
| src/Controls/tests/DeviceTests/Elements/NavigationPage/NavigationPageTests.iOS.cs | Updates handler-based tests and verifies back handling after cancellation. |
| src/Controls/src/Core/Handlers/Shell/ShellSectionHandler.iOS.cs | Unifies Shell back policy and adopts shared gesture delegation. |
Review effort: Balanced
Comment on lines
+533
to
+535
| _shouldBegin ??= manager._delegate.ShouldBeginInteractivePop(); | ||
| manager._interactivePopGesturePending = _shouldBegin.Value; | ||
| return manager._interactivePopGesturePending; |
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
github-actions
Bot
added
the
s/agent-review-in-progress
|
Azure Pipelines will not run the associated pipelines, because the pull request was updated after the run command was issued. Review the pull request again and issue a new run command. |
Collaborator
PR Test Gate
🧪 Gate analysis — click to expand⚠ Inconclusive. The PR could not be merged with its target branch; no tests ran. No interpretable test-group summary was produced. No additional test execution is inferred. Setup did not capture a verified PR head; no commit is claimed as tested. No immutable without-fix baseline was prepared. |
MauiBot
removed
the
s/agent-review-in-progress
kubaflo
added a commit
that referenced
this pull request
Oct 30, 2026
Reuse explicit NavigationRenderer registration for renderer-disposal continuations. Remove concrete ShellRenderer callback assumptions from three Android icon tests, reusing the native Shell fixture's toolbar lookup through a virtual hook. Preserve all assertions and exercise both real implementations. Android native controls: fixed 79/79, original concrete callbacks 76/79 with the same three failures, restored 79/79. iPhone simulator disposal cases pass 2/2, revert to the same two InvalidCastException failures with original registration, and pass 2/2 after restoration. Four separate gesture cases remain pending the existing upstream #39746 implementation; no duplicate is included. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com> Copilot-Session: d595bf0b-ce58-4118-be08-7fe144955a64
This comment has been minimized.
This comment has been minimized.
|
Azure Pipelines will not run the associated pipelines, because the pull request was updated after the run command was issued. Review the pull request again and issue a new run command. |
This was referenced Oct 30, 2026
kubaflo
added a commit
that referenced
this pull request
Oct 30, 2026
Merge original #39746 head09afac82be8c into the validation branch, retaining all original commits and authorship rather than copying or duplicating its implementation. Resolve the NavigationPageTests conflict by keeping both the renderer-disposal tests/helper and the original native shouldPopItem helper. No GitHub/protected-branch pull request is merged by this operation. Native iPhone11Pro iOS26.5 validation:540/540 tests pass across both NavigationPage and both Shell variants, including the four previously failing handler gesture cases. Integrated Apple Core and shared MacCatalyst Controls compile with zero warnings/errors. Existing local Xcode27.1 override and wrapper result-filename limitation remain documented; actual native XML identities and outcomes were verified. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com> Copilot-Session: d595bf0b-ce58-4118-be08-7fe144955a64
kubaflo
added a commit
that referenced
this pull request
Nov 1, 2026
<!-- Please let the below note in for people that find this PR -->
> [!NOTE]
> Are you waiting for the changes in this PR to be merged?
> It would be very helpful if you could [test the resulting
artifacts](https://github.com/dotnet/maui/wiki/Testing-PR-Builds) from
this PR and let us know in a comment if this change resolves your issue.
Thank you!
<!--
!!!!!!! MAIN IS THE ONLY ACTIVE BRANCH. MAKE SURE THIS PR IS TARGETING
MAIN. !!!!!!!
-->
### Description of Change
<!-- Enter description of the fix in this section -->
This prerequisite intentionally targets protected `net11.0`, so #39688
inherits corrections through reviewed integration rather than bypassing
branch protection. **Current publication status: full current-head CI is
not yet all-three-pipeline green.** Both owned correction refs remain at
actual #39779 published head `691808743d1b3915c9af359c9a37dae0588f51c2`;
GitHub REST confirms that head and the parent checkout is clean.
Exact-source
[device1657679](https://dev.azure.com/dnceng-public/public/_build/results?buildId=1657679)
**completed succeeded all9/9 jobs** at `2026-10-31T12:28:45.1319334Z`.
Exact-source
[maui-pr1657678](https://dev.azure.com/dnceng-public/public/_build/results?buildId=1657678)
remains running with29/31 jobs passed;
[UI1657680](https://dev.azure.com/dnceng-public/public/_build/results?buildId=1657680)
remains running with genuine current iOS badge failures, a distinct
Catalyst DNS setup failure, and other work still running. The fresh
current native audit has61 remaining app XMLs pending review: all9 Helix
job `Finished` fields are nonnull, all71 work items are finished with
exit0, and the62-XML catalog exists, but no parsed current pass
certification is claimed yet; the complete f26 all62 audit below remains
older-head evidence only. Azure `triggerInfo.pr.sourceSha` is `691`,
while `92a13b…` is the merge SHA, not the source head. Protected
`net11.0` and #39688 remain at
`c13dddb0a6b5b57a31a7fed4d4779dc01b31a9ef`; no bypass, force-push,
approval, or merge occurred. #39688 receives these fixes only after
#39779 passes required checks and is merged with explicit user
authorization.
**Current `691` iOS Mono same-code recovery receipt:** device1657679
OS26 Mono General completed1292=1278 passed+14 existing skips+0
failures/errors, with all declarations reconciled. Exact
`MemoryTests_TabbedViewHandler.WindowDoesNotLeak("Window Does Not
Leak")` passed in0.2446597s on `dci-mac-build-475`; receipt: parent
`files/691-ios26-general-results.xml`. The current Window test, helper,
and40-GC behavior are unchanged from historical `d673`, where the same
case failed in1.1958577s, and from the passing f26/local cohorts. This
fresh same-code cohort recovery is not a root-cause fix, certified undo
experiment, product patch, or evidence for changing the fixture, helper,
GC count, wait budget, or skip. Native iOS Mono raw138 succeeded with0
errors; OS26 job `ea1e82de…` and OS15 job `54d03775…` both have8
finished work items. The other15 app XMLs remain pending audit, so no
complete current native-report certification is claimed.
**Current `691` Catalyst UI setup receipt:** UI1657680 Catalyst Shell1
task `bcc8dcd8-bcf8-5783-51a6-38f65fe366c5` / raw2612 failed before
tests while downloading the MSBuild logger ZIP from
`vstsagenttools.blob.core.windows.net`: DNS reported `nodename nor
servname provided, or not known`; telemetry to `dev.azure.com` also
recorded DNS failure. Both attempts produced no readable NUnit XML or
TRX, so the native Shell1 cases are **unmeasured**, not four badge-test
failures or evidence of a native product bug or global outage. The
Catalyst UI stage remains running and no stage2 retry has been admitted;
classification waits for stage completion and all failures. No network,
runtime, source, URI, timeout, budget, or configuration change is
claimed.
**Current `691` iOS badge visual-failure receipt:** UI1657680 iOS
CoreCLR latest Shell1 raw4524 and publisher4526 completed initial173=163
passed+6 existing skips+4 failures; the exact four-case targeted NUnit
retry completed4 failures+0 passes. Run46147754 reports result100169
`ShellBadgeCanBeCleared` at2.30%, result100170 `InitialBadgeIsVisible`
at2.31%, result100171 `SetAtRuntime` at2.31%, and result100172
`MultipleTabsCanHaveBadges` at2.31%; all four retained the same
percentage across both attempts and the publisher recorded four real
failures. Parent receipts: `files/691-ios-shell1-failed4524.log` and
`files/691-ios-shell1-publisher4526.log`. This is a current native iOS
badge visual mismatch, distinct from Catalyst Shell1's DNS pre-test
failure with no XML. No new expected colors, precise underlying
mechanism, golden change, mask, tolerance, or skip is claimed before
current-image inspection.
**Historical superseded `d673` native/UI receipt:** on device1657603 iOS
Mono raw141, OS26 General completed1292=1277 passed+14 existing skips+1
failure: `MemoryTests_TabbedViewHandler.WindowDoesNotLeak`, with the
Window alive at1.1958577s. The same exact OS26 case passed on `f26`
in0.1247654s. The source creates `ContentPage`, not `TabbedPage` (the
latter appears only through inherited registration), so no Tabbed
product root, infrastructure cause, isolated local red,
runtime/SDK/source fix, helper change,40-GC change, skip, or wait-budget
change is claimed. A fresh local original-fixture iOS Mono Core
simulator OS26.5 XML completed65=64 passed+1 existing skip, with the
exact Window case passing in0.2360164s; the wrapper exited1 only because
it expected `testResults.xml` while the verified fresh output was
`xunit-test-*.xml`. An initial build without the existing CI opt-in
stopped at NETSDK1242/no tests and is not red; the successful run used
the existing `UseMonoRuntime=true` and
`_DisableCheckForUnsupportedMonoMobileRuntime=true` settings from
`stage-device-tests.yml`. Current `691` native CI is rerunning the case.
Historical d673 main1657602 was canceled after the new-head push, after
only partial progress (seven base jobs while integration work was still
running); no all-green claim is made. The old observer exited on a
JSON-read error while fetching the canceled timeline, which is not a
test failure.
**Historical superseded `d673` Windows focus/color receipt:** Windows
Shell2 run46144002 completed197=183 passed+13 existing skips+1 failure.
Exact result100185 `CancelButtonColor` passed its strict `Focused`
precondition and unchanged new golden, providing native acceptance for
that focus/color case only. Sole failure result100195 `Uppercase` failed
both attempts visually at0.70%; actual attachment17 (`01e09879…`,22830
bytes) and AX (`3d424a28…`) prove `HasKeyboardFocus=True`,
`FocusStatus=Focused`, and exactly195 RGB changes in bbox
`(168,266)-(217,276)`, confined to the event label versus an empty-label
golden, with zero changes outside it—including all uppercase glyphs and
colors. The `691` update makes the strict `Focused.After(2000,100)`
assertion unconditional inside the existing Windows-only outer guard and
updates only those195 label pixels; it changes no scope, skip,
tolerance, mask, or uppercase visual requirement. The native PNG is
byte-identical to the captured actual. Managed WinUI test build passed
before publication with0 warnings/errors in23.75s and after publication
via `./.dotnet/dotnet build
src/Controls/tests/TestCases.WinUI.Tests/Controls.TestCases.WinUI.Tests.csproj
-c Release --no-restore -v:q` with0 warnings/errors in18.26s; original
PNG byte comparison, scoped formatting, and `git diff --check` passed,
with three unrelated formatter whitespace changes restored. Actual
execution of the new `691` assertion remains pending fresh CI; this is
not full UI certification. Historical `d673` iOS Shell1
raw4295/publisher4310/run46144012 still has the same four badge failures
(173=163 passed+6 skips+4 failures; targeted retry4/4 failed), with no
safe badge fix or golden change.
**Historical superseded `cd` main-gate receipt:** maui-pr1657489
completed succeeded at `2026-10-31T09:39:23.6411636Z`. The fresh
complete Azure timeline contains1233 records and verifies31/31 jobs
succeeded, with no skipped or failed jobs. The run is qualified to exact
historical counterpart source
`cd7a8cb8a21c8d798c9588da7ec86a3ae54379df`; the original source Windows
Pack raw267 records190 warnings/0 errors. This closed only the main gate
on that historical head, **not all-three-pipeline promotion acceptance
or current `691` certification**. Earlier main-running/29-or30-of31
statements below are preserved historical checkpoints superseded by this
complete receipt and the current publication status above. Historical
`cd` devices failed7/9; historical `cd` UI had genuine failures.
Combined `79` manual1657568 and `f26` manual1657552 remain historical
scoped evidence, not current certification.
#### Current-head device failures
Device1657491 Android Mono
[raw149](https://dev.azure.com/dnceng-public/public/_apis/build/builds/1657491/logs/149),
Helix job prefix `76c02089`, contains actual Core XML3090=3017 passed+72
existing skips+1 failure. Exact
`AssertionExtensionsTests.EarlierReferenceCollectedDuringLaterWait(useWaitForGC:
True)` failed in4.999277s with `System.Object` retained. The other four
actual app XML reports are Controls1345=1332 passed+13 existing skips,
Essentials424/424 passed, Graphics33/33 passed, and Blazor68=64 passed+4
existing skips. Parent receipt: `files/cd-device-android-mono-149.log`
plus native XML.
Windows
[raw165](https://dev.azure.com/dnceng-public/public/_apis/build/builds/1657491/logs/165),
Helix job prefix `5ded47b7`, records pre-test `Add-AppxPackage` failures
for Controls-packaged `a000OR5` and Core-packaged `a000ORS`:
`0x80073CF0`, missing `AppxManifest`/`0x80070003`, raw console
lines59-60. These work items have no XML: their native cases are
**unmeasured**, not product-test failures or proof of an infrastructure
outage. Parent receipt: `files/cd-device-windows-165.log`.
The parent independently matched and ported already-validated isolated
#39529 commit prefix `f11b88966`, identical fixture blob
`132ef9846b52df6bf59b7fb29c1317a5c24d31cc`: one fixture file,25 added/13
removed lines, with a stronger real-later-reference-alive assertion.
There are no helper,40-GC, case or runtime changes. After validation,
this port was committed as `f26c935851d398ea5cb4e4415fbe0c2f0fa6e4e4`
and normally pushed **only to parent branch
`kubaflo-net11-promotion-compile-fixes`**. The #39779 counterpart
remains exactly `cd` to preserve ongoing native UI receipts; it has not
inherited this port. Historical `f8` successes below remain separate.
#### Parent-only GC port and manual validation scope
Native full Core Android Mono AOT rebuild after the port passed in96.94s
with6 existing warnings/0 errors, producing native APK ABIs. The
existing runner temporarily injected only the exact-class module
initializer; no injection was committed. Three fresh native Mono AOT
ARM64/API34 processes each passed the same9/9 cases with0
failures/skips, including both regression variants and seven safety
cases; these are three processes, not27 distinct tests. A fourth
post-push process passed the same9/9. The old ARM64/API34 AOT fixture
also passed9/9 in one explicit-instrumentation process, so **no local
baseline-red or isolated local causality is claimed**. The actual
published `cd` x64 Mono True-variant CI failure above and original
#39529 causal controls support the port; ARM64 local preservation does
not certify x64 or current PR CI.
Exact commands/results supplied by the parent:
```sh
PATH="$PWD/.dotnet:$PATH" DOTNET_ROOT="$PWD/.dotnet" UseMonoRuntime=true _DisableCheckForUnsupportedMonoMobileRuntime=true pwsh .github/skills/run-device-tests/scripts/Run-DeviceTests.ps1 -Project Core -Platform android -IncludeClasses Microsoft.Maui.DeviceTests.AssertionExtensionsTests -TestFilter Category=Memory -DeviceUdid emulator-5554 -Rebuild -BuildOnly -OutputDirectory /Users/kubaflo/.copilot/session-state/d595bf0b-ce58-4118-be08-7fe144955a64/files/gc-upstream-fixed-build
```
Passed full native Core rebuild96.94s,6 existing warnings/0 errors, Mono
AOT/native APK. Baseline used the identical build command without
`-BuildOnly`, with output directory `files/cd-gc-baseline`: build
passed, but the default runtime invocation printed Android AM
help/exit82 and produced no XML/no tests. That invocation is not native
acceptance, a crash or causal red.
```sh
for repeat in 1 2 3; do
./.dotnet/dotnet /Users/kubaflo/.nuget/packages/microsoft.dotnet.xharness.cli/11.0.0-prerelease.26517.1/tools/net11.0/any/Microsoft.DotNet.XHarness.CLI.dll android test --app artifacts/bin/Core.DeviceTests/Release/net11.0-android/com.microsoft.maui.core.devicetests-Signed.apk --package-name com.microsoft.maui.core.devicetests --instrumentation com.microsoft.maui.core.devicetests.TestInstrumentation --device-id emulator-5554 --output-directory "/Users/kubaflo/.copilot/session-state/d595bf0b-ce58-4118-be08-7fe144955a64/files/gc-fixed-$repeat" --timeout 00:05:00 --arg TestFilter=Category=Memory --arg "results-file-name=gc-fixed-$repeat.xml"
done
```
Each XHarness invocation exited0; each actual XML reports9/9 passed,0
failures/skips, exclusively the exact class. Post-push used the
identical single XHarness command without the loop, output directory
`files/gc-fixed-postpush` and `--arg
results-file-name=gc-fixed-postpush.xml`: exit0, actual XML9/9 passed.
The baseline explicit-instrumentation single command used
`files/cd-gc-baseline-explicit` and `--arg
results-file-name=cd-gc-baseline-explicit.xml`: old ARM64/API34 AOT
fixture9/9 passed, not local red.
- `./.dotnet/dotnet format Microsoft.Maui.sln --no-restore --include
src/Core/tests/DeviceTests/AssertionExtensionsTests.cs --exclude
Templates/src --exclude-diagnostics CA1822 --verbosity quiet` — passed.
- `git diff --check` — passed.
- `git hash-object
src/Core/tests/DeviceTests/AssertionExtensionsTests.cs` — returned
exactly `132ef9846b52df6bf59b7fb29c1317a5c24d31cc`, matching the
original upstream `f11` source blob.
- `az pipelines run --org https://dev.azure.com/dnceng-public --project
public --id 314 --branch
refs/heads/kubaflo-net11-promotion-compile-fixes` — admitted [manual
device1657552](https://dev.azure.com/dnceng-public/public/_build/results?buildId=1657552),
reason `manual`, exact
`sourceVersion=f26c935851d398ea5cb4e4415fbe0c2f0fa6e4e4`. Admission is
not a full native pass and is **not #39779 current CI**.
Only that fixture file was committed and parent-only pushed. Shared
helpers,40-GC behavior, cases, runtime, skips, production and APIs are
unchanged. Current `cd` device1657491 remains failed7/9 jobs; main/UI
remain running. No counterpart push or current-CI fix is claimed.
#### Latest restore serialization and Windows focused-state oracle
The `565` correction adds only seven lines in `eng/Build.props`, setting
default `ProjectToBuild.RestoreInParallel=false`. Original `f8` Pack
Windows build1657400/raw262/binlog prefix `8a3d7198` demonstrated
concurrent `BuildTasksOnly` portable/full Package restores overwriting
shared assets evaluated for different target frameworks. This serializes
those restores without changing the existing build-task/full ordering.
Native isolated build1657433 Pack Windows/raw267 **succeeded with190
warnings/0 errors**; after-binlog SHA
`133fbcbb3eac09bf0a97de3891791e5e2cbffcecd57d4c30bb322d5cff6294c5`,
ExecuteMSBuild task78, confirms `RestoreBuildInParallel=False`. Pack Mac
also succeeded. No pipeline tests are added and no speculative
stream-owner/locking patch is included.
**Current-head native packing acceptance:** on actual PR build1657489,
both Pack Windows and Pack macOS succeeded. The parent verified
`triggerInfo.pr.sourceSha=cd7a8cb8a21c8d798c9588da7ec86a3ae54379df`
using `az`. [Windows raw
log267](https://dev.azure.com/dnceng-public/public/_apis/build/builds/1657489/logs/267)
records `Build succeeded` at line2192,190 warnings at line2405 and0
errors at line2406, with packages published and `Finishing: Pack &
Publish`; receipt: parent `files/cd-native-windows-pack-267.log`. This
validates serialized restore on the actual current `cd` PR head, not
only the earlier isolated `565` manual run. Overall
main1657489/UI1657492 remain running and device1657491 completed
failed7/9 jobs; this is not an all-green claim. The new Windows
focused-state oracle now fails before the screenshot, as recorded below;
there is no native UI acceptance.
`cd7a8cb8a21` changes only the Windows CancelButtonColor test and its
native golden. All195 exact RGB differences between the old expectation
and native capture are confined to `FocusStatus`, formerly empty versus
source-required `Focused`; there are **zero differences outside that
label**, and the orange cancel-button rectangle is pixel-identical.
Native AX reports `HasKeyboardFocus=True`; the source Focused handler
writes `Focused`, and the test taps, clears and types `Testing`. A
Windows-only delayed assertion requires that label before the screenshot
(2s/100ms), then only those195 label pixels are reconciled. The
unchanged actual local comparer reported0.70% against the old image
while CI reported0.76%; these are not identical metrics. The new image
matches locally. No cancel-color production import, color change,
threshold relaxation or skip is included. Managed Windows NUnit test
compilation is not Windows UI runtime acceptance.
**Current `cd` UI failure receipts:** Windows Shell2
[raw4404](https://dev.azure.com/dnceng-public/public/_apis/build/builds/1657492/logs/4404),
publisher4408/run46141148, exact result100193, fails the strict
`Focused` assertion with an empty label in both initial and retry
attempts **before the screenshot**. Complete197 results=183 passed+13
existing NotExecuted+1 failure; every other ShellSearch case passes,
including Uppercase. Fresh AX attachment3, prefix `caf50e45`,22176
bytes, shows Edit `AutomationId=TextBox`, `HasKeyboardFocus=True`, while
the `FocusStatus` TextBlock Name is empty. Thus there is no native
cancel-orange or new-golden acceptance: this is
focus-ownership/event-metadata divergence, not a visual-color fix. It
matches the peer `cc6248` line-owner-clear/new-focus-sync mechanism, but
that implementation is **not imported**; original corrected
native-fixture SHAs are awaited.
Latest iOS26.5 CoreCLR Shell1
[raw4463](https://dev.azure.com/dnceng-public/public/_apis/build/builds/1657492/logs/4463),
publisher4467/run46141232, has initial173 results=163 passed+6 existing
skips+4 same badge failures. All four failures were selected for
targeted retry; no retry acceptance is claimed. Unsupported badge probes
remain rejected, with no safe fix or badge-golden change. Overall
main/UI remain running and current `cd` devices remain failed7/9 jobs.
Latest exact commands/results, supplied by the parent session:
- `./build.sh -restore -ci -configuration Release -warnAsError false
/p:SkipInitInternalTooling=true /p:ValidateXcodeVersion=false
/bl:/Users/kubaflo/.copilot/session-state/d595bf0b-ce58-4118-be08-7fe144955a64/files/restore-race-after.binlog`
— passed after `565`,0 warnings/0 errors,5.97s.
- `./build.sh -restore -ci -configuration Release -warnAsError false
/p:SkipInitInternalTooling=true /p:ValidateXcodeVersion=false
/bl:/Users/kubaflo/.copilot/session-state/d595bf0b-ce58-4118-be08-7fe144955a64/files/restore-race-postpush.binlog`
— passed after publishing `565`,0 warnings/0 errors,5.78s.
- `./build.sh -restore -ci -configuration Release -warnAsError false
/p:SkipInitInternalTooling=true /p:ValidateXcodeVersion=false` — passed
after publishing `cd`,0 warnings/0 errors,7.83s.
- `./.dotnet/dotnet build
src/Controls/tests/TestCases.WinUI.Tests/Controls.TestCases.WinUI.Tests.csproj
-c Release -v:q` — passed before `cd`,0 warnings/0 errors,37.17s.
- `./.dotnet/dotnet build
src/Controls/tests/TestCases.WinUI.Tests/Controls.TestCases.WinUI.Tests.csproj
-c Release -v:q --no-restore` — passed after publishing `cd`,0
warnings/0 errors,22.04s. These are managed test-compilation results,
not Windows-native UI runs.
- `./.dotnet/dotnet format whitespace Microsoft.Maui.sln --no-restore
--include
src/Controls/tests/TestCases.Shared.Tests/Tests/FeatureMatrix/ShellSearchHandlerFeatureTests.cs
--verbosity quiet` — passed. Three unrelated trailing-whitespace edits
were restored; final C# diff is only five added lines.
#### Latest complete earlier-head native evidence
Native `f8` device1657401 completed **all9/9 Helix jobs**, with62
complete XML reports:46,202 tests=44,980 passed+1,222 existing
exclusions,0 failures/errors. This is earlier-head evidence, not
certification of `cd`.
Native `f8` Mac Entry raw4511/publisher4517/run46138420 completed121=118
passed+3 existing exclusions,0 failures/errors; all four previously
failing targets passed. This replaces the earlier pending native Mac
acceptance status below. Native Mac CollectionView1
raw4182/publisher4185/run46137950 completed69=68 passed+1 existing
exclusion,0 failures/errors. The black-capture cases recovered in a
fresh-worker cohort; no precise cause or source black-capture fix is
claimed.
Windows CV1 raw4562/publisher4564/run46138612 completed355 final
results=341 passed+13 existing exclusions+1 real Horizontal Issue38059
failure; nine setup cases recovered. All13 updated expectations and
Issue36422's post-screenshot position assertion passed. Horizontal
layout still shows the wrong two vertical columns instead of the
required horizontal span4; its golden is not refreshed and no Windows
production layout patch is included.
`f8` Shell2 Windows run46138102 failed CancelButtonColor at0.76% in both
attempts before the focused-label correction. `f8` iOS badge tests
failed in both attempts with2.30%/2.31% differences. The source-required
blue/green badge still renders red through the native API; the
limitation remains unresolved, and all unsafe probes were removed.
Initial `f8` main1657400 Pack failed as described above. Its Helix
Resizetizer `DebugColorizedWithNamed` DeleteTMP IO failure had native796
results=793 passed+2 existing exclusions+1 failure. Debug monitor303
resubmitted only the failed unchanged work item as
`6e8fd615-2895-4b7e-8c4c-cb58d1529722`; the monitor passed with0 failed
and uploaded339 logical results, **not a native result count**. The
parent has now verified native retry receipt
`files/f8-resizetizer-debug-retry-results.xml`:796=794 passed+2 existing
skips,0 failures/errors. Exact SVG test type
`Microsoft.Maui.Resizetizer.Tests.SkiaSharpSvgToolsTests+Resize`, method
`ColorizedWithNamedReturnsColored`, passed in0.3028857s; this is not the
same-named raster test. Source log303 confirms only one failed work item
was resubmitted with unchanged payload. This is historical `f8` native
recovery, not current `cd` CI certification. No IO-lock source/infra
attribution or full current-head native pass is claimed.
The sections below retain the detailed earlier implementation and
validation history. Historical pending/in-progress statuses are
superseded only by the exact newer results above; absent, partial and
older-head measurements remain distinct.
#### Build and fixture corrections
- Use the existing `barBackground` Paint in the iOS 26 tab-bar layout
guard instead of nonexistent `barBackgroundColor`.
- Qualify WinUI alignment enums and `Thickness` in
`NativeIconBadgeState`. Extend the existing, narrowly scoped `RS0030`
exception to native-grid badge removal as well as insertion; these are
WinUI `Grid`, not `MauiPanel`, so `CachedChildren` is unavailable. The
analyzer stays enabled elsewhere.
- Reuse explicit `NavigationRenderer` registration for the
renderer-disposal continuation test, including inherited fixtures.
- Remove concrete `ShellRenderer` callback assumptions from three
Android icon tests. A virtual toolbar hook reuses native Shell's
hierarchy traversal, preserving the original color/aspect-ratio/drawing
assertions.
Raw `az` logs reproduce compile failures in1656326,1656328
and1656327/1656398, then native-grid analyzer errors in1656491/1656492
and inherited-fixture failures in1656522.
#### Reuse upstream implementations rather than duplicate them
Merge `597bb90b38e` integrates original #39746 head `09afac82be8`,
retaining its four commits and authorship. Its parents are `1382f5f3980`
and `09afac82be8`. The sole conflict was independent additions in
`NavigationPageTests.iOS.cs`; both the renderer-disposal tests/helper
and original native `shouldPopItem` interop helper were retained.
Navigation admission, cancellation, back-command and callback-count
assertions are preserved. This is a feature-branch integration, **not a
protected GitHub PR merge**.
Merge `43c74d473f1` retains original #39761 head `2583b4a48ee`. Default
appearance is captured before renderer initialization applies custom
backgrounds. Lazy capture in both public `SetAppearance` methods is also
preserved: removing that fallback broke8/16 unchanged native tests for
direct tracker callers. With it, all16 pass and the combined
navigation/Shell/appearance set passes556/556. The original two iOS
reset expectations were exercised by the gradient-to-solid-to-default UI
case.
#### Native tint regression and case-specific visual reconciliation
`1382f5f3980` reapplies existing per-item icon-color logic after
bottom-menu reconstruction. Previously, rebuilding tabs left the
selected label magenta but its font icon dark, and explicitly colored
font icons lost their color. Two native regressions cover
selected/unselected tint, explicit source colors, rebuilding and
subsequent selection. Native Shell exercises the shared manager too.
Visual changes have **no tolerance relaxation, dropped assertions or new
skips**:
- RTL Shell: exact net11 PNG restored from `7de2a9f24f9`; native API30
build1654686/log3169 line1078 and corrected-source execution pass.
- TabbedPage: reviewed native11 captions on the matching API30 profile
after fixing actual tint. Independent bitmap-color assertions catch a
color defect that occupies a small screenshot area.
- SwipeView: original raster colors required by net11 `GetIconTintColor`
semantics. No global fixture color override is included. The earlier
pre-merge PNG restoration was superseded because main changed that
fixture's layout. API30 and iOS26 captures preserve native raster
colors; the iOS change is limited to the two grocery icons.
- Windows default CV2 grouping/spacing: reviewed captures from
run46111660/results100055/100056. `CollectionView1` is a category shard,
not a handler variant; the dedicated Windows CV1 stage is separate. All
grouped data remains present and the first item is visible. Issue36422
additionally has a stable label AutomationId and an independent
no-upward-shift assertion.
#### Strict UI recovery and SafeArea polling
`baf64263282` reuses isolated upstream SafeArea commit `6e47ece4882`
from #39742. Delayed constraints avoid recording intermediate NUnit
failures while retaining **15 seconds / 500-ms polling / 1 pixel**,
subsequent checks and restoration.
The retry merger retained a discovery-time ignored BoxView case as
failed after fixture setup failed, although the complete retry recovered
ten cases and preserved the original ignore. Recovery now requires exact
assembly/full-name identity, the same discovery-time ignore reason, a
complete fixture, an actual pass and no cancellation. Ordinary
failure-to-skipped still fails. Replaying unchanged real CI reports
through the actual old/new C# helper changes219 passed/1 failed/1 not
executed to219 passed/0 failed/2 preexisting not executed. No pipeline
tests or new ignores are added.
#### Existing Material3 indicator fix retained in native Shell
`907a4135c51` ports the existing #38445 `UpdateActiveIndicatorColor`
callback to native `ShellItemHandler`; native TabbedPage already retains
its theme subscription. The omitted Shell callback left light `#E8DEF8`
after a dark switch instead of `#4A4458`.
Only two API36 expectations are reconciled with current fixture color
roles and native11 captions after correcting the actual indicator.
Code-only native diff bounds are `(313,2525)-(537,2638)`. With corrected
expectations unchanged, removing only the production callback makes
Shell fail again while TabbedPage passes. Open/recently merged work was
checked, including #38611, whose default-capture mechanism is different.
#### Network-independent HTML history and retained diagnostic artifacts
`7d49f3b5865` keeps the original initial `HtmlWebViewSource`, null
`BaseUrl` and JavaScript anchor navigation, but navigates to a small
bundled HTML document rather than an external GitHub page. Every
navigation wait's returned Boolean is asserted; no history assertion is
removed. On the native iPhone simulator, the fixture passes, fails with
the original `LoadHtmlString` production history implementation, passes
after restoration, and passes after publication.
`a5ac3382e87` additionally resolves that document beneath
`Contents/Resources` on Catalyst, matching the actual app-bundle
resource layout while preserving the initial null `BaseUrl`. Hosted Mac
WebView task4196 on build1657252 completed49/49; exact Issue30381
passed35s, with result and artifact publication successful.
External-page accessibility was not measured during the earlier command
timeout, so a precise earlier Catalyst hang cause is not claimed. The
local Catalyst app compiled with0 warnings/errors; it was not launched
through host XCTest.
`7d49f3b5865` shortens only the isolated diagnostic-directory name,
retaining the complete128-bit GUID as URL-safe base64. Raw Catalyst
artifact log4422 proves `TF205023`:407-character retry attachment paths
exceeded Azure's399-character limit. The directory shrinks from38 to24
characters, producing393-character paths. Artifact names/formats,
separate raw evidence and retry behavior are unchanged. The actual UI
Cake script compiles in dry-run mode. Subsequent ordinary Azure artifact
publication succeeded, but that is **not** proof that the
longest393-character retry path was exercised. No pipeline tests are
added.
#### Qualified native expectations and Entry fixture isolation
`f8f4e691aa9` changes only15 PNG expectations and one fixture line:
- Thirteen Windows CV1-configuration expectations use exact native
captures from build1657252/run46132638. Source requirements were
reviewed independently: Issue23377 native AX rectangles measure
exactly80-DIP item spacing rather than the old doubled spacing;
Issue25859 retains10-DIP gaps; Issue23708 retains all5/5/3 grouped
items; Issue39534 retains all eight full11-character dates after six
swaps; delayed cards and vertical spacing preserve their required
dimensions and complete visible items. Selected colors/items remain
correct. The Issue38730 fixture explicitly uses `CollectionView2`, even
in the CV1 configuration; its desktop three-column layout is not being
mistaken for the legacy handler.
- The two Mac appearance expectations retain the title transition,
content geometry and restored default colors. Meaningful differences in
the reset capture were confined to native navigation/tab-bar separator
rows; the large-title capture differs in the native separator while
preserving the standard title and body layout.
- The Entry fixture calls `EntryControl.Unfocus()` **before** resetting
its view model and pushing the options page. This is an editing-state
isolation candidate. The three final Mac Entry failures passed in the
initial attempt but failed during the full-fixture retry, with
unexpected clear-button glyphs; their goldens remain unchanged. Four
related iPhone-simulator tests passed before and after publication, and
the Catalyst app compiles. **Native Mac acceptance was pending at the
initial checkpoint; compilation and iPhone preservation were not causal
proof of a Mac fix.** Later exact native Entry acceptance is recorded
above.
All15 captures replay through the repository's unchanged
`MagickNetVisualComparer`; all12 existing stale expectations reproduce
their CI-reported differences, and nine deliberately corrupted controls
remain rejected. This qualifies the image expectations; it is **not** a
new end-to-end UI run. Issue36422's post-screenshot position assertion
was still awaiting new CI at that checkpoint; its later actual pass is
recorded above.
The incorrect Windows Issue38059 horizontal-grid expectation is
deliberately not refreshed: both old and current captures show a
two-column vertical grid despite the fixture replacing its layout with
horizontal span4. No Windows handler patch is included. A local Windows
build probe restored missing inputs but stopped at `MakePri.exe` with an
execution-format error on macOS; no Windows build pass or native
validation bypass is claimed.
No badge, black Mac capture, Windows RTL, or failing Entry expectation
is refreshed. No public framework API, generated manifest or unrelated
formatting churn is added. Imported navigation-manager interface changes
are internal.
#### Validation commands and boundaries
- `git diff --check` — passed for corrective commits and merge
resolutions, including the latest change set.
- `./.dotnet/dotnet format Microsoft.Maui.sln --no-restore --exclude
Templates/src --exclude-diagnostics CA1822 --include
src/Core/src/Platform/iOS/TabbedViewExtensions.cs
src/Controls/src/Core/Toolbar/Toolbar.Windows.cs --verbosity quiet` —
passed with the pinned SDK. Scoped formatting was repeated for changed
test/merge files; existing indentation was retained without whole-file
whitespace churn.
- `./.dotnet/dotnet format whitespace Microsoft.Maui.sln --no-restore
--include
src/Controls/tests/TestCases.HostApp/FeatureMatrix/EntryControl/EntryControlPage.xaml.cs
--verbosity quiet` — passed for the latest one-line fixture change at
`f8`.
- `./build.sh -restore -build -configuration Debug -projects
"$PWD/Microsoft.Maui.BuildTasks.slnf" -warnAsError false` — passed,0
warnings/errors, at the initial correction. The earlier relative-path
attempt failedMSB3202; absolute-path retry passed.
- `./.dotnet/dotnet build src/Core/src/Core.csproj -f net11.0-ios27.0
--no-restore --nologo -v:quiet` — passed,0 warnings/errors, including
integrated gesture code. Initial Preview6 NETSDK1140 was resolved by
repository provisioning.
- `DEVELOPER_DIR=/Applications/Xcode-27.1.0.app/Contents/Developer
./.dotnet/dotnet build src/Controls/src/Core/Controls.Core.csproj -f
net11.0-maccatalyst27.0 -c Debug -p:ValidateXcodeVersion=false -v:q` —
passed,0 warnings/errors, after appearance integration. An initial
unversioned-target attempt failedNETSDK1005; the actual versioned target
passed. Compilation is not a Catalyst-native test run.
- `DEVELOPER_DIR=/Applications/Xcode-27.1.0.app/Contents/Developer
./.dotnet/dotnet build
src/Controls/tests/TestCases.HostApp/Controls.TestCases.HostApp.csproj
-f net11.0-maccatalyst -c Debug -p:ValidateXcodeVersion=false -v:q` —
passed,0 warnings/errors, for the resource follow-up and latest Entry
candidate. No host XCTest launch.
- **Native Android fixture controls:** Release CoreCLR
`ShellHandlerTests_Shell`: fixed79/79; original callbacks76/79 with
exactly three CI failures; restored79/79.
- **Native Android tint controls:** API30 arm64 `TabbedPageTests`:
fixed64/64; production callback removed62/64 with missing native
magenta/lime pixels; restored64/64. Combined TabbedPage/native
Shell143/143. Built through `Run-DeviceTests.ps1 -Project Controls
-Platform android -Configuration Release -BuildOnly -IncludeClasses ...`
and executed through pinned XHarness `.TestInstrumentation`; fresh XML
identities/totals checked.
- **Focused API30 UI:** `BuildAndRunHostApp.ps1 -Platform android
-Configuration Release -DeviceUdid emulator-5570 -TestFilter
'FullyQualifiedName~Issue38765|FullyQualifiedName~VerifyShellMenuItemsAlignedInRTL|FullyQualifiedName~VerifyCollectionViewContentWithIconImageSwipeItem|FullyQualifiedName~Issue36422'`
—4/4, including after publication, on1080x1920/420dpi with explicit
serial, PLATFORM_VERSION11 and Pixel_2_API_30. Actual Appium device
verified. Earlier API34 attempts failed unchanged profile guards and are
not visual passes.
- **Apple disposal controls:** both cases pass, reproduce
InvalidCastException with original registration and pass after
restoration. Full handler class fixed/restored23/27, reverted21/27
before the separate four gesture failures were integrated; hosted c0 XML
confirmed disposal in inherited variants.
- **Combined native Apple:** `pwsh
.github/skills/run-device-tests/scripts/Run-DeviceTests.ps1 -Project
Controls -Platform ios -Configuration Debug -iOSVersion 26 -DeviceUdid
AA7BCC0E-CD5B-481A-BF91-402EB2CDFD6C -SkipXcodeVersionCheck
-IncludeClasses
'Microsoft.Maui.DeviceTests.NavigationPageTests,Microsoft.Maui.DeviceTests.NavigationPageNavigationHandlerTests,Microsoft.Maui.DeviceTests.ShellTests,Microsoft.Maui.DeviceTests.ShellNavigationHandlerTests,Microsoft.Maui.DeviceTests.SafeShellTabBarAppearanceTrackerTests,Microsoft.Maui.DeviceTests.ShellNavBarAppearanceTrackerTests'`
— fresh XML556 passed/0 failed, exact counts27+27+243+243+10+6.
Appearance repeated16/16 after publication. XHarness wrapper exits1
because it expects testResults.xml while fresh output is
xunit-test-*.xml; no successful wrapper invocation is claimed.
iPhone11Pro/iOS26.5 uses supported Xcode27.1 override, not exact CI
Xcode certification.
- **Focused iOS UI:** existing gradient, reset, orientation, keyboard
and raster cases5/5. No physical Apple or host XCTest UI automation.
- **Material3:** `./.dotnet/dotnet build
src/Controls/tests/TestCases.HostApp/Controls.TestCases.HostApp.csproj
-f net11.0-android -c Release -t:Install -p:UseMaterial3=true
'-p:AdbTarget=-s emulator-5570' -v:q` — passed on isolated API36
Pixel3XL arm64,1440x2960/560dpi. `PLATFORM_VERSION=16
DEVICE_SKIN=pixel_3_xl ANDROID_SERIAL=emulator-5570
DEVICE_UDID=emulator-5570 ./.dotnet/dotnet test
src/Controls/tests/TestCases.Android.Tests/Controls.TestCases.Android.Tests.csproj
-c Release -p:UseMaterial3=true --filter
'FullyQualifiedName~Issue38234'` — original0/2 with
CI-identical1.27%/2.55% differences; fixed2/2; callback removed1/2
precisely Shell; restored2/2 and post-push2/2. Initial capability36
selected no device; OS version16 selected verified API36.
- **Bundled HTML history:**
`DEVELOPER_DIR=/Applications/Xcode-27.1.0.app/Contents/Developer
ValidateXcodeVersion=false pwsh .github/scripts/BuildAndRunHostApp.ps1
-Platform ios -Configuration Debug -DeviceUdid
AA7BCC0E-CD5B-481A-BF91-402EB2CDFD6C -TestFilter
'FullyQualifiedName~Issue30381'` —1/1; original history production
behavior reverted0/1 with Expected True/Actual False; restored1/1;
post-push1/1. Temporary production change is removed.
- **Entry candidate preservation:**
`DEVELOPER_DIR=/Applications/Xcode-27.1.0.app/Contents/Developer
ValidateXcodeVersion=false pwsh .github/scripts/BuildAndRunHostApp.ps1
-Platform ios -Configuration Debug -DeviceUdid
AA7BCC0E-CD5B-481A-BF91-402EB2CDFD6C -TestFilter
'FullyQualifiedName~VerifyTextWhenFontSizeSetCorrectly|FullyQualifiedName~VerifyPlaceholderWithShadow|FullyQualifiedName~VerifyEntryControlWhenPlaceholderTextSet|FullyQualifiedName~VerifyClearButtonVisibilityWhenIsReadOnlyTrue'`
— exact4/4 passed before and after publication of `f8f4e691aa9`; fresh
TRX names/outcomes checked. This is not Mac runtime acceptance.
- **Cake configuration:** `DOTNET_ROLL_FORWARD=Major ./.dotnet/dotnet
tool run dotnet-cake eng/devices/ios.cake --target=uitest --dryrun
--configuration=Debug
--project="$PWD/src/Controls/tests/TestCases.iOS.Tests/Controls.TestCases.iOS.Tests.csproj"
--appproject="$PWD/src/Controls/tests/TestCases.HostApp/Controls.TestCases.HostApp.csproj"`
— passed; no tasks executed. Actual directory expression measured24
characters and observed path393<399.
#### Current hosted gates and unresolved failures
Current exact `cd7a8cb8a21` runs are
[maui-pr1657489](https://dev.azure.com/dnceng-public/public/_build/results?buildId=1657489),
completed succeeded31/31 jobs,
[UI1657492](https://dev.azure.com/dnceng-public/public/_build/results?buildId=1657492),
still running with genuine failures, and
[device1657491](https://dev.azure.com/dnceng-public/public/_build/results?buildId=1657491),
completed failed7/9 jobs with the exact native Core failure and
unmeasured Windows deployment work items recorded above. Source SHA is
verified; no all-three-gate green claim is made. The earlier
`f8f4e691aa9` runs were
[maui-pr1657400](https://dev.azure.com/dnceng-public/public/_build/results?buildId=1657400),
[device1657401](https://dev.azure.com/dnceng-public/public/_build/results?buildId=1657401)
and
[UI1657402](https://dev.azure.com/dnceng-public/public/_build/results?buildId=1657402),
all admitted with the exact earlier source SHA; complete newer evidence
from those runs is recorded above.
On previous `a5ac3382e87`, **main1657242 succeeded31/31**,
finishing2026-10-31T04:18:36UTC, and **device1657251 succeeded9/9**,
finishing04:11:27UTC. All62 actual native app/queue reports were
independently audited:46,202 tests=44,980 passed+1,222 existing
exclusions,0 failures/errors. Two Windows FlexLayout reports had zero
selected tests and are explicitly not passing measurements.
Unchanged-payload targeted retries recovered Android Mono Essentials
(exact424/424 x86), a Catalyst queue timeout where Core had never
executed, and a pre-test Windows MSIX deployment failure. Both exact
published MSIX files had valid CRC/manifests and matching x64 runtime
dependencies; no speculative package/file-lock/source fix or
worker-cause claim is made.
On prior907a, **main1657009 completed31/31** and **device1657010
completed all nine raw-submission-correlated Helix jobs**, including
Windows Graphics packaged/unpackaged. A transient Windows Release
Resizetizer sharing violation recovered through one targeted
failed-monitor retry: only the failing work item was resubmitted with
unchanged payload; complete XML796=794 passed+2 original skips,
including both exact ResizeQuality cases. No speculative file-lock
source fix or infrastructure cause is claimed.
UI still needs repair. The final Mac CollectionView category on1657252
published69 initial results (41 passed,27 failed,1 existing exclusion),
then its retry was canceled at the job deadline. All27 failure
identities and44 original diagnostic attachments were retained.
Representative actual images were entirely black—including the
full-desktop screenshot—while AX still contained the expected grouped
content and window geometry. Those captures cannot certify a layout
regression or a new golden; the capture/app-responsiveness cause remains
unproven. Generic Appium timeout wording is not proof of an infinite
layout loop. No black golden or timeout/ignore workaround is included.
The later fresh-worker native69-case recovery above does not establish a
cause.
**Badge colors are a real native rendering failure, not a safe golden
refresh:** source Home badgeBlue rendersRed on Catalyst, and an
independent iPhone native bitmap rendersRed despite the correct Green
native property getter. Per-item StandardAppearance was rejected because
it colors every badge like the selected tab; global clearing, ordering
and late-assignment probes also did not fix independent colors. All
experiments and failing reproduction-only changes were removed. UIKit
reports match this limitation ([native
report](https://stackoverflow.com/questions/78997012/ios-18-per-item-badge-color-on-uitabbar)),
but no safe supported per-item fix is certified. No Red golden, mask,
tolerance relaxation or new ignore is published.
The Mac Picker31731 case still fails waiting for the root page in both
complete attempts; its original native screenshot shows the picker
popover still open. A precise focus/presentation/navigation cause has
not been demonstrated. Windows Issue38059 horizontal-layout replacement
also remains unresolved, and its incorrect expectation is not refreshed.
The earlier Windows Flyout RTL glyph-corruption receipt establishes
prior inconsistent rendering, not the precise cause or the current-head
result. Its golden is not refreshed. Expected and actual
BackButtonHandled are bothFalse; an earlier True comparison was
corrected. Windows CancelButtonColor passed on907a's predecessor; later
exact `f8` native evidence supports only the focused-label oracle
correction described above, not a speculative color-production change.
Human-reviewed integration into protected `net11.0` is still required
before #39688 inherits corrections. No protected PR has been merged or
approved.
#### Latest parent-only publication and current native UI receipts
Parent branch `kubaflo-net11-promotion-compile-fixes` is now normally
pushed at `79eddf2b2d0dab4d4d7c4e5e26a2a2ee9856f2f5`, containing `f26`
GC plus original `cc6248` Windows focus production logic (8 lines) and
its corrected three-variant native fixture (final74 lines). This adds
only two new files/82 lines; no golden, SDK, shared-helper or tolerance
changes. The original corrected fixture trees were located publicly in
#39742 at `ff8f3461055c922d3cf5ad504bbd8a6d88dad42b` and
`ae09556e2a771a2a2a0c73292b6051a8b59c608f`. Retained peer build1657386
receipts cover all three variants in both package modes, but are
**upstream-scope evidence, not new combined-head acceptance or
CancelButtonColor certification**. The earlier not-imported wording
above describes the prior checkpoint; only the parent now contains the
focus port. #39779 counterpart and linked checkout remain exactly `cd`,
without this new port.
- `./.dotnet/dotnet format Microsoft.Maui.sln --no-restore --include
src/Controls/src/Core/Handlers/Shell/ShellItemHandler.Windows.cs
src/Controls/tests/DeviceTests/Elements/Shell/ShellTests.Windows.cs
--exclude Templates/src --exclude-diagnostics CA1822 --verbosity quiet`
— passed.
- `git diff --check` — passed.
- Historical combined `79` manual device1657568 Windows job
`2bb3c12a-e73a-4044-b592-b131972acf99` was created at09:41:16, matching
the exact Windows phase at09:39. Original Controls packaged and
unpackaged Shell XMLs each completed77/77 passed,0 skips/failures. Exact
`ShellTests.SearchHandlerFocusTracksDetachment` variants passed in both
modes: removal/refocus0.8307094s packaged and0.8259344s unpackaged;
replacement/refocus0.5663891s/0.5465569s;
disconnect0.4454088s/0.6898395s. Receipts: parent
`files/79-native-windows-packaged-Shell.xml` and
`files/79-native-windows-unpackaged-Shell.xml`. This is combined-`79`
native acceptance for all six focus oracles after the ported production
change—not merely upstream build1657386 evidence—but it is not current
`691` head/native/all-three-pipeline acceptance and does not execute the
new Uppercase assertion. Windows job-level `Finished` remains null: one
packaged MauiBlazorWebView work item (`a000OQ0`) is still running with
no files and console404, while the other nine apps plus the
administrative item exited0. Other app XMLs have not been fully audited,
so no10/10 certification, native hang/crash, or source-cause claim is
made; cached Shell-glob search returning0 is not a measurement, while
the authoritative49-file API listing supplied the original Shell XML
links. Historical combined-`79` CoreCLR retry
`dd84ab5e-00f0-4234-9b53-c07d5d6a5888` finished at10:20:53 with all five
apps plus admin exit0, but its actual retry XML audit remains pending,
so no full test counts are claimed. Catalyst Core OS26 remains
unmeasured after the prior cancellation; no overall `79` gate pass is
claimed.
Current `cd` Windows CV1
[raw4584](https://dev.azure.com/dnceng-public/public/_apis/build/builds/1657492/logs/4584),
publisher4586/run46141746, has initial355=332 passed+10 failed+13
existing skips. Nine setup failures recovered in the10-case targeted
retry (9 passed/1 failed); final355=341 passed+13 existing skips+1 exact
`Issue38059_UntemplatedHorizontalGridLayout` failure. All13 qualified
goldens and Issue36422's post-screenshot assertion pass. The incorrect
horizontal golden is not refreshed. Complete355-result JSON was retained
by the parent.
Current Catalyst Shell1
[raw4518](https://dev.azure.com/dnceng-public/public/_apis/build/builds/1657492/logs/4518),
publisher4520/run46141376, has initial148=144 passed+4 same badge
failures,0 skips; all four targeted retries fail. Complete148-result
JSON and the existing iOS26.5 complete173-result JSON were retained.
There is still no safe badge fix or badge-golden change. Main was29/31
jobs with remaining jobs running at the latest read, not a full-green
measurement; overall main/UI remain running and current `cd` devices
remain failed7/9 jobs. Parent-only/manual and upstream validation do not
certify this counterpart.
#### Hosted parent-only GC after-acceptance
On manual `f26` device1657552, Android Mono
[raw157](https://dev.azure.com/dnceng-public/public/_apis/build/builds/1657552/logs/157)
succeeded. Helix job `a0a8fd7d-fdd6-45aa-a092-3312e36053be`, Core work
item on worker `a000ADW`, produced [complete original native
XML](https://helixr1107v0xdeko0k025g8.blob.core.windows.net/dotnet-maui-refs-heads-kubaflo-net11-promotioa0a8fd7dfdd645aaa0/com.microsoft.maui.core.devicetests-Signed/1/testResults.xml?helixlogtype=result):3090=3018
passed+72 existing skips,0 failures/errors. The parent parsed original
bytes and reconciled declared/actual counts; receipt
`files/f26-native-android-mono-core-results.xml`. All nine
`AssertionExtensionsTests` pass: exact earlier-reference True
variant4.4212107s, False5.1103807s; retained-order
False8.5293263s/True8.3305855s; all null/empty controls pass.
Before `cd` Core on worker `a000ADL` was3090=3017 passed+72 existing
skips+1 True-variant failure4.999277s; after `f26` on `a000ADW` is the
complete native cohort above. This is hosted before/after cohort
evidence, **not an isolated native undo/revert control or full promotion
certification**. The parent completed the full native audit for source
`f26` manual device1657552, which **completed succeeded9/9**.
Across62/62 logical app reports,154 canonical original XMLs
declare46,202 tests=44,980 passed+1,222 existing skips+0
failures/errors, with every assembly-declared count reconciled. All71
individual work items—62 apps plus9 administrative items—finished with
exit0; administrative items are excluded from test totals. Azure
reports9/9 jobs complete and all11/11 individual Windows work items
complete, while Windows job-level `Finished` metadata remains
null/stale; this does **not** claim all nine job `Finished` fields. The
five-app Android Mono cohort remains Core3090=3018 passed+72 skips,
Controls1345=1332 passed+13 skips, Essentials424/424 passed,
Graphics33/33 passed, and Blazor68=64 passed+4 skips. Actual new Windows
packaged results are nonempty full-runtime XMLs: Controls1048=1015
passed+33 existing skips and Core2146=2035 passed+111 existing skips.
The earlier PRETEST Appx missing-manifest condition did not recur in
this `f26` cohort; no source, cause, package, worker, or
infrastructure-outage attribution is made. Receipts: parent
`files/f26-all-device-native-audit.json`,
`files/f26-all-device-native-audit-originals/`, and the earlier scoped
XML receipts. This certifies only the older `f26` GC native cohort—not
current `d673`, combined `79` focus repair, or all-three-pipeline
acceptance. Earlier8/9/Windows-running statements are historical
superseded checkpoints.
#### Latest combined-parent CoreCLR failure and bounded retry
On `79` manual1657568, Android CoreCLR
[raw141](https://dev.azure.com/dnceng-public/public/_apis/build/builds/1657568/logs/141),
Helix `833b315d-8390-4238-a247-f3dfe3e63c0f`, Controls worker `a000ADY`,
produced complete native XML1345=1331 passed+13 existing skips+1
failure. Exact
`HybridWebViewTests_ContentRootResolution.RootedPath_DoesNotResolve(path:
//content/page.html)` failed in26.2155066s at the25000ms
`WaitForHybridWebViewLoaded` **before the path oracle**. This is not
evidence of a rooted-path security failure, a product source bug or an
infrastructure outage. The parent audited all five actual app
XMLs:4960=4870 passed+89 existing skips+1 failure; Core3090=3018
passed+72 existing skips, with all nine GC-class cases passing;
Essentials424 passed; Graphics33 passed; Blazor68=64 passed+4 existing
skips. Receipts: parent `files/79-native-android-coreclr-*.xml`.
The `f26` CoreCLR comparison from
[raw149](https://dev.azure.com/dnceng-public/public/_apis/build/builds/1657552/logs/149),
Helix `0b63b622-3b9b-47bf-a4ae-f9cdbd260854`, worker `a000ADU`,
exercised the same unchanged Android Controls logic: complete1345=1332
passed+13 existing skips,0 failures; exact same `//content/page.html`
case passed in0.452862s, as did all other rooted-path cases. Receipt:
parent `files/f26-native-android-coreclr-controls-results.xml`. No wait,
URL, SDK or source change was made.
One bounded CoreCLR stage retry was positively verified admitted as
attempt2 for stage ref `devicetests_android_coreclr`, stage ID
`ed3a755a-04c9-545d-584b-d7b617cca329` (previous attempt1), with
`forceRetryAllJobs=false`; recovery is pending. An initial
GUID-as-stage-ref PATCH returned empty CLI output but did not admit an
attempt according to the raw timeline; correcting the actual identifier
admitted attempt2. This is one effective retry, not multiple admitted
recoveries.
`az devops invoke --org https://dev.azure.com/dnceng-public --area build
--resource stages --route-parameters project=public buildId=1657568
stageRefName=devicetests_android_coreclr --http-method PATCH
--api-version 7.1 --in-file
/Users/kubaflo/.copilot/session-state/d595bf0b-ce58-4118-be08-7fe144955a64/files/retry-failed-helix-stage.json
-o none` — effective admission verified by actual stage attempt2;
request body `state=retry`, `forceRetryAllJobs=false`. This is not
native recovery, full combined-head acceptance, Windows focus acceptance
or promotion certification. Historical pending statements above are
checkpoints; this receipt updates only the combined-parent CoreCLR
failure and admitted pending retry. Counterpart and checkout remain
`cd`.
### Issues Fixed
<!-- Please make sure that there is a bug logged for the issue being
fixed. The bug should describe the problem and how to reproduce it. -->
Refs #39688, #38765, #38234, #30381, and reused implementations #39746,
#39761, #38445. This prerequisite does not auto-close an issue.
<!--
Are you targeting main? All PRs should target the main branch unless
otherwise noted.
-->
---------
This branch was previously deployed
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Issue summary
On iOS, when navigating from a root page to a child page within either NavigationPage or Shell, overriding OnBackButtonPressed on the child page to return true was not respected. The page would still pop back to the previous/root page upon tapping the native back button or performing an interactive swipe-back gesture.
Root Cause
Interactive pop gesture handling and back button evaluation logic were fragmented and duplicated between NavigationViewHandler and ShellSectionHandler.
ShellSectionHandler implemented its own private GestureDelegate which did not consistently hook into the underlying navigation lifecycle or block the native interactive pop gesture when the page or BackButtonBehavior handled the back action.
The native interactive pop gesture recognizer was not consulting the back-press policy early enough during the gesture evaluation phase (ShouldBegin).
Description of Change
INavigationManagerDelegate.cs
NavigationControllerManager.cs
NavigationViewHandler.iOS.cs
ShellSectionHandler.iOS.cs
Reference PR
.NET 10 Reference PR #38071
Validated the behaviour in the following platforms
Screenshots