Skip to content

Improved style inheritance#31317

Merged
kubaflo merged 6 commits intodotnet:inflight/currentfrom
kubaflo:Issue31280
Apr 1, 2026
Merged

Improved style inheritance#31317
kubaflo merged 6 commits intodotnet:inflight/currentfrom
kubaflo:Issue31280

Conversation

@kubaflo
Copy link
Copy Markdown
Contributor

@kubaflo kubaflo commented Aug 24, 2025

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 from this PR and let us know in a comment if this change resolves your issue. Thank you!

Description of Change

Corrects style inheritance in Style.ApplyCore by unapplying the base style before reapplying it.

Issues Fixed

Fixes #31280

Copilot AI review requested due to automatic review settings August 24, 2025 12:23
@kubaflo kubaflo requested a review from a team as a code owner August 24, 2025 12:23
@kubaflo kubaflo self-assigned this Aug 24, 2025
Copy link
Copy Markdown
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull Request Overview

This PR fixes a style inheritance issue where derived styles were not properly inheriting from their base styles when properties were overridden. The fix ensures that when a style is applied, the base style is first unapplied and then reapplied, allowing proper inheritance behavior.

  • Corrects style inheritance by unapplying base styles before reapplying them in Style.ApplyCore
  • Adds comprehensive unit tests to verify style inheritance works correctly across multiple levels
  • Validates that overridden properties in derived styles are properly applied while maintaining base style properties

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File Description
src/Controls/src/Core/Style.cs Core fix adding base style unapplication before reapplication in ApplyCore method
src/Controls/tests/Xaml.UnitTests/Issues/Issue31280.xaml Test XAML defining hierarchical styles with property overrides to validate inheritance
src/Controls/tests/Xaml.UnitTests/Issues/Issue31280.xaml.cs Unit test class verifying that style inheritance works correctly for both compiled and non-compiled XAML

@dotnet-policy-service dotnet-policy-service bot added the community ✨ Community Contribution label Aug 24, 2025
@dotnet-policy-service
Copy link
Copy Markdown
Contributor

Hey there @@kubaflo! Thank you so much for your PR! Someone from the team will get assigned to your PR shortly and we'll get it reviewed.

@github-actions
Copy link
Copy Markdown
Contributor

github-actions bot commented Mar 16, 2026

🚀 Dogfood this PR with:

⚠️ WARNING: Do not do this without first carefully reviewing the code of this PR to satisfy yourself it is safe.

curl -fsSL https://raw.githubusercontent.com/dotnet/maui/main/eng/scripts/get-maui-pr.sh | bash -s -- 31317

Or

  • Run remotely in PowerShell:
iex "& { $(irm https://raw.githubusercontent.com/dotnet/maui/main/eng/scripts/get-maui-pr.ps1) } 31317"

Copy link
Copy Markdown
Contributor

@StephaneDelcroix StephaneDelcroix left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this looks good, but Style changes have a tendency to create regressions, I'd recommend targeting net11.0

also, xaml unittests no longer use nunit but xunit

StephaneDelcroix added a commit that referenced this pull request Mar 18, 2026
Cherry-pick of PR #31317 (commit 1d7b946) targeting net11.0.
Fixed test to use xUnit patterns (XamlInflator, [Theory], Assert.Equal)
instead of NUnit patterns ([TestFixture], [TestCase], Assert.AreEqual).
Renamed test files from Issue31280 to Maui31280 per naming conventions.

Fixes #31280

Co-authored-by: Jakub Florkowski <kubaflo123@gmail.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
@StephaneDelcroix StephaneDelcroix changed the base branch from main to net11.0 March 18, 2026 11:11
Copy link
Copy Markdown
Contributor

@StephaneDelcroix StephaneDelcroix left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Multi-Model Consensus Review (5 models: claude-opus-4.6 ×2, claude-sonnet-4.6, gemini-3-pro-preview, gpt-5.3-codex)

CI Status: ✅ All checks passing

Consensus Findings

No consensus bugs found (5/5 models clean).

One model raised a potential re-entrancy concern via RemoveDynamicResource → OnBasedOnResourceChanged but this was not confirmed by any other model and does not reach consensus threshold.

Test Coverage

✅ Includes XAML test (Issue31280.xaml) and unit test (Issue31280.xaml.cs) exercising multi-level style inheritance with property overrides.

Verdict: ✅ Approve

Minimal, correct fix with proper safety guard. No regressions identified.

Copy link
Copy Markdown
Contributor

@StephaneDelcroix StephaneDelcroix left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

rename "Issue31280" to "Maui31280". IssueXXXXX used to track issues in the Xamarin.Forms repository, MauiXXXXX on maui

@kubaflo kubaflo changed the base branch from main to inflight/current March 18, 2026 13:36
@kubaflo kubaflo changed the base branch from inflight/current to main March 18, 2026 13:39
@kubaflo
Copy link
Copy Markdown
Contributor Author

kubaflo commented Mar 18, 2026

/azp run maui-pr

@azure-pipelines
Copy link
Copy Markdown

Azure Pipelines successfully started running 1 pipeline(s).

@MauiBot
Copy link
Copy Markdown
Collaborator

MauiBot commented Mar 22, 2026

🤖 AI Summary

📊 Expand Full Review7b440a8 · Update Maui31280.xaml.cs
🔍 Pre-Flight — Context & Validation

Issue: #31280 - Styles based on a style that is based on another style that uses AppThemeBinding do not inherit properties correctly.
PR: #31317 - Improved style inheritance
Platforms Affected: iOS, Android, Windows, macOS
Files Changed: 1 implementation, 2 test

Key Findings

  • The PR changes src/Controls/src/Core/Style.cs to unapply a base style before reapplying it inside Style.ApplyCore, targeting inherited-style behavior when AppThemeBinding is involved.
  • The linked issue is verified on all platforms and describes a three-level style inheritance case where the third style incorrectly resolves to the original base color instead of the overridden inherited color.
  • The PR adds a XAML unit test under src/Controls/tests/Xaml.UnitTests/Issues/ to validate the inheritance chain and expected TextColor values.
  • Prior human review called out repository conventions: XAML unit tests now use xUnit and MAUI issue tests should use the MauiXXXXX naming pattern rather than IssueXXXXX.
  • No prior PRAgent phase output was found in PR comments; this run starts from pre-flight.

Edge Cases / Discussion Notes

  • The issue reproduces only when the base style uses AppThemeBinding; simple literal values do not exhibit the problem.
  • The reported workaround is to avoid AppThemeBinding in the base style.
  • Because this PR uses XAML unit tests rather than HostApp UI tests, gate verification will need to use the XAML unit test project rather than the UI-test verification path.

Fix Candidates

# Source Approach Test Result Files Changed Notes
PR PR #31317 Unapply base style before reapplying it during Style.ApplyCore, plus add a regression XAML unit test covering multi-level style inheritance with AppThemeBinding. ⏳ PENDING (Gate) src/Controls/src/Core/Style.cs, src/Controls/tests/Xaml.UnitTests/Issues/Maui31280.xaml, src/Controls/tests/Xaml.UnitTests/Issues/Maui31280.xaml.cs Original PR

🚦 Gate — Test Verification

Gate Result: ❌ FAILED

Platform: android (host context); verification executed with platform-agnostic XAML unit tests because this PR adds no HostApp/UI test coverage.
Mode: Exact PR Validation
Test Project: src/Controls/tests/Xaml.UnitTests/Controls.Xaml.UnitTests.csproj
Test Filter: StyleInheritanceShouldWork

  • Tests FAIL without fix: ✅
  • Tests PASS with fix: ❌

Evidence

  • The underlying regression is real: when the Style.ApplyCore change is removed, the regression reproduces and the deepest style resolves TextColor to green instead of red.
  • However, the PR as submitted does not currently pass the new test project build because src/Controls/tests/Xaml.UnitTests/Issues/Maui31280.xaml.cs still imports NUnit.Framework in an xUnit test project.
  • Exact PR-state validation fails with compile errors before test execution:
    • error CS0246: The type or namespace name 'NUnit' could not be found
    • error CS0246: The type or namespace name 'CollectionAttribute' could not be found
  • As a result, the PR fix is not merge-ready in its current form even though the Style.cs logic change itself appears to address the reported behavior once the test file is corrected.

🔧 Fix — Analysis & Comparison

Fix Candidates

# Source Approach Test Result Files Changed Notes
1 try-fix (claude-opus-4.6) Detect style re-application in IStyle.Apply and call UnApplyCore before reapplying so stale AppThemeBinding state is cleared by the style itself. ✅ PASS* 2 files Passed once the test file import was corrected to xUnit.
2 try-fix (claude-sonnet-4.6) Replace the static OnBasedOnResourceChanged callback with a style-owned callback that captures the specific Style instance and reapplies the correct owner in ApplyCore. ✅ PASS* 2 files Passed once the test file import was corrected to xUnit.
3 try-fix (gpt-5.3-codex) Keep the callback static but route OnBasedOnResourceChanged to the matching style node in the BasedOn chain before unapply/reapply. ✅ PASS* 1 file Passed once the test file import was corrected to xUnit.
4 try-fix (gemini-3-pro-preview) Validate resource identity in OnBasedOnResourceChanged so nested style updates are ignored unless they match the root style's BaseResourceKey. ✅ PASS* 2 files Passed once the test file import was corrected to xUnit.
5 try-fix (claude-opus-4.6 round 2) Invert ApplyCore ordering so derived setters run before BasedOn, attempting to let derived values win via specificity/ordering instead of cleanup. ❌ FAIL 2 files Fundamentally incompatible with persistent AppThemeBinding listeners.
6 try-fix (claude-sonnet-4.6 round 2) Flatten the BasedOn chain into a winner map, unapply overridden base setters, and add a re-entrancy guard for the callback. ✅ PASS* 2 files Works, but substantially more complex than the PR fix.
PR PR #31317 Unapply the base style before reapplying it inside Style.ApplyCore. ❌ FAIL (Gate) 3 files Exact PR state fails to compile because Maui31280.xaml.cs uses NUnit.Framework in an xUnit project.

* All passing try-fix attempts also corrected the test file import from using NUnit.Framework; to using Xunit;, which is required for the regression test project to compile.

Cross-Pollination

Model Round New Ideas? Details
claude-opus-4.6 1 Yes Self-cleaning re-application in IStyle.Apply
claude-sonnet-4.6 1 Yes Make _basedOnResourceProperty callback style-owned instead of static/top-style-driven
gpt-5.3-codex 1 Yes Resolve callback to matching owner within the BasedOn chain
gemini-3-pro-preview 1 Yes Validate resource identity to filter callback cross-talk
claude-opus-4.6 2 Yes Invert ApplyCore order so base runs after derived at lower priority
claude-sonnet-4.6 2 Yes Flatten the BasedOn chain into one effective setter map
gpt-5.3-codex 2 No NO NEW IDEAS
gemini-3-pro-preview 2 No NO NEW IDEAS
claude-opus-4.6 3 Yes Add per-property style-precedence tokens to specificity/value provenance
claude-sonnet-4.6 3 Yes Apply AppThemeBinding through binding-engine precedence instead of value assignment
gpt-5.3-codex 3 Yes Preserve lowered specificity inside AppThemeBinding callback updates
gemini-3-pro-preview 3 Yes Remove stale bindings in Setter.Apply before setting plain values

Exhausted: Partial — exploration found additional theoretical directions, but the exact PR state already fails gate due to a concrete test compile issue, so further alternative execution was not necessary to reach a recommendation.
Selected Fix: Keep the PR's Style.ApplyCore logic and also correct src/Controls/tests/Xaml.UnitTests/Issues/Maui31280.xaml.cs to use xUnit (using Xunit;). That is the simplest merge-ready path.


📋 Report — Final Recommendation

⚠️ Final Recommendation: REQUEST CHANGES

Phase Status

Phase Status Notes
Pre-Flight ✅ COMPLETE Small Style fix plus new XAML regression test; issue verified on all platforms.
Gate ❌ FAILED Exact PR state does not build the new XAML test on the android-hosted review environment because Maui31280.xaml.cs imports NUnit.Framework in an xUnit project.
Try-Fix ✅ COMPLETE 6 executed attempts: 5 passing alternatives (all required the xUnit import correction), 1 failing alternative that confirmed cleanup of persistent AppThemeBinding state matters.
Report ✅ COMPLETE

Summary

The Style.ApplyCore logic change appears directionally correct: when I validated the behavior with the test file corrected to xUnit, the regression failed without the Style.cs change and passed with it. However, the PR as submitted is not merge-ready because the newly added test file still uses NUnit.Framework, so the XAML unit test project fails to compile before the regression can run.

Root Cause

There are two separate issues in the current PR state:

  1. Behavioral bug being fixed by the PR: multi-level BasedOn style application combined with AppThemeBinding can leave the deepest style inheriting the original base value (Green) instead of the intermediate override (Red).
  2. New test wiring bug in the PR itself: src/Controls/tests/Xaml.UnitTests/Issues/Maui31280.xaml.cs imports NUnit.Framework, but Controls.Xaml.UnitTests uses xUnit. That causes compile errors (CS0246 for NUnit, CollectionAttribute, etc.) in exact PR-state validation.

Fix Quality

The implementation in src/Controls/src/Core/Style.cs is small and plausible, and the try-fix phase found multiple alternative passing strategies — but none were clearly better once complexity was considered. The simplest path is to keep the PR’s Style.ApplyCore logic and fix the regression test file to use xUnit (using Xunit;). Until that test import issue is corrected, I would not approve the PR.


@MauiBot MauiBot added s/agent-changes-requested AI agent recommends changes - found a better alternative or issues s/agent-reviewed PR was reviewed by AI agent workflow (full 4-phase review) labels Mar 22, 2026
@kubaflo
Copy link
Copy Markdown
Contributor Author

kubaflo commented Mar 23, 2026

/azp run maui-pr-uitests

@azure-pipelines
Copy link
Copy Markdown

Azure Pipelines successfully started running 1 pipeline(s).

PureWeen and others added 2 commits March 25, 2026 09:44
…otnet#34548)

<!-- 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!

## Description

Adds a [gh-aw (GitHub Agentic
Workflows)](https://github.github.com/gh-aw/introduction/overview/)
workflow that automatically evaluates test quality on PRs using the
`evaluate-pr-tests` skill.

### What it does

When a PR adds or modifies test files, this workflow:
1. **Checks out the PR branch** (including fork PRs) in a pre-agent step
2. **Runs the `evaluate-pr-tests` skill** via Copilot CLI in a sandboxed
container
3. **Posts the evaluation report** as a PR comment using gh-aw
safe-outputs

### Triggers

| Trigger | When | Fork PR support |
|---------|------|-----------------|
| `pull_request` | Automatic on test file changes (`src/**/tests/**`) |
❌ Blocked by `pre_activation` gate |
| `workflow_dispatch` | Manual — enter PR number | ✅ Works for all PRs |
| `issue_comment` (`/evaluate-tests`) | Comment on PR | ⚠️ Same-repo
only (see Known Limitations) |

### Security model

| Layer | Implementation |
|-------|---------------|
| **gh-aw sandbox** | Agent runs in container with scrubbed credentials,
network firewall |
| **Safe outputs** | Max 1 PR comment per run, content-limited |
| **Checkout without execution** | `steps:` checks out PR code but never
executes workspace scripts |
| **Base branch restoration** | `.github/skills/`,
`.github/instructions/`, `.github/copilot-instructions.md` restored from
base branch after checkout |
| **Fork PR activation gate** | `pull_request` events blocked for forks
via `head.repo.id == repository_id` |
| **Pinned actions** | SHA-pinned `actions/checkout`,
`actions/github-script`, etc. |
| **Minimal permissions** | Each job declares only what it needs |
| **Concurrency** | One evaluation per PR, cancels in-progress |
| **Threat detection** | gh-aw built-in threat detection analyzes agent
output |

### Files added/modified

- `.github/workflows/copilot-evaluate-tests.md` — gh-aw workflow source
- `.github/workflows/copilot-evaluate-tests.lock.yml` — Compiled
workflow (auto-generated by `gh aw compile`)
- `.github/skills/evaluate-pr-tests/scripts/Gather-TestContext.ps1` —
Test context gathering script (binary-safe file download, path traversal
protection)
- `.github/instructions/gh-aw-workflows.instructions.md` — Copilot
instructions for gh-aw development

### Known Limitations

**Fork PR evaluation via `/evaluate-tests` comment is not supported in
v1.** The gh-aw platform inserts a `checkout_pr_branch.cjs` step after
all user steps, which may overwrite base-branch skill files restored for
fork PRs. This is a known gh-aw platform limitation — user steps always
run before platform-generated steps, with no way to insert steps after.

**Workaround:** Use `workflow_dispatch` (Actions UI → "Run workflow" →
enter PR number) to evaluate fork PRs. This trigger bypasses the
platform checkout step entirely and works correctly.

**Related upstream issues:**
- [github/gh-aw#18481](github/gh-aw#18481) —
"Using gh-aw in forks of repositories"
- [github/gh-aw#18518](github/gh-aw#18518) —
Fork detection and warning in `gh aw init`
- [github/gh-aw#18520](github/gh-aw#18520) —
Fork context hint in failure messages
- [github/gh-aw#18521](github/gh-aw#18521) —
Fork support documentation

### Fixes

- Fixes dotnet#34602

---------

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Jakub Florkowski <kubaflo123@gmail.com>
## Summary

Enables the copilot-evaluate-tests gh-aw workflow to run on fork PRs by
adding `forks: ["*"]` to the `pull_request` trigger and removing the
fork guard from `Checkout-GhAwPr.ps1`.

## Changes

1. **copilot-evaluate-tests.md**: Added `forks: ["*"]` to opt out of
gh-aw auto-injected fork activation guard. Scoped `Checkout-GhAwPr.ps1`
step to `workflow_dispatch` only (redundant for other triggers since
platform handles checkout).

2. **copilot-evaluate-tests.lock.yml**: Recompiled via `gh aw compile` —
fork guard removed from activation `if:` conditions.

3. **Checkout-GhAwPr.ps1**: Removed the `isCrossRepository` fork guard.
Updated header docs and restore comments to accurately describe behavior
for all trigger×fork combinations (including corrected step ordering).

4. **gh-aw-workflows.instructions.md**: Updated all stale references to
the removed fork guard. Documented `forks: ["*"]` opt-in, clarified
residual risk model for fork PRs, and updated troubleshooting table.

## Security Model

Fork PRs are safe because:
- Agent runs in **sandboxed container** with all credentials scrubbed
- Output limited to **1 comment** via `safe-outputs: add-comment: max:
1`
- Agent **prompt comes from base branch** (`runtime-import`) — forks
cannot alter instructions
- Pre-flight check catches missing `SKILL.md` if fork isn't rebased on
`main`
- No workspace code is executed with `GITHUB_TOKEN` (checkout without
execution)

## Testing

- ✅ `workflow_dispatch` tested against fork PR dotnet#34621
- ✅ Lock.yml statically verified — fork guard removed from `if:`
conditions
- ⏳ `pull_request` trigger on fork PRs can only be verified post-merge
(GitHub Actions reads lock.yml from default branch)

---------

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…taType is compiled (dotnet#34717)

## Description

Adds regression tests for dotnet#34713 verifying the XAML source generator
correctly handles bindings with `Converter={StaticResource ...}` inside
`x:DataType` scopes.

Closes dotnet#34713

## Investigation

After thorough investigation of the source generator pipeline
(`KnownMarkups.cs`, `CompiledBindingMarkup.cs`, `NodeSGExtensions.cs`):

### When converter IS in page resources (compile-time resolution ✅)

`GetResourceNode()` walks the XAML tree, finds the converter resource,
and `ProvideValueForStaticResourceExtension` returns the variable
directly — **no runtime `ProvideValue` call**. The converter is
referenced at compile time.

### When converter is NOT in page resources (runtime resolution ✅)

`GetResourceNode()` returns null → falls through to `IsValueProvider` →
generates `StaticResourceExtension.ProvideValue(serviceProvider)`. The
`SimpleValueTargetProvider` provides the full parent chain, and
`TryGetApplicationLevelResource` checks `Application.Current.Resources`.
The binding IS still compiled into a `TypedBinding` — only the converter
resolution is deferred.

### Verified on both `main` and `net11.0`

All tests pass on both branches.

## Tests added

| Test | What it verifies |
|------|-----------------|
| `SourceGenResolvesConverterAtCompileTime_ImplicitResources` |
Converter in implicit `<Resources>` → compile-time resolution, no
`ProvideValue` |
| `SourceGenResolvesConverterAtCompileTime_ExplicitResourceDictionary` |
Converter in explicit `<ResourceDictionary>` → compile-time resolution,
no `ProvideValue` |
| `SourceGenCompilesBindingWithConverterToTypedBinding` | Converter NOT
in page resources → still compiled to `TypedBinding`, no raw `Binding`
fallback |
| `BindingWithConverterFromAppResourcesWorksCorrectly` × 3 | Runtime
behavior correct for all inflators (Runtime, XamlC, SourceGen) |

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
StephaneDelcroix added a commit that referenced this pull request Mar 30, 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!

### Description of Change

Cherry-pick of #31317 targeting `net11.0`. Corrects style inheritance in
`Style.ApplyCore` by unapplying the base style before reapplying it.

Test updated from NUnit to xUnit patterns (`XamlInflator`, `[Theory]`,
`Assert.Equal`) and renamed from `Issue31280` to `Maui31280` per
conventions.

### Issues Fixed

Fixes #31280

Co-authored-by: Jakub Florkowski <kubaflo123@gmail.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
PureWeen and others added 2 commits March 30, 2026 09:35
<!-- 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!

## Description

Adds arcade inter-branch merge workflow and configuration to automate
merging `net11.0` into the `release/11.0.1xx-preview3` branch.

### Files added

| File | Purpose |
|------|---------|
| `github-merge-flow-release-11.jsonc` | Merge flow config — source
`net11.0`, target `release/11.0.1xx-preview3` |
| `.github/workflows/merge-net11-to-release.yml` | GitHub Actions
workflow — triggers on push to net11.0, daily cron, manual dispatch |

### How it works

Uses the shared [dotnet/arcade inter-branch merge
infrastructure](https://github.com/dotnet/arcade/blob/main/.github/workflows/inter-branch-merge-base.yml):
- **Event-driven**: triggers on push to `net11.0`, with daily cron
safety net
- **ResetToTargetPaths**: auto-resets `global.json`, `NuGet.config`,
`eng/Version.Details.xml`, `eng/Versions.props`, `eng/common/*` to
target branch versions
- **QuietComments**: reduces GitHub notification noise
- Skips PRs when only Maestro bot commits exist

### Incrementing for future releases

When cutting a new release (e.g., preview4), update:
1. `github-merge-flow-release-11.jsonc` → change `MergeToBranch` value
2. `.github/workflows/merge-net11-to-release.yml` → update workflow
`name` field

Follows the same pattern as `merge-main-to-net11.yml` /
`github-merge-flow-net11.jsonc`.

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
<!--
!!!!!!! MAIN IS THE ONLY ACTIVE BRANCH. MAKE SURE THIS PR IS TARGETING
MAIN. !!!!!!!
-->



<!-- Enter description of the fix in this section -->

### 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. -->

Fixes dotnet#33355

### Description of Change

This report has the goal to provide a detailed progress on the solution
of the memory-leak

The test consists doing 100 navigations between 2 pages, as the image
below suggest

<img width="876" height="502" alt="image"
src="https://github.com/user-attachments/assets/e9e80768-dd40-4445-9fc8-90469579236c"
/>


Running the gc-dump on the desired objects this is what I found.

> BaseLine is the dump when the app starts.

| | | | | |
| ------------------------------------ | ------- | ------------ |
-------------- | ------------------------ |
| Object Type | Count | Size (Bytes) | Expected Count | Status |
| **Page2** | **100** | 84,000 | 1 | ❌ **LEAKED** (99 extra) |
| **PageHandler** | **103** | 9,888 | 4 | ❌ **LEAKED** (99 extra) |
| **StackNavigationManager** | **102** | 16,320 | 1 | ❌ **LEAKED** (101
extra) |
| **StackNavigationManager.Callbacks** | **102** | 5,712 | 1 | ❌
**LEAKED** (101 extra) |
| **NavigationViewFragment** | **102** | 5,712 | ~2 | ❌ **LEAKED** (100
extra) |

So the first fix was to call `Disconnect` handler, on the
`previousDetail` during the `FlyoutPage.Detail_set`. The PageChanges and
Navigated events will not see this, since this `set` is not considered a
navigation in .Net Maui.

After that we see the following data

| | | | | |
| ------------------------------------ | ------- | ------------ |
---------------------- | ------------ |
| Object Type | Count | Size (Bytes) | vs Baseline | Status |
| **Page2** | **100** | 84,000 | Same | ❌ **LEAKED** |
| **PageHandler** | **103** | 9,888 | Same | ❌ **LEAKED** |
| **StackNavigationManager** | **102** | 16,320 | Same | ❌ **LEAKED** |
| **StackNavigationManager.Callbacks** | **1** | 56 | ✅ **FIXED!** (was
102) | ✅ **Good!** |
| **NavigationViewFragment** | **102** | 5,712 | Same | ❌ **LEAKED** |

So, calling the Disconnect handler will fix the leak at
`StackNavigationManager.Callbacks`. Next step was to investigate the
`StackNavigationManager` and see what's holding it.

On `StackNavigationManager` I see lot of object that should be cleaned
up in order to release other references from it. After cleaning it up
the result is

|   |   |   |   |   |
|---|---|---|---|---|
|Object Type|Count|Size (Bytes)|vs Baseline|Status|
|**Page2**|**1**|840|✅ **FIXED!** (was 100)|✅ **Perfect!**|
|**PageHandler**|**4**|384|✅ **FIXED!** (was 103)|✅ **Perfect!**|
|**StackNavigationManager**|**102**|16,320|❌ Still leaking (was 102)|❌
**Unchanged**|
|**StackNavigationManager.Callbacks**|**1**|56|✅ Fixed (was 102)|✅
**Good!**|
|**NavigationViewFragment**|**102**|5,712|❌ Still leaking (was 102)|❌
**Unchanged**|

So something is still holding the `StackNavigationManager` and
`NavigationViewFragment` so I changed the approach and found that
`NavigationViewFragment` is holding everything and after fixing that,
cleaning it up on `Destroy` method. here's the result

|   |   |   |   |   |
|---|---|---|---|---|
|Object Type|Count|Size (Bytes)|vs Previous|Status|
|**Page2**|**1**|840|✅ Same|✅ **Perfect!**|
|**PageHandler**|**4**|384|✅ Same|✅ **Perfect!**|
|**StackNavigationManager**|**1**|160|🎉 **FIXED!** (was 102)|🎉
**FIXED!**|
|**StackNavigationManager.Callbacks**|**1**|56|✅ Same|✅ **Perfect!**|
|**NavigationViewFragment**|**102**|5,712|⚠️ Still present|⚠️
**Remaining**|

With that there's still the leak of `NavigationViewFragment`, looking at
the graph the something on Android side is holding it, there's no root
into managed objects, as far the gcdump can tell. I tried to cleanup the
`FragmentManager`, `NavController` and so on but without success (maybe
I did it wrong).

There's still one instance of page 2, somehow it lives longer, I don't
think it's a leak object because since its value is 1. For reference the
Page2 graph is
```
Page2 (1 instance)
 └── PageHandler
     └── EventHandler<FocusChangeEventArgs>
         └── IOnFocusChangeListenerImplementor (Native Android)
             └── UNDEFINED

```


Looking into www I found that android caches those Fragments, sadly in
our case we don't reuse them. The good part is each object has only 56
bytes, so it shouldn't be a big deal, I believe we can take the
improvements made by this PR and keep an eye on that, maybe that's fixed
when moved to Navigation3 implementation.
@kubaflo
Copy link
Copy Markdown
Contributor Author

kubaflo commented Apr 1, 2026

Code Review — PR #31317

Independent Assessment

What this changes: In Style.ApplyCore, adds an ((IStyle)basedOn).UnApply(bindable) call before ((IStyle)basedOn).Apply(...) so that when a style inheritance chain is (re-)applied, stale property values from a previous base-style application are cleared before the fresh application. Adds a XAML unit test exercising a 3-level style chain with AppThemeBinding in the root style.

Inferred motivation: When a derived style overrides a property set by a base style that uses AppThemeBinding, the override gets lost during re-application because the base style's old values aren't cleared first.

Reconciliation with PR Narrative

Author claims: "Corrects style inheritance in Style.ApplyCore by unapplying the base style before reapplying it." Fixes #31280.
Agreement: The core fix matches the claim and the approach is sound. Calling UnApply before Apply ensures a clean slate. UnApply on a never-applied style is safe (no-ops through specificities.TryGetValue guard). Reviewer @StephaneDelcroix requested renaming to Maui31280 and migration to xUnit — the renaming was done but the xUnit migration has issues.

Findings

❌ Critical — Wrong using import causes compilation failure

File: Maui31280.xaml.cs:3
The test file imports using NUnit.Framework; but uses xUnit APIs throughout: [Collection("Issue")] (xUnit), [Theory] (ambiguous), and Assert.Equal (xUnit — NUnit uses Assert.AreEqual). This project uses xUnit. Should be using Xunit;. This is almost certainly the cause of the CI build failures.

⚠️ Important — Missing AppThemeBinding test infrastructure setup

File: Maui31280.xaml.cs:14-16 (the Tests class)
The XAML uses {AppThemeBinding Light=Green, Dark=Green} but the test class doesn't set up the required infrastructure. Every other test in this project that uses AppThemeBinding (e.g., Maui23201.xaml.cs, Maui19388.xaml.cs) follows this pattern:

[Collection("Issue")]
public class Tests : IDisposable
{
    public Tests()
    {
        Application.SetCurrentApplication(new MockApplication());
        DispatcherProvider.SetCurrent(new DispatcherProviderStub());
    }

    public void Dispose()
    {
        AppInfo.SetCurrent(null);
        Application.SetCurrentApplication(null);
        DispatcherProvider.SetCurrent(null);
    }
    // ...
}

Without this, the AppThemeBinding may not resolve correctly, meaning the test could pass for the wrong reasons or fail unexpectedly.

⚠️ Minor — Missing newlines at end of files

Files: Maui31280.xaml and Maui31280.xaml.cs
Both files are missing the trailing newline (\ No newline at end of file in diff). All other files in this directory have trailing newlines.

Devil's Advocate

  • Is the UnApply-before-Apply safe on first application? Yes — UnApplyCore early-returns via specificities.TryGetValue when the style was never applied, and _targets.Remove is a no-op. Safe.
  • Re-entrancy concern? UnApply calls RemoveDynamicResource which could trigger OnBasedOnResourceChanged, potentially causing recursive unapply/apply. However, OnBasedOnResourceChanged looks up this style (the parent) from (bindable as IStyleElement).Style, and the _targets.TryGetValue guard should prevent double-processing. The multi-model review (5 models) found no confirmed bug here.
  • Should this target net11.0 instead of main? @StephaneDelcroix flagged this: "Style changes have a tendency to create regressions, I'd recommend targeting net11.0." This is a process/risk decision for the team. The fix is minimal (2 lines), but the Style system is notoriously fragile.

Verdict: NEEDS_CHANGES

Confidence: high
Summary: The core 2-line fix in Style.cs is correct and well-motivated. However, the test file has a wrong using import (NUnit.Framework instead of Xunit) that likely causes CI build failures, and it's missing the required AppThemeBinding infrastructure setup (MockApplication, DispatcherProvider) that all similar tests in the project use. Fix these two issues and CI should go green.

@kubaflo
Copy link
Copy Markdown
Contributor Author

kubaflo commented Apr 1, 2026

🟡 .NET MAUI Review - Changes Suggested

Expand Full Review - 7b440a8 - Improved style inheritance

Independent Assessment

What this changes: In Style.ApplyCore, adds an ((IStyle)basedOn).UnApply(bindable) call before ((IStyle)basedOn).Apply(...) so that when a style inheritance chain is (re-)applied, stale property values from a previous base-style application are cleared before the fresh application. Adds a XAML unit test exercising a 3-level style chain with AppThemeBinding in the root style.

Inferred motivation: When a derived style overrides a property set by a base style that uses AppThemeBinding, the override gets lost during re-application because the base style's old values aren't cleared first.

Reconciliation with PR Narrative

Author claims: "Corrects style inheritance in Style.ApplyCore by unapplying the base style before reapplying it." Fixes #31280.
Agreement: The core fix matches the claim and the approach is sound. Calling UnApply before Apply ensures a clean slate. UnApply on a never-applied style is safe (no-ops through specificities.TryGetValue guard). Reviewer @StephaneDelcroix requested renaming to Maui31280 and migration to xUnit — the renaming was done but the xUnit migration has issues.

Findings

❌ Critical — Wrong using import causes compilation failure

File: Maui31280.xaml.cs:3
The test file imports using NUnit.Framework; but uses xUnit APIs throughout: [Collection("Issue")] (xUnit), [Theory] (ambiguous), and Assert.Equal (xUnit — NUnit uses Assert.AreEqual). This project uses xUnit. Should be using Xunit;. This is almost certainly the cause of the CI build failures.

⚠️ Important — Missing AppThemeBinding test infrastructure setup

File: Maui31280.xaml.cs:14-16 (the Tests class)
The XAML uses {AppThemeBinding Light=Green, Dark=Green} but the test class doesn't set up the required infrastructure. Every other test in this project that uses AppThemeBinding (e.g., Maui23201.xaml.cs, Maui19388.xaml.cs) follows this pattern:

[Collection("Issue")]
public class Tests : IDisposable
{
    public Tests()
    {
        Application.SetCurrentApplication(new MockApplication());
        DispatcherProvider.SetCurrent(new DispatcherProviderStub());
    }

    public void Dispose()
    {
        AppInfo.SetCurrent(null);
        Application.SetCurrentApplication(null);
        DispatcherProvider.SetCurrent(null);
    }
    // ...
}

Without this, the AppThemeBinding may not resolve correctly, meaning the test could pass for the wrong reasons or fail unexpectedly.

⚠️ Minor — Missing newlines at end of files

Files: Maui31280.xaml and Maui31280.xaml.cs
Both files are missing the trailing newline (\ No newline at end of file in diff). All other files in this directory have trailing newlines.

Devil's Advocate

  • Is the UnApply-before-Apply safe on first application? Yes — UnApplyCore early-returns via specificities.TryGetValue when the style was never applied, and _targets.Remove is a no-op. Safe.
  • Re-entrancy concern? UnApply calls RemoveDynamicResource which could trigger OnBasedOnResourceChanged, potentially causing recursive unapply/apply. However, OnBasedOnResourceChanged looks up this style (the parent) from (bindable as IStyleElement).Style, and the _targets.TryGetValue guard should prevent double-processing. The multi-model review (5 models) found no confirmed bug here.
  • Should this target net11.0 instead of main? @StephaneDelcroix flagged this: "Style changes have a tendency to create regressions, I'd recommend targeting net11.0." This is a process/risk decision for the team. The fix is minimal (2 lines), but the Style system is notoriously fragile.

Verdict: NEEDS_CHANGES

Confidence: high
Summary: The core 2-line fix in Style.cs is correct and well-motivated. However, the test file has a wrong using import (NUnit.Framework instead of Xunit) that likely causes CI build failures, and it's missing the required AppThemeBinding infrastructure setup (MockApplication, DispatcherProvider) that all similar tests in the project use. Fix these two issues and CI should go green.

Corrects style inheritance in Style.ApplyCore by unapplying the base
style before reapplying it, fixing multi-level style property override.

- Adds UnApply call before Apply in ApplyCore when basedOn is not null
- Adds XAML unit test verifying multi-level style inheritance
- Uses xUnit test patterns (XamlInflator, [Theory], Assert.Equal)

Fixes dotnet#31280

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
@kubaflo
Copy link
Copy Markdown
Contributor Author

kubaflo commented Apr 1, 2026

/azp run maui-pr-uitests maui-pr-devicetests

@azure-pipelines
Copy link
Copy Markdown

No pipelines are associated with this pull request.

@kubaflo kubaflo changed the base branch from main to inflight/current April 1, 2026 16:12
@kubaflo kubaflo merged commit c9d0fc3 into dotnet:inflight/current Apr 1, 2026
7 of 17 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

community ✨ Community Contribution s/agent-changes-requested AI agent recommends changes - found a better alternative or issues s/agent-reviewed PR was reviewed by AI agent workflow (full 4-phase review)

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Styles based on a style that is based on another style that uses AppThemeBinding do not inherit properties correctly.

6 participants