[Andrew Leader (andrewleader)]

Description

Adding the log files needed for developers to capture Windows ML logs for diagnostic purposes. The instructions for using these logs will be published on Microsoft Learn after this PR is merged. Microsoft Learn docs are already authored and tested and reviewed.

Target Release

Immediately (applicable to 1.8).

Checklist

Note that /azp run currently isn't working for this repo.

• Samples build and run using the Visual Studio versions listed in the Windows development docs.
• Samples build and run on all supported platforms (x64, x86, ARM64) and configurations (Debug, Release).
• Samples set the minimum supported OS version to Windows 10 version 1809.
• Samples build clean with no warnings or errors.
• [For new samples]: Samples have completed the sample guidelines checklist and follow standardization/naming guidelines.
• Microsoft employees only: you can validate your changes by following the instructions here: https://www.osgwiki.com/wiki/WindowsAppSDK-Samples

[Copilot]

Pull request overview

Adds a new Samples/WindowsML/capture-logs diagnostics drop (profiles + helper script) and links it from the Windows ML sample docs so developers can capture Windows ML/ONNX diagnostic traces and generate a summarized rundown log.

Changes:

• Adds a WPR profile (WinML.wprp) for capturing Windows ML + ONNX/ORT provider events.
• Adds a PowerShell helper (Get-WinMLRundown.ps1) to export ETL data via WPA exporter and generate a consolidated log.
• Updates repo documentation to surface the new “Capture Logs” resource under Windows ML.

Reviewed changes

Copilot reviewed 5 out of 6 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
Samples/WindowsML/capture-logs/WinML.wprp	New WPR profile defining providers/collectors for WinML/ONNX/ORT diagnostics.
Samples/WindowsML/capture-logs/README.md	Documents the capture-logs artifacts and points to Microsoft Learn instructions.
Samples/WindowsML/capture-logs/Get-WinMLRundown.ps1	Exports ETL → CSV and merges/renames fields into a single rundown log.
Samples/WindowsML/README.md	Adds a “Diagnostics” section linking to capture-logs.
README.md	Adds a top-level link to the Windows ML capture-logs resource.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Invoke-WebRequest -UseBasicParsing is not supported in PowerShell 6+/7+ (the parameter was removed), which will cause this script to fail when run under modern PowerShell. If you need compatibility with both Windows PowerShell 5.1 and PowerShell 7, gate the parameter based on availability or avoid it entirely.

Suggested change

	Invoke-WebRequest -Uri $adkUrl -OutFile $installerPath -UseBasicParsing
	if ($PSVersionTable.PSVersion.Major -lt 6) {
	Invoke-WebRequest -Uri $adkUrl -OutFile $installerPath -UseBasicParsing
	} else {
	Invoke-WebRequest -Uri $adkUrl -OutFile $installerPath
	}

[Copilot]

The script invokes wpaexporter.exe but doesn’t check whether it succeeded (e.g., via $LASTEXITCODE) before continuing. If export fails, the script falls through to “No CSV files found.” and exits, which can mask the underlying failure; consider checking the exit code and writing a clearer error + returning a non-zero exit code.

Suggested change

	# Step 2: Get all CSV files
	$csvFiles = Get-ChildItem -Path $OutputFolder -Filter *.csv
	if ($csvFiles.Count -eq 0) {
	Set-Content -Path $LogFilePath -Value "No CSV files found."
	exit
	if ($LASTEXITCODE -ne 0) {
	$errorMessage = "wpaexporter.exe failed with exit code $LASTEXITCODE."
	Write-Host $errorMessage
	if ($LogFilePath) {
	Set-Content -Path $LogFilePath -Value $errorMessage
	}
	exit $LASTEXITCODE
	}
	# Step 2: Get all CSV files
	$csvFiles = Get-ChildItem -Path $OutputFolder -Filter *.csv
	if ($csvFiles.Count -eq 0) {
	Set-Content -Path $LogFilePath -Value "No CSV files found after export."
	exit 1

[Copilot]

Several error paths call exit without an explicit non-zero status (e.g., when no CSV files/data/Process column are found). This can cause CI/scripts to treat failures as success. Use a non-zero exit code for these failure cases (or throw) so callers can reliably detect failure.

Suggested change

	exit
	}
	if (-not ($allData | Get-Member -Name "Process")) {
	Set-Content -Path $LogFilePath -Value "No 'Process' column found."
	exit
	exit 1
	}
	if (-not ($allData | Get-Member -Name "Process")) {
	Set-Content -Path $LogFilePath -Value "No 'Process' column found."
	exit 1

[Copilot]

The script downloads an installer (via fwlink) and runs a silent install (/quiet) of the Windows ADK feature. That’s a high-risk pattern for a sample utility because it executes downloaded code without any explicit user confirmation or integrity verification. Consider removing auto-install and requiring manual install, or at minimum add an explicit opt-in prompt and validate the downloaded binary (e.g., Authenticode signature/publisher) before executing it.