Avoid reporting CA1823 for inline arrays

[sharwell]

Fixes #6789

[stephentoub]

Does this need to be updated whenever the repo updates the version it depends on preview to preview?

[sharwell]

Not really. Normally the signatures in the reference assemblies don't change to the point of affecting the unit tests so they just keep working. Plus, it'll end up in dotnet/roslyn-sdk soon.

[mavasani]

Another potential approach we take for tests is to add an additional source file to the test with custom stub implementation of the newer APIs added to the Framework which are not yet available via the currently referenced assemblies. We can do a sweeping cleanup later once we move the tests to reference packages with required APIs.

[sharwell]

Another potential approach we take for tests is to add an additional source file to the test with custom stub implementation of the newer APIs

This is necessary when the APIs haven't been published, but in this case we can reference the real APIs and avoid the risk involved in cases where a deviation in our test implementation impacts the behavior of the analyzer.

[mavasani]

What is the difference between this extension, and the one right below?

roslyn-analyzers/src/Utilities/Compiler/Extensions/ISymbolExtensions.cs

Lines 662 to 680 in 1d9fd4e

	/// <summary>
	/// Returns a value indicating whether the specified symbol has the specified
	/// attribute.
	/// </summary>
	/// <param name="symbol">
	/// The symbol being examined.
	/// </param>
	/// <param name="attribute">
	/// The attribute in question.
	/// </param>
	/// <returns>
	/// <see langword="true"/> if <paramref name="symbol"/> has an attribute of type
	/// <paramref name="attribute"/>; otherwise <see langword="false"/>.
	/// </returns>
	/// <remarks>
	/// If <paramref name="symbol"/> is a type, this method does not find attributes
	/// on its base types.
	/// </remarks>
	public static bool HasAttribute(this ISymbol symbol, [NotNullWhen(returnValue: true)] INamedTypeSymbol? attribute)

[sharwell]

➡️ I'll send a follow-up to address this.

[sharwell]

➡️ To answer the question, the previous HasAttribute method was delegating to the params form of HasAnyAttribute. The new method avoids the allocation that all calls to HasAttribute were facing.

[mavasani]

Changes LGTM, though we should consolidate all the HasXXXAttribute helpers instead of adding a new one.