SA1600 is not raised for record primary constructors that are missing param tags

[SapiensAnatis]

Today, the following code will not cause any Stylecop diagnostics to be raised:

/// <summary>
/// Record.
/// </summary>
public record MyRecord(int Parameter);

This is in spite of the missing <param> tag for Parameter. Stylecop should report a diagnostic unless this is included:

/// <summary>
/// Record.
/// </summary>
/// <param name="Parameter">Parameter.</param>
public record MyRecord(int Parameter);

Because Parameter declares a property as well as a constructor parameter, it was agreed in #3770 that it should report SA1600 instead of the usual SA1611 for missing <param> tags.

[SapiensAnatis]

I've made some initial progress towards tackling this issue, but am currently dealing with a strange problem where my test is reporting each diagnostic twice 😔

I get this output for a record like the one in the example above that declares a single undocumented parameter:

Microsoft.CodeAnalysis.Testing.Verifiers.EqualWithMessageException : Context: Diagnostics of test state
Mismatch between number of diagnostics returned, expected "1" actual "2"

Diagnostics:
// /0/Test0.cs(5,28): warning SA1600: Elements should be documented
VerifyCS.Diagnostic().WithSpan(5, 28, 5, 34),
// /0/Test0.cs(5,28): warning SA1600: Elements should be documented
VerifyCS.Diagnostic().WithSpan(5, 28, 5, 34),

I won't make this a huge comment containing everything I've done (I could open a draft PR if need be) but it seems when I debug the tests all my breakpoints are hit twice. This seems to happen with other analyzers as well. I was wondering if that could be the cause, or is this normal when testing analyzers?

[sharwell]

I think there are some versions of Roslyn where callbacks execute twice, resulting in duplicate diagnostics. You can include the diagnostic twice in the expected results and I can investigate further.