Another overlooked detail in the primary constructor's decision process

[sonyps5201314]

Input code

    record WebPair7(string name, string? value, ref readonly object encode)
    {
        public string? Value { get; } = name;
        public string Name { get; } = value;
        string? WebValue { get; } = name != null ? "111" : value;
        string? WebValue2;
    }

Erroneous output

// TestClassPropertyInit, Version=1.0.0.0, Culture=neutral, PublicKeyToken=null
// TestClassPropertyInit.WebPair7
using System.Runtime.CompilerServices;

internal record WebPair7
{
	public string name { get; init; }

	public string? value { get; init; }

	public object encode { get; init; }

	public string? Value { get; }

	public string Name { get; }

	private string? WebValue { get; }

	private string? WebValue2;

	public WebPair7(string name, string? value, ref readonly object encode)
	{
		this.name = name;
		this.value = value;
		this.encode = encode;
		Value = name;
		Name = value;
		WebValue = ((name != null) ? "111" : value);
		base..ctor();
	}

	[CompilerGenerated]
	public void Deconstruct(out string name, out string? value, out object encode)
	{
		name = this.name;
		value = this.value;
		encode = this.encode;
	}
}

The line base..ctor(); appeared.

Details

This commit in my original PR actually included this example and provided a fix, but it seems you might have overlooked it. Based on my local testing, even after your recent 10+ revisions regarding primary constructors, the current implementation is still not as robust as what was in my original PR.

Therefore, I strongly recommend once again that you review the individual commit history of my PR. There are many details in my code that you haven't captured yet, which I have also explained multiple times in previous GitHub Issue comments.

Thank you. Perhaps it is not my place to offer advice, and I realize this might be hard to accept emotionally, but I simply wish for ILSpy to be as perfect as possible.

[sonyps5201314]

@siegfriedpammer I initially intended to find a counter-example to demonstrate that my original logic handled scenarios that your latest changes might have overlooked. However, I unexpectedly discovered a new issue instead: using your latest code, ILSpy throws an error when the C# language version is switched to 8.0 or lower:

I'll wait for this to be fixed before I proceed with further verification.

[siegfriedpammer]

If you find any other issues, please create a new issue to keep everything clean. Thanks for your understanding.