[Regression] IDE0031: False positive on .NET 11 Preview1

[bkoelman]

Version Used:
.NET 11 Preview 1 with VS Insiders 2026 [11513.90]

This is a regression. The problem does not occur using stable .NET 10 and stable VS 2026.

Steps to Reproduce:
A suggestion is offered to convert:

if (_session != null)
{
    _session.Disposed -= SessionOnDisposed;
    _session = null;
}

to:

_session?.Disposed -= SessionOnDisposed;
_session = null;

Diagnostic Id:
IDE0031: Null check can be simplified

Expected Behavior:
No suggestion to change the existing code.

Actual Behavior:
A suggestion is offered. We've increased the severity to Warning in Release builds. Combined with warnaserror, this breaks the build.

[CyrusNajmabadi]

I'm not understanding why this is being described as a false positive. This is the pattern that analyzer looks for.

[bkoelman]

Why is the changed code better? It simplifies the C# code, at the cost of lower runtime performance. It doesn't eliminate branching (in the context of branch prediction in speculative execution), while introducing an assignment that is no longer conditional.

[CyrusNajmabadi]

Why is the changed code better? It simplifies the C# code,

That's exactly the purpose of it.

If you don't like that, you can disable the rule. :-)

[bkoelman]

I like how the rule used to work, just not this new aspect. I wonder if suggestions should be offered that degrade runtime performance or introduce additional execution. I also find the resulting code less readable.

@stephentoub Would you consider my concern justified?

[CyrusNajmabadi]

I wonder if suggestions should be offered that degrade runtime performance

Many rules do this. It's a non-goal for ide analyzers to have no runtime impact. If that is problematic, the rule can be disabled or suppressed. However, this is very by design.

Note: the ide analyzers have not and likely never will be concerned with the cost of an assignment.

You want far more strict analyzers for that.