LDM Notes for November 12th, 2025

[333fred]

https://github.com/dotnet/csharplang/blob/main/meetings/2025/LDM-2025-11-12.md

Agenda

• Unsafe evolution

[huoyaoyuan]

I'm still very concerned about the entire feature on unsafe propagation.

public unsafe void Method1(int[] array)
{
    fixed (int* ptr = array)
    {
        // proceed with the pointer
    }
}

void Method2()
{
    int[] array = WhatEver();
    Method1(array);
}

If I'm understanding correctly, calling Method1 will require unsafe context since the method is marked as unsafe. People usually tend to put the unsafe modifier at method level to reduce new scopes within methods. If Method2 is in a different codebase, then the author of Method1 may not be aware that it's a source breaking change for all its callers, and the correct fix is to change Method1.

[333fred]

People usually tend to put the unsafe modifier at method level to reduce new scopes within methods.

You're not wrong, but this is a general problem. We are making a breaking change, and when you opt into the new rules, you will need to fix your methods.

[ds5678]

I notice that the language design team seems much more open to breaking changes now than in past years (C# 12 and earlier). What caused this shift?

[CyrusNajmabadi]

The breaks follow our stated philosophy around when it can be ok. We've never been "no breaks ever" (there have been beans going back decades). It's always been: the value of the break had to be worth it.