Are there any plans for CsWin32 to support the new C# memory safety features?

[littletoxic]

https://devblogs.microsoft.com/dotnet/improving-csharp-memory-safety/

How will the upcoming C# memory safety improvements affect CsWin32?

[jevansaks]

Good question! Things should be mostly unchanged, but there may be more cases where opted-in callers will have to do the safety attestations. We might need to change the codegen to help as well, since there are places where we put unsafe on the generated methods even though they're not unsafe by the new rules.

If someone wanted to scout ahead and see what opt-ing in cswin32 to the new mode looks like, I would be interested to see the results. :)

[yhcedpn]

It appears that .NET 11 primarily restricts unsafe regions to operations like pointer dereferencing, thereby narrowing the scope covered by the unsafe keyword; propagating pointers no longer seems to be considered unsafe. Additionally, the inclusion of <safety> XML comments in the code is required.

This is highly beneficial for CsWin32, and I look forward to the maintainers adding a milestone for updating to the .NET 11 safety model.

[yhcedpn]

It might require modifying some code and adding <safety> comments, but I think the workload should be minimal.

PS: I heard that dotnet format can automatically apply this update—is that true?

[jevansaks]

Well, it's different. Yes, pointer de-referencing is inherently unsafe but the responsible thing to do is to make sure that any methods you make which do unsafe things that require the caller to think about safety should also be unsafe. We probably need to think a little harder about where we can remove unsafe and where it should be added.

[yhcedpn]

Sounds like this requires professional developers to check each method, figure out their call chains, and decide the safe and unsafe boundaries.

Anyway, I’m really looking forward to the project updating to .NET 11!