Cannot suppress obsolete or experimental errors that use a custom diagnostic ID

[anthonymastreanvae]

Description

I cannot suppress obsolete or experimental errors that use a custom diagnostic ID with a local or global suppress message attribute. Only preprocessor directives and Global AnalyzerConfig seems to work... but I can't always use those options (for example in some compliance & regulatory situations).

Reproduction Steps

using System.Diagnostics.CodeAnalysis;

namespace N;

[Experimental("EXAMPLE0001")]
public class A
{
}

[Obsolete("B is obsolete", DiagnosticId = "EXAMPLE0002")]
public class B
{
}

public class C
{
    [SuppressMessage("", "EXAMPLE0001")]
    private readonly A c;

    [SuppressMessage("", "EXAMPLE0002")]
    private readonly B b;
}

Expected behavior

I expect to be able to use the SuppressMessageAttribute locally or globally to suppress the custom diagnostic ID.

Actual behavior

I am unable to use the SuppressMessageAttribute locally or globally to suppress the custom diagnostic ID.

Regression?

No response

Known Workarounds

No response

Configuration

No response

Other information

No response

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/area-system-runtime
See info in area-owners.md if you want to be subscribed.

[KalleOlaviNiemitalo]

SuppressMessageAttribute and UnconditionalSuppressMessageAttribute do not affect the standard obsoletion warning CS0618, either. So the behaviour with custom obsoletions is consistent with that.

The SuppressMessageAttribute documentation says

Does not apply to compiler diagnostics.

so, this seems to be by design.

[anthonymastreanvae]

Does not apply to compiler diagnostics.

Sure, but applying a custom diagnostic ID means it's no longer a "compiler diagnostic", right? I only recently uncovered some documentation on these attributes...

It is also possible to suppress the diagnostic by usual means, such as an explicit compiler option or #pragma.

It's not clear that "by usual means" does NOT include SuppressMessageAttribute.

https://learn.microsoft.com/en-us/dotnet/csharp/language-reference/proposals/csharp-12.0/experimental-attribute#reported-diagnostic

[KalleOlaviNiemitalo]

Anyhow, I don't think there is anything that can be done in dotnet/runtime to change this behaviour. This issue should be moved to dotnet/roslyn.

[anthonymastreanvae]

Anyhow, I don't think there is anything that can be done in dotnet/runtime to change this behaviour. This issue should be moved to dotnet/roslyn.

Ah, thank you... I was uncertain where this behavior was defined.

[KalleOlaviNiemitalo]

dotnet/runtime defines the attribute classes so that they can be used in programs, but it does not control what effect the attributes have at compile time.

dotnet/roslyn contains the C# compiler in which the warnings and suppression checks are implemented.

dotnet/csharplang is where C# language changes are designed. The feature speclets are maintained there. If the speclet said that custom obsoletions must not be suppressible using attributes, then I think changing that would require an issue in dotnet/csharplang to update the speclet and another issue in dotnet/roslyn to implement it.

dotnet/csharpstandard is where the C# standard is drafted so that ECMA can approve it. This is lagging badly behind dotnet/csharplang; the latest ECMA standard is for C# 7.3. The current drafts are for C# 8 and C# 9. Although the ObsoleteAttribute.DiagnosticId property was introduced in .NET 5, which uses C# 9, that feature is not in the C# 9 draft yet. Therefore, dotnet/csharpstandard does not yet specify whether custom obsoletions should be suppressible using attributes, and it's too early to file issues about that there.

[anthonymastreanvae]

Can someone who is a member of both teams move the bug or do I need to open a new one in dotnet/roslyn?