#error version

[The-Futurist]

I added this to a small test project in VS 2026 insiders (up to date) to see what specific version of C# is being used, but it behaves pretty weird:

It does show me the version but it also seems to say that using #error is itself an error, as if using it is an error.

It also underlines version as somehow being wrong:

But it understands it fine because it dies report the C# version number in the error window!

[CyrusNajmabadi]

Yes. That's the entire point of it.

but it also seems to say that using #error is itself an error,

Correct. You used #error which is literally the way of saying "this code is in error".

It also underlines version as somehow being wrong:

Yes. That's where it reports the error diagnostic.

[CyrusNajmabadi]

And, regardless, this has been appropriately closed. The lang/compiler feature is working entirely as intended. Going another route would just be 'make work' that wouldn't actually provide benefit.

[The-Futurist]

And, regardless, this has been appropriately closed. The lang/compiler feature is working entirely as intended. Going another route would just be 'make work' that wouldn't actually provide benefit.

I understand that it isn't going to change, the replies from Microsoft here make that crystal clear, I'm no longer interested in seeing this improve because the matter has been - it seems - closed.

Having said that, your arguments that it is entirely reasonable, being used as originally intended and so on are laughable and deserve to be challenged.

In the problem domain there is nothing erroneous about reporting the compiler's version information, it is informational. Causing compilation to fail with an error is not a desirable side effect of reporting version information surely?

There is already a #pragma which could easily have been leveraged for something like reporting compile time technical information.

So go ahead, continue to use #error but don't jump through hoops trying to present a bad idea as a good idea, it's a hack and why can't you simply admit that?

[CyrusNajmabadi]

Causing compilation to fail with an error is not a desirable side effect

Yes. It is the main purpose. That's why error was chosen. Not warning. It's not suppressible. It's a primary need of the feature and why we approved it.

It also matches how error works for all other cases. It's always an error.

If we wanted it to be fail, we would have made this part of #warning instead.

There is already a #pragma

Pragmas don't fail. You can have whatever you want after it. So we can't error here as that would break potential existing usages by the ecosystem.

This was considered and explicitly rejected.

it's a hack

Anything else used would have to replicate this behavior at much higher cost. Reusing existing compiler pieces for their intended purpose is not a hack in my mind. But you do you :-)

[The-Futurist]

Causing compilation to fail with an error is not a desirable side effect

Yes. It is the main purpose. That's why error was chosen. Not warning. It's not suppressible. It's a primary need of the feature and why we approved it.

Why would anyone want compilation to fail when all one wants is for the compiler to report it's version number? Sound's quite ridiculous.

It also matches how error works for all other cases. It's always an error.

Using #error is fine for errors but not well suited to reporting a compiler's version number, that's a hack.

If we wanted it to be fail, we would have made this part of #warning instead.

There is already a #pragma

Pragmas don't fail. You can have whatever you want after it.

Exactly, sounds ideal for reporting version info then.

So we can't error here as that would break potential existing usages by the ecosystem.

Causing compilation to fail and reporting a version number are distinct requirements.

This was considered and explicitly rejected.

it's a hack

Anything else used would have to replicate this behavior at much higher cost. Reusing existing compiler pieces for their intended purpose is not a hack in my mind. But you do you :-)

The hack is relying on a directive that causes compilation to always fail, for reporting compiler version. There's no relationship I can see between these to concepts.

[333fred]

We're done here. Locking the thread to prevent further derailment. The compiler is working as intended and needs no further changes.