CS0220 on uint.MaxValue + (nuint)42

[lostmsu]

Version Used: 3.10.0-1.21125.6 (93b8af2)

Steps to Reproduce: uint.MaxValue + (nuint)42

Expected Behavior:
uint + nuint => promote uint to nuint, then add.

Actual Behavior: error CS0220: The operation overflows at compile time in checked mode

Workaround:

uint uintMax = uint.MaxValue;
nuint size = uintMax + (nuint)42;

[Youssef1313]

Expected Behavior:
uint + nuint => promote uint to nuint, then add.

• There is no implicit conversion between uint to nuint per spec. Only explicit conversion exists.
• I can't see how converting uint to nuint would prevent the overflow from happening.
• Your workaround doesn't prevent the overflow from happening. The overflow would happen, but the compiler doesn't report it because uintMax isn't a compile-time constant, and there is no kind of data flow analysis done here.

[cston]

Thanks for reporting @lostmsu.

An error is reported for uint.MaxValue + (nuint)42 because the expression overflows when nuint is 32 bits.

As @Youssef1313 described, the second example avoids the compile time error, but will still overflow at runtime when executed on a 32-bit platform.

Closing as "by design".

[lostmsu]

@cston this would make sense, if nuint was always 32 bit, and I was trying to define a const nuint field or variable. But I am just trying to pass that value to a function, that expects nuint.

In fact, this directly contradicts the spec: https://docs.microsoft.com/en-us/dotnet/csharp/language-reference/proposals/csharp-9.0/native-integers#constants

Constant folding operations are evaluated with Int32 and UInt32 operands rather than native ints for consistent behavior regardless of compiler platform. If the operation results in a constant value in 32-bits, constant folding is performed at compile-time. Otherwise the operation is executed at runtime and not considered a constant.

My take on spec here is that because the result might be an UInt32 overflow (an might not be, depending on target platform), it is illegal for the C# compiler to force constant folding on that + op.

[cston]

@lostmsu,

this would make sense, if nuint was always 32 bit, and I was trying to define a const nuint field or variable. But I am just trying to pass that value to a function, that expects nuint.

Good point, thanks.

From the LDM notes, the compiler should report a warning not an error:

Whenever we can safely produce a constant value under 32-bits, we do constant folding. Otherwise, the result is non-constant, and under checked, the code produces a warning and the result is non-constant.

I'll investigate.