Normative use of a .NET constant not required by this C# spec

[RexJaeschke]

[This issue was spun-off from Issue #1378 sub-issue 7.]

In 23.5.6.2, "The CallerLineNumber attribute,", we have the following:

The attribute System.Runtime.CompilerServices.CallerLineNumberAttribute is allowed on optional parameters when there is a standard implicit conversion (§10.4.2) from the constant value int.MaxValue to the parameter’s type. This ensures that any non-negative line number up to that value can be passed without error.

However, int.MaxValue is not part of this specification. How might we reword this normative text to get around that? Or might we add the minima/maxima constants for the built-in types?

[KalleOlaviNiemitalo]

"from the constant value 0x7fffffff" perhaps?

[Nigel-Ecma]

Maybe we should stick with base-10, §8.3.6 defines the maximum int value as 2147483647 and its representation as 2’s complement – which is of course @KalleOlaviNiemitalo’s 0x7fffffff.

Maybe just simplify:

The attribute System.Runtime.CompilerServices.CallerLineNumberAttribute is allowed on optional parameters when there is a standard implicit conversion (§10.4.2) from the maximum value of int (§8.3.6) to the parameter’s type.

Should the wording be reconsidered? The text requires “a standard implicit conversion”, which excludes user-defined conversions leaving the pre-defined ones, which “always succeed and never cause exceptions” (§10.2.1); and also user-defined types as the parameter type. So any error this wording “ensures” cannot happen must be a compile time one, and given the restriction on conversions that can only come from an implicit constant conversion (§10.2.11) which support conversions from constant expressions of type int to smaller types iff there is no loss of magnitude. In other words the restriction is the parameter type cannot be sbyte, byte, short or ushort?

If that logic is correct we can consider:

The attribute System.Runtime.CompilerServices.CallerLineNumberAttribute is allowed on optional parameters of any integral type (§8.3.6) capable of representing all non-negative values of int (§8.3.6).

(non-negative rather than postive to include zero)

Which is a least short, but not short :-)