https://issues.dlang.org/show_bug.cgi?id=14950
Kenji Hara <k.hara...@gmail.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Keywords|rejects-valid |diagnostic
Hardware|x86_64 |All
OS|Linux |All
Severity|normal |minor
--- Comment #1 from Kenji Hara <k.hara...@gmail.com> ---
This is a minor diagnostic issue.
In D, a derived enum type cannot overflow the value range of its base type.
http://dlang.org/enum
> If there is no AssignExpression and it is not the first EnumMember, it is
> given the value of the previous EnumMember+1. If the value of the previous
> EnumMember is EnumBaseType.max, it is an error. If the value of the previous
> EnumMember+1 is the same as the value of the previous EnumMember, it is an
> error. (This can happen with floating point types.)
With the definition of A and B
> enum A {
> start,
> end
> }
>
> enum B {
> start = A.end,
> end
> }
The base type of B is implicitly deduced to A, because B.start is defined by
using A.end.
So B.end is beyond the value of A.max (== A.end), and compiler reports an
overflow error.
But current error message is a little confusing. The current message
> ./test.d(9): Error: enum member test.B.end initialization with (B.start + 1)
> causes overflow for type 'int'
should become:
./test.d(9): Error: enum member test.B.end initialization with (B.start + 1)
causes overflow for type 'A'
----
To avoid the error, you need to specify the base of each enum types explicitly.
You're intended that the two enums are for C library interfaces, so the best
for that would be int.
A fixup example:
enum A : int {
start, // == int(0)
end // == int(1)
}
enum B : int {
start = A.end, // == cast(int)A.end == int(0)
end // == int(1)
}
Then, the B definition never overflows from its base type range: 0 to int.max.
--
