On Sunday, 31 May 2026 at 15:42:32 UTC, Nick Treleaven wrote:
On Sunday, 31 May 2026 at 08:54:39 UTC, Dom Disc wrote:
On Friday, 29 May 2026 at 10:56:51 UTC, Nick Treleaven wrote:
Can you give an example of what you think should work but
doesn't?
```d
byte x = 7;
x = -x; // cannot convert -x of type int to byte
```
At the point of `-x`, the compiler doesn't know that `x !=
x.min`, so VRP correctly prevents this. Do you think the
compiler should allow it? How would it decide that? (If `x` is
const, it can be assigned to a byte).
-byte.min == byte.min (twos complement - also the same result as
if the int-value 128 is truncated to signed eight bit), so yes,
this should be allowed. -x should always have same type as x.
But I would prefer that T.min is NaN of signed types. Instead
byte.min should be -127 (or general T.min == -T.max). This would
also fix abs().
