On Tuesday, 11 March 2025 at 10:23:59 UTC, Olivier Pisano wrote:

It's also useful for rounding: n&-16

On Thursday, 13 March 2025 at 07:29:29 UTC, Kagamin wrote:

Again: Please use n&~15 instead. It produces the same result, but without relying on ugly and confusing implicit signed/unsigned conversions.

On Thursday, 13 March 2025 at 07:29:29 UTC, Kagamin wrote:

typeof(16) is int, signed unary minus would not be deprecated.

On Friday, 14 March 2025 at 12:17:59 UTC, Nick Treleaven wrote:

Exactly, I proposed to deprecate unary minus for 16U (uint), not for 16 (int).

writeln(-16U); //writes '4294967280'

On Friday, 14 March 2025 at 16:09:55 UTC, Olivier Pisano wrote:

writeln(-16U); //writes '4294967280'

I have just discovered today a LLVM bug for one platform I'm developing to that it fails to convert correctly long to double:

ulong b = 1_000_000;
double a = b;
writeln(a); //Prints 0.10

I'm generally against disabling implicit conversions. But maybe it could be useful for that situation. That could be under a compilation flag?