On Thursday, 17 March 2016 at 09:59:41 UTC, Dominikus Dittes
Scherkl wrote:
Or you can use an improved opCmp implementation in the
compiler, that only add additional runtime cost, if someone is
stupid enough to compare signed with unsigned values - but
yield the correct result:
I should also point out that, while it would have been eminently
sensible for D to just implement signed/unsigned comparison
correctly in the first place - fixing this now would be a silent
breaking change.
I'm fairly confident that it would cause real-world problems, as
I myself can recall trying to write code that intentionally
leveraged the current bizarre behavior to eke out a little more
speed, in the past.
Such code could be trivially fixed with the addition of a
`cast(uint)` or the like in the right place, BUT only if someone
actually remembered that it needed to be fixed: neither the
compiler, nor dfix could automatically detect the rare code that
deliberately depends on the current behavior.
So, I think Walter and Andrei made the right call by saying to
make it a warning. Maybe later though, after a *long* deprecation
period, we could fix it to check for negative values like it
should.
