Not only is the headline tired, it's wrong per the author's own analysis (emphasis mine): > Finally I checked the optimization flags. The problem seems to be r

Not only is the headline tired, it's wrong per the author's own analysis (emphasis mine): > Finally I checked the optimization flags. The problem seems to be reproducible with code generated without optimization (-Onone) for ARM (because for x86 it is OK). At the same time, result is right with optimized code (-O, -Ounchecked) - this is rather strange to me, I'd expect unoptimized code free of errors like this, rather that optimized one. That is, there's an outright correctness bug in Swift, or at very least a huge break in the principle of least surprise.