4 ms·
Especially interesting because (as one of the compiler team members noted on Zulip) this is a miscompilation relatively easy to stumble into in safe code (yikes
by bayesnet 3mo ago
Especially interesting because (as one of the compiler team members noted on Zulip) this is a miscompilation relatively easy to stumble into in safe code (yikes). Looks like the cause was a late pass in MIR optimization pipeline — I would think these are carefully vetted for soundness so am surprised that this slipped in there.