https://issues.dlang.org/show_bug.cgi?id=14467
--- Comment #5 from Steven Schveighoffer <schvei...@yahoo.com> --- Thank you, that is indeed the root cause. I have to investigate why that change was made, as most of the lifetime.d assumes GC.query will get the correct block info. Doing a double-lookup does not sound appealing to me as a fix, we need a function that finds the block info even for interior pointers. At this point, the array runtime is going to be severely broken performance-wise. --