https://bugs.winehq.org/show_bug.cgi?id=38558
--- Comment #11 from katsunori.kumatani@gmail.com --- Created attachment 61807 --> https://bugs.winehq.org/attachment.cgi?id=61807 Be conservative when dst overlaps into src
This solves the bug when dst overlaps into src from a lower address. The opposite case, when src overlaps into dst from a lower address, is not an issue with the former code path (16 bytes at a time).
Minimal performance impact of only one branch before the loop, should not affect small strings much.