https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125524
--- Comment #10 from uecker at gcc dot gnu.org --- There was certainly a reason for the restriction in documentation and I also do not remember the details. So no guarantee that the example will work, but that the increment of the pointer is lost is certainly a bug (the FE should insert the .ACCESS_WITH_SIZE internal function only to reads but not writrres)
