------- Additional Comments From baldrick at free dot fr 2005-01-15 12:27 ------- Subject: Re: ACATS CXA4009, CXA4020 - valgrind detects bad code (overlapping memcpy)
> Confirmed, the problem is in gimplify_modify_expr_to_memcpy (but this is not > a regression even > though this is a middle-end problem). Are you sure? The memcpy is already there in the generic tree (based on the dumps), and doesn't gimplification happen later? Also, the memcpy problem goes back to GNAT 3.15p, which is based on gcc 2.8.1. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19419
