[Bug tree-optimization/100366] [11/12/13 Regression] spurious warning - std::vector::clear followed by std::vector::insert(vec.end(), ...) with -O2

["zeranoe at gmail dot com" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100366

Kyle Schwarz <zeranoe at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |zeranoe at gmail dot com

--- Comment #14 from Kyle Schwarz <zeranoe at gmail dot com> ---
I also ran into this with:
> auto data = std::make_shared<std::vector<uint8_t>>();
> data->insert(data->end(), {0xAA, 0xBB, 0xCC});

I do not hit it with:
> std::vector<uint8_t> data;
> data.insert(data.end(), {0xAA, 0xBB, 0xCC});

Bisecting shows it was introduced with
81d6cdd335ffc60c216a020d5c99306f659377a2. In gcc/gimple-ssa-warn-access.cc
pass_waccess::check() there's now a call to the new
pass_waccess::check_builtin() which throws this warning (BUILT_IN_MEMMOVE,
check_memop_access()). The old behavior had these checks scattered throughout
builtins.c but now with it being part of pass_waccess::check() it is called for
each bb as part of pass_waccess::execute().