[Bug middle-end/87296] [8/9 Regression] -Wstringop-overflow false positive due to bogus MEM_REF type

[msebor at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87296

--- Comment #4 from Martin Sebor <msebor at gcc dot gnu.org> ---
The MEM_REF type must mean something, otherwise why have it at all? (If it's
completely meaningless it could/should just be null or void or error_mark_node
instead to prevent using it.)  Is it just that array bounds in the type are
unreliable?  Unless it really is completely useless (in which case we should
make it so) it would be helpful to summarize the conditions when it can safely
be relied on.