https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87489
--- Comment #7 from Andrew Church <achurch+gcc at achurch dot org> ---
Would it be reasonable to have the FE warning trigger only on a literal null
value and not on variables whose values are known to be null? I don't know the
history behind -Wnonnull warning at two separate points, but if the FE warning
is intended to cover cases in which (for example) argument order is
inadvertently switched to pass an otherwise-reasonable null to a nonnull
parameter, then not warning on variables doesn't seem like it would make much
of a difference. And I would not expect the compiler to catch data flow bugs
("variable declared as constant null" -> "constant null variable passed to
nonnull function parameter") at -O0 anyway.
