[Bug tree-optimization/100430] False positive for -Warray-bounds and pointers

[msebor at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100430

Martin Sebor <msebor at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |msebor at gcc dot gnu.org

--- Comment #3 from Martin Sebor <msebor at gcc dot gnu.org> ---
At -O3 the call to f() is inlined earlier and the whole if statement in f() is
eliminated before the warning runs.   The store is eliminated because c is
static and nothing reads from it.  When it's made extern the store is retained
and the warning shows up at -O3 too.

At -O2 early inlining fails with:

will not early inline: void h()/3->void f(char*&)/2, growth 5 exceeds --param
early-inlining-insns divided by number of calls
Iterations: 0

Bumping up the early-inlining-insns parameter inlines the function at -O2 and
the warning disappears then too.

All these interactions with the optimizers are inherent in the middle end
warnings because they so heavily depend on them.  The static analyzer can do
better because it doesn't depend on the optimizers.