[Bug target/105523] Wrong warning array subscript [0] is outside array bounds

["david at westcontrol dot com" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105523

--- Comment #25 from David Brown <david at westcontrol dot com> ---
(In reply to Andrew Pinski from comment #24)
> (In reply to LIU Hao from comment #22)
> > Yes, GCC should be told to shut up about dereferencing artificial address
> > values.
> 
> NO.
> Take:
> ```
> static inline int f(int *a)
> {
>   return a[10];
> }
> 
> int g()
> {
>   return f(0);
> }
> ```
> The warning is there for the above case really (and similar ones with struct
> offsets). Where you originally have a null pointer and have an offset from
> there; by the time the warning happens, the IR does not know if it was
> originally from an offset of a null pointer or if the value was written in.
> The paramater is there to "tune" the heurstic to figure out if it is null
> pointer deference or if it is some real (HW) address. Maybe
> -fno-delete-null-pointer-checks should imply --param=min-pagesize=0, though
> some folks want the warning indepdent of trying to delete null pointer
> checks.

It is worth noting, I think, that although on a target like the AVR (and most
embedded systems without an MMU) the address 0 is a real part of memory, and
can really be read and/or written, any code that tries to dereference a 0
pointer is almost always wrong.  I don't want gcc to consider 0 as an
acceptable address on these targets - I want it to warn me if it sees a null
pointer dereference.  If I really want to target address 0, as I might
occasionally do, I'll use a pointer to volatile - /then/ I'd like gcc to
believe me without question.

I don't know if every embedded developer feels the same way.  (Georg-Johann
could chime in with his opinion.)