On Fri, Feb 09, 2018 at 11:40:29AM +0100, Richard Biener wrote:
I.e., having to track all pointers to d between the call to
strncpy and the assignment of the nul and make sure none of
them ends up used in a string function. It didn't seem
the additional complexity would have been worth the effort
(or the likely false negatives).
Well, I'd just walk immediate uses of the VDEF of the
strncpy call, not of the pointer argument. There will be exactly _one_ possible
store (gimple_vdef () is non-NULL) that you need to verify (with using
the current matching
logic). But it'll skip non-store statements for you.
Well, it should also punt on the immediate uses of the VDEF that
have NULL gimple_vdef and the alias oracle says that might alias with that,
i.e. warn about say
strncpy (p, ...);
foo (p);
p[whatever] = '\0';
where foo is pure, because it might read the unterminated string, but don't
warn about
strncpy (p, ...);
x = *q;
p[whatever] = '\0';
if q[0] can't alias with p.
Or just warn if there are any immediate uses of the strncpy VDEF that have
gimple_vdef NULL, or non-NULL and aren't the zero store you are looking for.