On 6 September 2016 at 20:15, John Snow <js...@redhat.com> wrote: > Well, I know some have religious arguments against making sanitizers happy, > but that's not my religion. > > If some systems appear to think that unlink must take a nonnull argument, I > think that's a bug -- but making the sanitizer happy doesn't cost us much > either IMO.
POSIX doesn't mandate that unlink() handles NULL, so it's just a quality-of-implementation issue in the library to handle it without barfing (or conversely a QoI issue to over-conservatively mark it as requires-nonnull in the system headers, if you like). In any case my general view with all these sanitizing/lint tools is that it's usually worth taking a few unnecessary-looking detours in order to get to zero-warnings so you can easily flag up when new code triggers a warning (which could well be a bug). > (And it's /just/ a test.) > > So, personally: > > Reviewed-by: John Snow <js...@redhat.com> Thanks! -- PMM
