On 02/23/2012 23:39, Paul Harris wrote:
> If that patch worked, then that looks like undefined behaviour.
What triggered me to think about the change to logical negation turns
out to be what you could consider "a FreeBSD problem" after all. :) I
know that in some operating systems (and I believe all/most Linux'es?)
NULL is defined to be numeric 0, so equating them is safe to do. FreeBSD
doesn't do that, so it isn't safe. (Please note, I'm not trying to start
a "whose interpretation of the C standard is more correct" fight, I'm
just saying that FreeBSD does it differently than some/many others do.)
Meanwhile, I tested that change with and without --dont-restore and it
worked as expected in both cases, so that change should go into the
repo. On a more general note, if there are any other places where 'foo
== 0' is being used where logical negation is intended, especially when
foo can be null, they should also be fixed. A quick grep of the new code
(minus str*cmp of course) indicated too many places for me to trace the
code paths myself, sorry. But changing this would definitely make the
code more robust.
FWIW, modulo some meaningless whitespace changes the code in that line
in 0.92 is the same as it is in 0.95.2, so I'm not sure what happened. I
suspect that some new #include is indirectly pulling in our _null.h.
That's not a bad thing (in fact, it's probably more correct) but the net
effect was a change in behavior. Fortunately it's easily remedied.
hth,
Doug
--
It's always a long day; 86400 doesn't fit into a short.
Breadth of IT experience, and depth of knowledge in the DNS.
Yours for the right price. :) http://SupersetSolutions.com/
--
To unsubscribe, send mail to [email protected].
