On 7/13/07, Scott McKellar <[EMAIL PROTECTED]> wrote:
--- Mike Rylander <[EMAIL PROTECTED]> wrote:
> BTW, I've attempted to find evidence of a platform where NULL is not
> all-0-bits and failed, though this only exposes my inability to tell
> google what I want to find.
See the C FAQ, question, 5.17:
http://c-faq.com/null/machexamp.html
See also questions 5.16 and 7.31.
This part of the FAQ probably hasn't been updated for over fifteen
years, and the machines mentioned are no doubt museum pieces by now.
I don't know if any current machines represent NULL with non-zeros.
Any such machines are probably far out of the mainstream, and will
likely never attempt to run Evergreen.
Heh ... I'll personally fix any NULL/0 assumptions that cause problems
if someone can get all the dependencies going on a Honeywell-Bull
mainframe. ;)
In all seriousness, though, I have no problem with pedantism.
Changing to malloc() with no memset or to a static array is fine with
me.
UPDATE: changed back to an array. No reason not to, and there is a
provable benefit to speed for a sufficient number of iterations.
Nevertheless I don't like making needlessly non-portable assumptions,
especially when those assumptions are neither obvious nor documented.
You may of course not wish to be as pedantic as I am.
Oh, it's not that all. The trailing NULL is still there, so
portability is maintained, I believe.
--miker
