On Mon, Sep 10, 2007 at 09:11:21AM +0000, Pierre Habouzit wrote:
> On Sun, Sep 09, 2007 at 10:30:38PM +0000, Junio C Hamano wrote:
> > Pierre Habouzit <madcoder@xxxxxxxxxx> writes:
> >
> > > @@ -89,14 +86,13 @@ unsigned long interpolate(char *result, unsigned long reslen,
> > > }
> > > }
> > > /* Straight copy one non-interpolation character. */
> > > + if (newlen < reslen)
> > > *dest++ = *src;
> > > src++;
> > > newlen++;
> > > }
> > >
> > > + if (reslen > 0)
> > > + *dest = '\0';
> > > + return newlen;
> > > }
> >
> > Are you sure about this *dest = '\0' assignment? Shouldn't it
> > be done only when "(newlen < reslen)" just like the straight
> > copy and substitution cases?
>
> Actually no, this test is the proper one, I should have kept the
> " + 1" in the previous ones though so that we never append a char to a
I mean the if (newlen < reslen) should read if (newlen + 1 < reslen),
the +1 in the previous test (the one with the memcopy) whould not have
the +1.
The rule is:
if (newlen + WHAT_WE_WANT_TO_ADD < reslen) {
// adding WHAT_WE_WANT_TO_ADD bytes is ok, there will be space
// for the \0
}
In the latter test, WHAT_WE_WANT_TO_ADD is 1, in the previous it's
valuelen and not valuelen + 1.
--
·O· Pierre Habouzit
··O madcoder@xxxxxxxxxx
OOO http://www.madism.org
Attachment:
pgp1YOXkpiEOM.pgp
Description: PGP signature
