On Fri, 4 Sep 1998, David Hartwig wrote:
> Vadim Mikheev wrote:
>
> > src/backend/port/snprintf.c:
> >
> > snprintf.c:84: structure has no member named `_flags'
> > snprintf.c:84: `__SWR' undeclared (first use this function)
> > snprintf.c:84: (Each undeclared identifier is reported only once
> > snprintf.c:84: for each function it appears in.)
> > snprintf.c:84: `__SSTR' undeclared (first use this function)
> > snprintf.c:85: structure has no member named `_bf'
> > snprintf.c:85: structure has no member named `_p'
> > snprintf.c:86: structure has no member named `_bf'
> > snprintf.c:86: structure has no member named `_w'
> > snprintf.c:89: structure has no member named `_p'
> > ^^^^^^^^^
> > This is about FILE structure...
> > BSD code was used for snprintf.c, but
> > FILE in BSD is not the same as in SunOS 5.5.1...
>
> I have the same problem on my AIX box. , Linux includes snprint as in the
> stdio library.
>
> I would prefer to drop snprintf altogether, and replace it with
> sprintf and a sufficient buffer. The places where snprintf is used
> are logically bound to string lengths less than 64 bytes.
snprintf() has become very popular in dealing with buffer overrun
security holes, at least that's where I've seen it most. Solaris 2.6 does
have snprintf, whereas 2.5.1 doesn't...
> Otherwise, snprintf becomes another porting issue. I have no idea how
> to ( or wheather I can) implement this using vprintf. The flags and
> data structures are very different.
I'm grabbing a different snprintf()...just looked through
sendmail's conf.c file, which has an snprintf() implementation in there
that looks more "Generic"...its not GPLd, so is safe to integrate...
Watch the commit messages...
