On Thu, Sep 18, 2008 at 02:44:24PM -0400, James Carlson wrote: > The spec defines which macros are required, but doesn't prohibit > others. I think it should be fine the way it is ... but it's worth a > pass through the conformance test group.
<SNIP!> I asked Don Cragun early on about this, and he said it's not a concern of X/Open, so we are in the clear w.r.t. bracketting macros. Let me find the note... ahh, here's the relevant passage: ===================== (Cut up to and including here.) ===================== The standards(5) man page correctly states that we implement XNS4, XNS5, and POSIX.1-2001. We have never tried to document a list of standards that we don't try to conform to on that man page, so there is no reason for it to list IEEE Std 1003.1g-2000. POSIX.1-2001 and later POSIX standards do not include AF_LOCAL, PF_LOCAL, AF_FILE, nor PF_FILE; but the namespace reservations in POSIX.1-2001 and later editions/revisions allow implementations to add macros and (where appropriate) structure member names) in <sys/socket.h> starting with AF_, MSG_, PF_, SCM_, SHUT_, SO, cmsg_, if_, ifc_, ifra_, ifru_, infu_, l_, msg_, sa_, and ss_ without hiding them when building applications in standards conforming environments. So, AF_FILE, AF_LOCAL, PF_FILE, and PF_LOCAL can be added to <sys/socket.h> without the #ifdef's that were needed to add timersub() to <sys/time.h>. ===================== (Cut up to and including here.) ===================== So *maybe* SUN_LEN() is still an issue, but the others certainly aren't. Da _______________________________________________ networking-discuss mailing list [email protected]
