"Craig A. Berry" <[EMAIL PROTECTED]> wrote on 12/19/2003 09:19:39 AM:
> Peter wrote:
> > >An impressively small patch. It looks good.
>
> Thanks. It would have been a good bit smaller except for the
> following hardware model on the testdrive system:
>
> $ write sys$output f$getsyi("HW_MODEL")
> 0
>
> and, since we had been assuming anything with HW_MODEL <= 1024 was a
> VAX, caused the configuration process to think the Itanium was a VAX.
> I assume they just haven't finished filling out the internal table or
> whatever SYS$GETSYI uses to know system specifics.
OK this is something to keep an eye on then. When I read it I was
wondering if the return values were negative numbers.
> A couple of other changes were necessitated by the fact that the
> compiler now identifies itself as "hp C" rather than "DEC C" or
> "Compaq C" so I had to add checks for that. This is something we'd
> have had to do for Alpha and VAX anyway.
What does CXX/VERSION say on the new arch nowadays? I may want
to push the "-Dtry_cxx" stuff a bit further. Have the __DECC,
__DECC_VER, __DECCXX, and __DECCXX_VER compiler macros been changed
or superseded?
> And then Willem said:
>
> >IMHO VMS_I64 is better: again, 3 characters. I think HP uses the same:
> >
> >VAX - AXP - I64
> >
> >VAX - ALPHA - IA64
>
> Yes, I wondered about that. I went with IA64 because that's what
> f$getsyi("ARCH_NAME") returns, but then we don't follow that on
> Alpha, where it just returns "Alpha" but we use "AXP". Consider the
> exact choice of letters subject to change, and I should have said
> that the initial support of the I-thing by whatever name is as
> experimental and subject to change as early adopter's kit itself.
Hmm, I guess I'll toss my vote in for C<$Config{'archname'} eq 'VMS_I64'>
too; but I am not attached to that string strongly and could easily live
with VMS_IA64 as patched now. I've always been fond of the letter
count symmetry and I used to install in a disk shared by a VAX and
and Alpha (years ago). The choice of archname enters into your @INC
path.
By the way I'd caution against the temptation to call
f$getsyi("ARCH_NAME") anywhere in configure.com itself if it can
be avoided. If I recall correctly that was not
supported in VMS as late a V5.5-2 on VAX. While it was supported
by V6.2 on VAX (maybe V6.0 - but I am not sure) and 1.5 on AXP,
the f$getsyi("HW_MODEL") call is more backward compatible. It might be
worthwhile to ask someone in OpenVMS engineering what the plans are
for HW_MODEL.
Peter Prymmer
