[RESEND] [PATCH] Better vxWorks support in longlong.h

[Kevin Ryde]

Mark Haigh <mhaigh at sonicwall.com> writes:
>
> Is that really so terrible?  Like I said, I'm trying to work with you
> here to get something that is acceptable to both the gcc guys and the
> gmp guys.

I don't want to research the perversions of old and new free and
non-free compilers and systems.  In gmp we're going to go with

  #if (HAVE_HOST_CPU_FAMILY_power || HAVE_HOST_CPU_FAMILY_powerpc) \
    && W_TYPE_SIZE == 32

where those HAVE's are setup by configure.  This is sure to be right
and it cuts through what is otherwise messy and uncertain.

On some systems the specs give nice defines you can rely on (eg. ia64)
but alas the power family as far as I can tell is not one of them.

For gcc, longlong.h is a relatively modest optimization as far as I
can see.  Assuming sensible people recompile gcc with itself once up
and running then it only really needs to support its own current
setups, __powerpc__ or whatever.  In gmp we want to go back a bit
further.

-- 
All followups to gmp-bugs at swox.com please.