Karim BELABAS on Tue, 18 Jun 2002 09:57:34 +0200 (MEST)

[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: failure of 2.2.3 build on HP-UX

On Mon, 17 Jun 2002, Ilya Zakharevich wrote:

> On Mon, Jun 17, 2002 at 03:23:57PM -0400, Igor Schein wrote:
>> For me, the Configure default on HPUX 11.x 32bit platform is portable kernel.
> Well, as you can see, it is not so for me and 10.20...  All I did was
>   ./Configure --graphic=builtin.X11-gnuplot-dynamic

That's because of the following statement in Configure:

        case "$osname-`uname -r`" in
          hpux-?.09.*)  asmarch=none  ;;
          hpux-?.10.*)  asmarch=$arch ;;
        esac ;;

so we only use assembler on HPUX-10.*. [ because the assembler kernel has
been breaking with each new release of HPUX we could test. It was OK for
HPUX-10.*. I think gcc had problems with it, not sure about that ].

> [...] But why did Configure forgot about it too?  The build is with
> assembler *and* with +z.

I never took the time to fix this properly (mostly because I never had access
again to HPUX). OK, let's try and get it right this time: what I think should
be done for HP is

  If we default to static binary (--static or interactive choice), build
*-sta.o (_without_ -fPIC of +z of course), and use assembler if HPUX version
is OK (see above).

  Otherwise build *-dyn.o with relevant flags, and use portable kernel.

[ it might be a good idea to separate *-sta.o and *-dyn.o for all
architectures, in fact ]


P.S: Btw, Igor, does the assembler kernel work with HPUX 11.x ?
Is the dummy.c hack described in MACHINES still necessary ? If so, we might
as well have it built it  (in e.g systems/hpux.c, #ifdef HPUX_SHL_HACK ?) ...

Karim Belabas                    Tel: (+33) (0)1 69 15 57 48
Dép. de Mathematiques, Bat. 425  Fax: (+33) (0)1 69 15 60 19
Université Paris-Sud             Email: Karim.Belabas@math.u-psud.fr
F-91405 Orsay (France)           http://www.math.u-psud.fr/~belabas
PARI/GP Home Page: http://www.parigp-home.de/