AKCL on HP-UX 7.0 B 9000/370

[Joachim Laubsch <laubsch@hpljl.hpl.hp.com>]

Has anyone fixed the bug that prevents successful loading of .o files
produced by compile-file, using HPs latest C/ANSI compiler?

The symptom is: If after successul compilation of a file you try to
load it, you get something like

 > (load "foo.o")
 > Loading foo.o

 > Rel_Info {r_segment = 4, r_symbolnum= 25, r_address = 8} -- Ignored
 > Rel_Info {r_segment = 4, r_symbolnum= 26, r_address = 40} -- Ignored
 > xbin/kcl: 8687 Illegal instruction - core dumped

A workaround (found by Adam Hammer, <hammer@cs.purdue.edu>) is to use
an older version of HPs C Compiler once AKCL is built:

 > "...  I had to tell akcl to use the following C
 > compiler parts before it would compile lisp functions:

 > /usr/local/akcl/cpp:
 >          $Revision: 64.32 $

 > /usr/local/akcl/cpass1:
 >          (20) $Revision: 64.14 $ $Date: 89/08/30 15:11:57 $

 > /usr/local/akcl/cpass2:
 >          (20) $Revision: 64.14 $ $Date: 89/08/30 15:11:57 $

 > ..."


Has anyone found a better solution? 

I suspect it has something to do with a.out file assumptions not being
met in the function COMPILER-BUILD, but have not penetrated this
interesting topic further.

Has anyone built AKCL using GNU's gcc on HP-UX 7.0 B 9000/370?

; -*- Joachim (email: laubsch@hplabs.hp.com tel: 415-857-7695).