genrecog from current cvs tree fails on openVMS/Alpha

genrecog from current cvs tree fails on openVMS/Alpha

[kkaempf]

genrecog from the current snapshot fails on openVMS/Alpha with

	genrecog: Two actions at one point in tree
	after 133 definitions

It took some time for me to figure this one out, it was the following
change that broke genrtl on openVMS/Alpha

Thu Jan 29 10:12:27 1998  Jeffrey A Law  (law@cygnus.com)

        * configure.in: Check for atoq and atoll.
        * rtl.c (read_rtx): Use HAVE_ATOLL and HAVE_ATOQ to select the
        proper routine for converting ascii into long long values.

Since HOST_BITS_PER_INT == HOST_BITS_PER_LONG == 32 and HOST_BITS_PER_WIDE_INT == 64
and neither HAVE_ATOLL nor HAVE_ATOQ are defined (VMS can't run configure!),
tmp_wide wasn't assigned a value in rtl.c:765

A correct fix for the vms-alpha configuration is adding HAVE_ATOLL to alpha/xm-vms.h
but this bug might break other targets also. I would add the following fix which
ensures tmp_wide gets a value:

*** rtl.c.orig  Tue Mar 17 17:07:13 1998
--- rtl.c       Tue Mar 17 17:08:43 1998
***************
*** 773,778 ****
--- 773,780 ----
  #else
  #ifdef HAVE_ATOQ
        tmp_wide = atoq (tmp_char);
+ #else
+       tmp_wide = (HOST_WIDE_INT)atol (tmp_char);
  #endif
  #endif
  #endif

Comments anyone ?

-- 
proGIS Software                 E-Mail: kkaempf@progis.de
Dipl.-Inform. Klaus K"ampf      Fax:    0241-47067-29
Jakobstr. 117                   Voice:  0241-47067-11
D-52064 Aachen                  WWW:	http://www.progis.de

Re: genrecog from current cvs tree fails on openVMS/Alpha

[Richard Henderson]

> A correct fix for the vms-alpha configuration is adding HAVE_ATOLL to
> alpha/xm-vms.h but this bug might break other targets also. I would add
> the following fix which ensures tmp_wide gets a value:

I took the approach of hand-rolling an atoll, since atol would in this
case loose data, which would cause completely different headaches.  I
also comitted the xm-vms.h change.


r~