Subject: OpenVMS XDR problem - have you compiled on OpenVMS/Alpha yet?
I don't know what the correct changes to xdrfloat.c should be. Have you
made the changes yet? Before I go diggin into the bowels of Engineering to
solve this one, I just wanted to check with you first.
From: US3RMC::"DOUG@xxxxxxxxxxxxxxxxxxxx " "DOUG" 14-Oct-93 21:06
Subj: Problem with XDR on Alpha under OpenVMS
Thanks for your speedy reply. We checked, and our XDR routines are current
with what is on Internet. And we are working under OpenVMS (V1.5). To be
a bit more specific about what I tried, I modified the xdrfloat.c program
(which has a date in it's header of 87/11/04) to change the
#ifdef vax to #if defined(vax) || defined(__ALPHA) and
#ifndef vax to #if !defined(vax) || !defined(__ALPHA)
When I attempt to compile this, I get the following compiler crash:
$ ccc XDRFLOAT.C
%GEM-F-ASSERTION, An invalid buffer length was specified
%TRACE-F-TRACEBACK, symbolic stack dump follows
Image Name Module Name Routine Name Line Number rel PC abs PC
DECC$COMPILE GEM_ST_IV GEM_ST_STORE_BU 541 00000530 00218C20
DECC$COMPILE GEMC_ME_STATIC_ add_constant_in 19579 000006F0 00153450
DECC$COMPILE INITIAL interpret_sourc 15111 00002744 0011C924
DECC$COMPILE DECLARE interpret_sourc 17870 00004354 00119BE4
DECC$COMPILE FE_MASTER fe_master 14461 00000394 000F6524
DECC$COMPILE GEMC_COMPILE gem_xx_compile 26094 00000310 000E0310
DECC$COMPILE GEM_CP_VMS GEM_CP_MAIN 1502 00000F10 001885B0
0 85EFA8F0 85EFA8F0
I guess the xdrfloat.c routine may be ok (the single precision real number
format is the same between the VAX and AXP)--but unless we can compile it
we can't try it! You mentioned you have someone a Digital who might be able
to shed some light on this? Thanks.
>From owner-netcdfgroup@xxxxxxxxxxxxxxxx 18 Tue, 1991 Jun 20:22:26
Date: Tue, 18 Jun 1991 20:22:26 -0700
From: stew@xxxxxxxxxxxxxxxxxxxx (Stewart Levin)
Subject: Re: netcdf on convex
Received: from oas.Stanford.EDU by unidata.ucar.edu with SMTP id AA06150
(5.65b/IDA-1.4.3 for cat >> /home/ftp/mail-archives/netcdfgroup); Tue, 18 Jun
1991 21:20:53 -0600
Received: from hanauma.stanford.edu by oas.Stanford.EDU (4.1/inc-1.0)
id AA19599; Tue, 18 Jun 1991 20:22:29 PDT
Received: by hanauma.stanford.edu (5.64/7.0)
id AA01106; Tue, 18 Jun 1991 20:22:26 -0700
Fiddled around a bit with the Makefile's and include files last night.
CFLAGS=-g -std -D_POSIX_SOURCE -D_CONVEX_SOURCE were needed and
I had to modify a few #include incantations. Also some places where
#ifdef convex were used, I added __convex__, too. float infinity fill
values pulled from <float.h>. Perhaps one or two other minor things.
link flags were -lrpd into pointer to double and pointer to union containers.
Can someone try this out? I'm willing to help.
hough not the
supplied routines). nctest got past the initial error you ran into, but
fails in the attrget test retrieving -4.5 ... instead of 4.5 ...