Xref: utzoo comp.sys.att:7205 unix-pc.general:3497
Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.milw.wisc.edu!bionet!ames!topgun.dspo.gov!lll-winken!gryphon!denwa!srhqla!kosman!kevin
From: kevin@kosman.UUCP (Kevin O'Gorman)
Newsgroups: comp.sys.att,unix-pc.general
Subject: Re: Nethack 3.00 on a Unix PC: assembler complains.
Message-ID: <941@kosman.UUCP>
Date: 4 Aug 89 14:41:30 GMT
References: <1989Jul31.033853.23599@ntvax.uucp> <2151@hub.UUCP>
Reply-To: kevin@kosman.UUCP (Root)
Followup-To: comp.sys.att
Organization: K.O.'s Manor - Vital Computer Systems, Oxnard, CA 93035
Lines: 30

In article <2151@hub.UUCP> todd@ivucsb.sba.ca.us (Todd Day) writes:
>In article <1989Jul31.033853.23599@ntvax.uucp> canoaf@ntvax.uucp (Augustine Cano) writes:
>~While trying to compile Nethack 3.00 on a 3b1 (67Mb HD/2Mb RAM, 3.51a OS)
>~I got the following:
>~
>~	cc -O -I../include -c apply.c
>~Assembler: apply.c
>~aline  359	: branch offset is too remote
>
>Evidentally, the optimizer is forgetting about how far away some things
>are.  I got the same problem.  Compile as much as you can with the optimizer
>(-O flag) and compile the ones that break by hand.  In this case,
>
>cc -I../include -c apply.c
>
>Nethack is the only program I've ever had do this.  Anyone know why this
>happens?

I think so.  The same thing happened to me in compiling the last version
of Nethack 2.0.  I didn't think to fiddle with -O, but took a good look
at what piece of code broke.

Anyway, I decided to try undefining the thing that controls whether the
compiler is supposed to use bitfields in the struct definitions.  When
bitfields were no longer being used, everything compiled with -O just
fine.

This does not explain just what it is that is broken, but it may help
explain why the problem shows up so rarely: not many software packages
use bitfields, so you won't see this too much.