Path: utzoo!utgpu!watmath!att!chinet!ignatz
From: ignatz@chinet.chi.il.us (Dave Ihnat)
Newsgroups: unix-pc.general
Subject: Re: Nethack 3.0 compile hassle
Summary: Absolutely wrong; not bitfields, optimization...
Message-ID: <9155@chinet.chi.il.us>
Date: 3 Aug 89 18:38:51 GMT
References: <217@labii.UUCP> <936@kosman.UUCP>
Reply-To: ignatz@chinet.chi.il.us (Dave Ihnat)
Distribution: na
Organization: Chinet - Public Access Unix
Lines: 23

In article <936@kosman.UUCP> kevin@kosman.UUCP (Root) writes:
>I deleted the rest of the message, but you're using the stock cc, right?
>
>You've also got a configuration switch defined in there somewhere that
>turns on the use of bitfields to make structs more compact.
>
>The problem is that in this configuration, cc generates short branch
>instructions for parts of 'if' commands, when it cannot really tell that
>a short branch will get there.
>
>You have two solutions: turn off bitfields, or turn to gcc.

Er...this has been hashed out in other messages, but as this just arrived,
I want to emphasize the answer:  It's Not Bitfields; it's the optimizer.
Keep your bitfields, and turn on every option your heart desires.  But,
for every module that generates one of these messages, either manually
compile it *without* optimization, or make a dependency in the makefile
which does so.  Thereafter, all works well with the stock development system.
(at least for 3.51; I'd guess for 3.5 as well.)

		Dave Ihnat
		ignatz@homebru.chi.il.us (preferred return address)
		ignatz@chinet.chi.il.us